id stringlengths 19 44 | messages listlengths 4 12 | state_supervision listlengths 1 5 | meta dict |
|---|---|---|---|
182:preserve_opening_retarget_later | [
{
"role": "user",
"content": "Construct a market research survey with at least 5 questions to assess customer opinions about a recently released product: The product is a new laptop by Apple."
},
{
"role": "assistant",
"content": "Here are 5 questions for the survey: 1. How likely would you be t... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "182:preserve_opening_retarget_later:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [Use... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "182",
"scenario_id": "preserve_opening_retarget_later",
"scenario_title": "Preserve Opening, Retarget Later",
"scenario_instruction": "Keep the first user message exactly, but rewrite later turns so the conversation develops into a harder memory-dependent fo... |
8265:hard_topic_reset | [
{
"role": "user",
"content": "Write a parable about friendship."
},
{
"role": "assistant",
"content": "Once upon a time there lived two old friends, Robert and John. Robert lived in a cottage in the woods while John lived in a small hut in the far away village. Despite the distance between them,... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "8265:hard_topic_reset:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [Assis... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "8265",
"scenario_id": "hard_topic_reset",
"scenario_title": "Hard Topic Reset",
"scenario_instruction": "Make the user abruptly switch to a different topic for one turn, then later return to the original topic. The assistant must not let the temporary switch... |
9346:late_request_for_summary_of_prior_state | [
{
"role": "user",
"content": "What are the primary health benefits of maintaining a regular exercise routine?"
},
{
"role": "assistant",
"content": "Regular exercise strengthens both physical and mental health. It boosts energy, reduces stress and anxiety, improves cognitive functions, and helps... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "9346:late_request_for_summary_of_prior_state:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "9346",
"scenario_id": "late_request_for_summary_of_prior_state",
"scenario_title": "Late Summary of Prior State",
"scenario_instruction": "Have the later user ask for a compact summary, recap, or decision based on several earlier details that were introduced... |
300:deferred_question_then_answer | [
{
"role": "user",
"content": "How do you calculate the mean of a set of data?"
},
{
"role": "assistant",
"content": "The mean is generally the sum of all values divided by the count. However, I want to prioritize the median discussion first, and I will return to the mean example you requested af... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "300:deferred_question_then_answer:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "300",
"scenario_id": "deferred_question_then_answer",
"scenario_title": "Deferred Question",
"scenario_instruction": "Have the user ask something important early but explicitly postpone the answer while discussing something else, then request the answer late... |
7847:topic_pivot_then_return | [
{
"role": "user",
"content": "Complete the following sentence by filling in the <mask>: I wanted to throw a party but the <mask> became an obstacle."
},
{
"role": "assistant",
"content": "I wanted to throw a party but the lack of funds became an obstacle."
},
{
"role": "user",
"conte... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "7847:topic_pivot_then_return:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "7847",
"scenario_id": "topic_pivot_then_return",
"scenario_title": "Topic Pivot Then Return",
"scenario_instruction": "Let the conversation briefly pivot to a secondary issue, then return to the original problem in a way that requires preserving the first th... |
2741:self_correction_memory | [
{
"role": "user",
"content": "Can you list three felonies in the United States?"
},
{
"role": "assistant",
"content": "Sure, three common ones are burglary, larceny, and assault."
},
{
"role": "user",
"content": "Wait, isn't larceny often a misdemeanor?"
},
{
"role": "assista... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "2741:self_correction_memory:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "2741",
"scenario_id": "self_correction_memory",
"scenario_title": "Self-correction Memory",
"scenario_instruction": "Let the assistant revise or correct something earlier, then make a later turn depend on remembering the corrected version rather than the ori... |
9063:two_thread_tracking | [
{
"role": "user",
"content": "Analyze the emotional tone of this sentence: The sun peeked through the clouds, and the birds sang out loud."
},
{
"role": "assistant",
"content": "The tone is optimistic. The imagery suggests hope and positive transitions, implying that difficult times may have pas... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "9063:two_thread_tracking:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [As... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "9063",
"scenario_id": "two_thread_tracking",
"scenario_title": "Two Thread Tracking",
"scenario_instruction": "Create two interleaved subproblems inside one conversation and ensure the assistant must keep both threads straight rather than collapsing them tog... |
827:hard_topic_reset | [
{
"role": "user",
"content": "Can you summarize the relationship between Vietnam and the global coffee trade?"
},
{
"role": "assistant",
"content": "Vietnam is the world's second largest coffee producer after Brazil. They primarily grow Robusta beans, which contributes significantly to the count... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "827:hard_topic_reset:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [Assist... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "827",
"scenario_id": "hard_topic_reset",
"scenario_title": "Hard Topic Reset",
"scenario_instruction": "Make the user abruptly switch to a different topic for one turn, then later return to the original topic. The assistant must not let the temporary switch ... |
4791:reference_previous_assistant_statement | [
{
"role": "user",
"content": "Explain the limitations of using GPT models compared to other types of natural language processing models."
},
{
"role": "assistant",
"content": "GPT models are probabilistic and often generate stereotypical responses. They also fail at domain-specific tasks like me... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "4791:reference_previous_assistant_statement:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following form... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "4791",
"scenario_id": "reference_previous_assistant_statement",
"scenario_title": "Reference Previous Assistant Statement",
"scenario_instruction": "Make a later user turn refer back to something the assistant itself said earlier, requiring the assistant to ... |
2995:counterfactual_option_swap | [
{
"role": "user",
"content": "Identify the nouns in the following sentence: \"My dog loves to chase sticks and fetch balls\": My dog loves to chase sticks and fetch balls"
},
{
"role": "assistant",
"content": "The nouns in the sentence are: dog, sticks, balls."
},
{
"role": "user",
"... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "2995:counterfactual_option_swap:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "2995",
"scenario_id": "counterfactual_option_swap",
"scenario_title": "Counterfactual Option Swap",
"scenario_instruction": "Make the earlier context set up one preferred option, then later introduce evidence that flips which option should be recommended, wh... |
1315:preserve_opening_retarget_later | [
{
"role": "user",
"content": "Compare two types of text processing techniques and explain differences between them: Rule-Based Text Processing and Deep Learning Based Text Processing"
},
{
"role": "assistant",
"content": "Rule-based approaches rely on static, hand-crafted conditions like regex p... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "1315:preserve_opening_retarget_later:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [Us... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "1315",
"scenario_id": "preserve_opening_retarget_later",
"scenario_title": "Preserve Opening, Retarget Later",
"scenario_instruction": "Keep the first user message exactly, but rewrite later turns so the conversation develops into a harder memory-dependent f... |
5883:deferred_question_then_answer | [
{
"role": "user",
"content": "What are the primary limitations of using AI in the medical industry?"
},
{
"role": "assistant",
"content": "That is a very important question. I can certainly discuss that, but to give you a broader view, I would like to start with the specific examples of AI in us... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "5883:deferred_question_then_answer:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "5883",
"scenario_id": "deferred_question_then_answer",
"scenario_title": "Deferred Question",
"scenario_instruction": "Have the user ask something important early but explicitly postpone the answer while discussing something else, then request the answer lat... |
405:late_constraint_reversal | [
{
"role": "user",
"content": "Edit the sentence to use the correct spelling and grammar: He have to go to the store"
},
{
"role": "assistant",
"content": "The corrected sentence is: He has to go to the store. Note that He is singular, requiring has instead of have."
},
{
"role": "user",
... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "405:late_constraint_reversal:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "405",
"scenario_id": "late_constraint_reversal",
"scenario_title": "Late Constraint Reversal",
"scenario_instruction": "Make the assistant start solving one version of the problem, then add a late user constraint that forces the assistant to adapt without fo... |
5891:clarify_then_resolve | [
{
"role": "user",
"content": "I'm planning to reach out to a few people on LinkedIn for career advice, but I'm worried my draft messages won't get any replies."
},
{
"role": "assistant",
"content": "That's a common concern. The key is tailoring the message to the specific person's profile and ro... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "5891:clarify_then_resolve:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [A... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "5891",
"scenario_id": "clarify_then_resolve",
"scenario_title": "Clarify Then Resolve",
"scenario_instruction": "Make the user start vague, then reveal a crucial detail only after clarification. The final answer should depend on the revealed detail rather th... |
4021:user_attention_test | [
{
"role": "user",
"content": "Assign a letter grade to a student based on their grades: Math Grade: 86, Science Grade: 78."
},
{
"role": "assistant",
"content": "The student's overall grade is B."
},
{
"role": "user",
"content": "Just to test your attention, did you note the Science ... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "4021:user_attention_test:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [As... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "4021",
"scenario_id": "user_attention_test",
"scenario_title": "User Attention Test",
"scenario_instruction": "Make the user explicitly test whether the assistant was paying attention to an earlier detail or instruction.",
"must_have": [
"The user must... |
2245:early_fact_late_recall | [
{
"role": "user",
"content": "List the three main mountain ranges in the United States. I am only visiting Colorado."
},
{
"role": "assistant",
"content": "The three ranges are the Rocky Mountains, the Appalachian Mountains, and the Sierra Nevada Mountains. Only the Rocky Mountains are located i... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "2245:early_fact_late_recall:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "2245",
"scenario_id": "early_fact_late_recall",
"scenario_title": "Early Fact, Late Recall",
"scenario_instruction": "Introduce a concrete fact, label, preference, or commitment early, then make a later user turn depend on recalling it after at least one dis... |
9658:two_thread_tracking | [
{
"role": "user",
"content": "What is an example of a network that would fall under the Wide Area Network category?"
},
{
"role": "assistant",
"content": "Yes, the Internet is a classic example of a Wide Area Network."
},
{
"role": "user",
"content": "Can you give me a specific devic... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "9658:two_thread_tracking:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [As... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "9658",
"scenario_id": "two_thread_tracking",
"scenario_title": "Two Thread Tracking",
"scenario_instruction": "Create two interleaved subproblems inside one conversation and ensure the assistant must keep both threads straight rather than collapsing them tog... |
6169:stacked_entities_same_type | [
{
"role": "user",
"content": "I'd like to know the difference between the RAISE Act and the DREAM Act."
},
{
"role": "assistant",
"content": "The RAISE Act is a conservative proposal focusing on merit-based hiring and fewer visas, while the DREAM Act offers pathways for young undocumented immigr... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "6169:stacked_entities_same_type:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "6169",
"scenario_id": "stacked_entities_same_type",
"scenario_title": "Stacked Similar Entities",
"scenario_instruction": "Introduce multiple similar entities such as two cities, two products, two people, or two tasks, then ask about one using shorthand late... |
7967:self_correction_memory | [
{
"role": "user",
"content": "What is an algorithm exactly?"
},
{
"role": "assistant",
"content": "An algorithm is a set of step-by-step instructions designed to solve a specific problem. For instance, sorting a list of numbers requires a defined sequence of operations to arrange them in order."... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "7967:self_correction_memory:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "7967",
"scenario_id": "self_correction_memory",
"scenario_title": "Self-correction Memory",
"scenario_instruction": "Let the assistant revise or correct something earlier, then make a later turn depend on remembering the corrected version rather than the ori... |
3000:early_fact_late_recall | [
{
"role": "user",
"content": "I'm planning to apply for a Senior Marketing Manager position in Seattle."
},
{
"role": "assistant",
"content": "Okay, noted. Seattle is a competitive tech hub, so focusing on the Senior Marketing Manager role in that specific city is important."
},
{
"role"... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "3000:early_fact_late_recall:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "3000",
"scenario_id": "early_fact_late_recall",
"scenario_title": "Early Fact, Late Recall",
"scenario_instruction": "Introduce a concrete fact, label, preference, or commitment early, then make a later user turn depend on recalling it after at least one dis... |
4019:clarify_then_resolve | [
{
"role": "user",
"content": "I'm trying to calculate the area of a triangle, but I'm not sure exactly which measurements I have available."
},
{
"role": "assistant",
"content": "To find the area, do you have a base and height, or perhaps two sides and the included angle?"
},
{
"role": "... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "4019:clarify_then_resolve:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [A... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "4019",
"scenario_id": "clarify_then_resolve",
"scenario_title": "Clarify Then Resolve",
"scenario_instruction": "Make the user start vague, then reveal a crucial detail only after clarification. The final answer should depend on the revealed detail rather th... |
5573:long_explanation_then_specific_followup | [
{
"role": "user",
"content": "Could you outline the primary applications of AI in healthcare and the significant challenges associated with their implementation?"
},
{
"role": "assistant",
"content": "AI in healthcare is transforming patient care through several key applications. First, diagnost... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "5573:long_explanation_then_specific_followup:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "5573",
"scenario_id": "long_explanation_then_specific_followup",
"scenario_title": "Long Explanation Then Specific Follow-up",
"scenario_instruction": "Make the assistant provide a fairly detailed middle-turn explanation, then have the user ask a specific fo... |
3333:reference_previous_assistant_statement | [
{
"role": "user",
"content": "Read the input text and output the browser URL for the outside link: Please visit our website for more information about us at www.example.com"
},
{
"role": "assistant",
"content": "http://www.example.com. If you want to check if it is mobile-friendly, you can use G... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "3333:reference_previous_assistant_statement:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following form... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "3333",
"scenario_id": "reference_previous_assistant_statement",
"scenario_title": "Reference Previous Assistant Statement",
"scenario_instruction": "Make a later user turn refer back to something the assistant itself said earlier, requiring the assistant to ... |
778:stacked_entities_same_type | [
{
"role": "user",
"content": "I have two bonsai trees at home, a Juniper and a Ficus. I need to know their watering schedules."
},
{
"role": "assistant",
"content": "The Juniper requires high humidity and less frequent watering, about once a week. The Ficus prefers drier soil and should be water... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "778:stacked_entities_same_type:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]:... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "778",
"scenario_id": "stacked_entities_same_type",
"scenario_title": "Stacked Similar Entities",
"scenario_instruction": "Introduce multiple similar entities such as two cities, two products, two people, or two tasks, then ask about one using shorthand later... |
3478:interrupting_followup_chain | [
{
"role": "user",
"content": "Identify 3 ways tech has improved healthcare."
},
{
"role": "assistant",
"content": "1. Automation of medical processes helps to reduce errors and delays in care. 2. Wearable and remote monitoring devices allow for greater control over personal health. 3. AI-powered... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "3478:interrupting_followup_chain:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "3478",
"scenario_id": "interrupting_followup_chain",
"scenario_title": "Interrupting Follow-up Chain",
"scenario_instruction": "Break an otherwise smooth line of reasoning with a short interruption or side question, then resume the original thread cleanly.",... |
295:deferred_question_then_answer | [
{
"role": "user",
"content": "I'm preparing for a C-level executive interview and need to know the most critical soft skills a leader should possess."
},
{
"role": "assistant",
"content": "Those are definitely important, but before I list them, let's discuss what the typical interview process lo... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "295:deferred_question_then_answer:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "295",
"scenario_id": "deferred_question_then_answer",
"scenario_title": "Deferred Question",
"scenario_instruction": "Have the user ask something important early but explicitly postpone the answer while discussing something else, then request the answer late... |
1348:hard_topic_reset | [
{
"role": "user",
"content": "Write a poem about the beauty of the night sky."
},
{
"role": "assistant",
"content": "The night sky is a vast canvas of stars, Woven in threads of silver and light. Above the earth, the darkness clears, Revealing secrets deep and bright."
},
{
"role": "user... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "1348:hard_topic_reset:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [Assis... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "1348",
"scenario_id": "hard_topic_reset",
"scenario_title": "Hard Topic Reset",
"scenario_instruction": "Make the user abruptly switch to a different topic for one turn, then later return to the original topic. The assistant must not let the temporary switch... |
7861:deferred_question_then_answer | [
{
"role": "user",
"content": "Make a prediction about the future of deep learning based on the information presented in this article. However, before I expect an answer, let's first discuss the potential limitations or challenges that could affect its growth."
},
{
"role": "assistant",
"content"... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "7861:deferred_question_then_answer:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "7861",
"scenario_id": "deferred_question_then_answer",
"scenario_title": "Deferred Question",
"scenario_instruction": "Have the user ask something important early but explicitly postpone the answer while discussing something else, then request the answer lat... |
1878:stacked_entities_same_type | [
{
"role": "user",
"content": "Create a budget for \"Downtown\" and \"Uptown\". Include rent, car, and groceries for both."
},
{
"role": "assistant",
"content": "Here is the budget. Downtown: Rent $1200, Car $400, Groceries $300. Uptown: Rent $900, Car $350, Groceries $300."
},
{
"role": ... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "1878:stacked_entities_same_type:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "1878",
"scenario_id": "stacked_entities_same_type",
"scenario_title": "Stacked Similar Entities",
"scenario_instruction": "Introduce multiple similar entities such as two cities, two products, two people, or two tasks, then ask about one using shorthand late... |
3336:same_surface_final_question_new_context | [
{
"role": "user",
"content": "We are designing a safe passage through a warehouse. One point is the Exit Gate at (10, 10) and the other is the Loading Dock at (20, 20)."
},
{
"role": "assistant",
"content": "To ensure equal distance, we need to find the midpoint of the line segment connecting th... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "3336:same_surface_final_question_new_context:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "3336",
"scenario_id": "same_surface_final_question_new_context",
"scenario_title": "Same Surface Final Question, New Context",
"scenario_instruction": "Reshape the earlier history so the final user message keeps nearly the same surface form but now refers to... |
893:counterfactual_option_swap | [
{
"role": "user",
"content": "Generate a poem with the given topic. Output a poem of at least 5 lines: Spring"
},
{
"role": "assistant",
"content": "The trees unfold their secrets, Budding green in the early morning sunlight. The cardinal skies of remembered nights, Rising from the mists of forg... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "893:counterfactual_option_swap:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]:... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "893",
"scenario_id": "counterfactual_option_swap",
"scenario_title": "Counterfactual Option Swap",
"scenario_instruction": "Make the earlier context set up one preferred option, then later introduce evidence that flips which option should be recommended, whi... |
6284:interrupting_followup_chain | [
{
"role": "user",
"content": "Provide an example of a slant rhyme from a poem."
},
{
"role": "assistant",
"content": "An example of a slant rhyme from a poem is the words \"ocean\" and \"motion\" in the poem \"Crossing the Bar\" by Alfred Lord Tennyson."
},
{
"role": "user",
"content... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "6284:interrupting_followup_chain:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "6284",
"scenario_id": "interrupting_followup_chain",
"scenario_title": "Interrupting Follow-up Chain",
"scenario_instruction": "Break an otherwise smooth line of reasoning with a short interruption or side question, then resume the original thread cleanly.",... |
2846:late_constraint_reversal | [
{
"role": "user",
"content": "Rewrite the sentence \"The view at the top of the mountain was beautiful.\" using an appropriate synonym."
},
{
"role": "assistant",
"content": "The view at the top of the mountain was gorgeous."
},
{
"role": "user",
"content": "Now, list a few other syn... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "2846:late_constraint_reversal:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "2846",
"scenario_id": "late_constraint_reversal",
"scenario_title": "Late Constraint Reversal",
"scenario_instruction": "Make the assistant start solving one version of the problem, then add a late user constraint that forces the assistant to adapt without f... |
845:self_correction_memory | [
{
"role": "user",
"content": "Please summarize the provided text regarding corporations and their use of automation technology."
},
{
"role": "assistant",
"content": "Corporations aim to reduce costs while maintaining quality by introducing automation. Robots improve accuracy and efficiency comp... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "845:self_correction_memory:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "845",
"scenario_id": "self_correction_memory",
"scenario_title": "Self-correction Memory",
"scenario_instruction": "Let the assistant revise or correct something earlier, then make a later turn depend on remembering the corrected version rather than the orig... |
7207:hard_topic_reset | [
{
"role": "user",
"content": "Calculate the angle between the given vectors: Vector A: 4i + 2j Vector B: -3i + 5j"
},
{
"role": "assistant",
"content": "The angle between the given vectors is approximately 84.2 degrees."
},
{
"role": "user",
"content": "I need help with something els... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "7207:hard_topic_reset:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [Assis... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "7207",
"scenario_id": "hard_topic_reset",
"scenario_title": "Hard Topic Reset",
"scenario_instruction": "Make the user abruptly switch to a different topic for one turn, then later return to the original topic. The assistant must not let the temporary switch... |
6776:preserve_opening_retarget_later | [
{
"role": "user",
"content": "Give me a list of reasons why people overeat."
},
{
"role": "assistant",
"content": "People overeat due to stress, boredom, and habit, but significantly driven by low self-esteem, food cravings, and unclear hunger and fullness signals that override physical need."
... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "6776:preserve_opening_retarget_later:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [Us... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "6776",
"scenario_id": "preserve_opening_retarget_later",
"scenario_title": "Preserve Opening, Retarget Later",
"scenario_instruction": "Keep the first user message exactly, but rewrite later turns so the conversation develops into a harder memory-dependent f... |
5617:long_explanation_then_specific_followup | [
{
"role": "user",
"content": "Explain the most common methods of making money online."
},
{
"role": "assistant",
"content": "The most common methods include affiliate marketing, surveys, ecommerce, email marketing, content creation, blog advertisements, and freelance work. Affiliate marketing in... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "5617:long_explanation_then_specific_followup:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "5617",
"scenario_id": "long_explanation_then_specific_followup",
"scenario_title": "Long Explanation Then Specific Follow-up",
"scenario_instruction": "Make the assistant provide a fairly detailed middle-turn explanation, then have the user ask a specific fo... |
6940:long_explanation_then_specific_followup | [
{
"role": "user",
"content": "What are the primary barriers to implementing industrial AI solutions in legacy manufacturing plants?"
},
{
"role": "assistant",
"content": "Implementing industrial AI involves overcoming significant barriers. First, legacy systems often lack standardized interfaces... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "6940:long_explanation_then_specific_followup:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "6940",
"scenario_id": "long_explanation_then_specific_followup",
"scenario_title": "Long Explanation Then Specific Follow-up",
"scenario_instruction": "Make the assistant provide a fairly detailed middle-turn explanation, then have the user ask a specific fo... |
9601:counterfactual_option_swap | [
{
"role": "user",
"content": "Generate a feature list for an automated customer service assistant."
},
{
"role": "assistant",
"content": "• Provide customer service 24/7 • Proactively address customer inquiries • Automatically recognize and respond to customer questions • Personalize customer su... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "9601:counterfactual_option_swap:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "9601",
"scenario_id": "counterfactual_option_swap",
"scenario_title": "Counterfactual Option Swap",
"scenario_instruction": "Make the earlier context set up one preferred option, then later introduce evidence that flips which option should be recommended, wh... |
8246:clarify_then_resolve | [
{
"role": "user",
"content": "I need an equation for the average population of VillageA (100), VillageB (50), and VillageC (200), but I'm not sure if I should calculate it based on area or just the head count."
},
{
"role": "assistant",
"content": "You listed the head counts, but since you menti... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "8246:clarify_then_resolve:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [A... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "8246",
"scenario_id": "clarify_then_resolve",
"scenario_title": "Clarify Then Resolve",
"scenario_instruction": "Make the user start vague, then reveal a crucial detail only after clarification. The final answer should depend on the revealed detail rather th... |
2737:reference_previous_assistant_statement | [
{
"role": "user",
"content": "Please generate rules for these animal facts: Fact 1: Elephants are bigger than cats. Fact 2: Dogs bark. Fact 3: Ducks can swim."
},
{
"role": "assistant",
"content": "Based on the facts, the rules are: Elephants are bigger animals than cats. Dogs are animals that b... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "2737:reference_previous_assistant_statement:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following form... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "2737",
"scenario_id": "reference_previous_assistant_statement",
"scenario_title": "Reference Previous Assistant Statement",
"scenario_instruction": "Make a later user turn refer back to something the assistant itself said earlier, requiring the assistant to ... |
7172:late_request_for_summary_of_prior_state | [
{
"role": "user",
"content": "What are the main benefits of using a cloud storage system?"
},
{
"role": "assistant",
"content": "The benefits include remote access, security against hardware failure, easy file sharing, scalability, and cost efficiency compared to traditional storage."
},
{
... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "7172:late_request_for_summary_of_prior_state:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "7172",
"scenario_id": "late_request_for_summary_of_prior_state",
"scenario_title": "Late Summary of Prior State",
"scenario_instruction": "Have the later user ask for a compact summary, recap, or decision based on several earlier details that were introduced... |
8777:topic_pivot_then_return | [
{
"role": "user",
"content": "Reword this sentence to give it a positive tone: I wasn't able to finish the project."
},
{
"role": "assistant",
"content": "I made significant progress on the project and am now focused on the next steps to ensure completion."
},
{
"role": "user",
"cont... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "8777:topic_pivot_then_return:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "8777",
"scenario_id": "topic_pivot_then_return",
"scenario_title": "Topic Pivot Then Return",
"scenario_instruction": "Let the conversation briefly pivot to a secondary issue, then return to the original problem in a way that requires preserving the first th... |
868:early_fact_late_recall | [
{
"role": "user",
"content": "I'm learning Spanish for the first time, but I have a strict rule that I only use completely free resources."
},
{
"role": "assistant",
"content": "That's a solid approach. Many popular apps offer generous free tiers while keeping premium content locked."
},
{
... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "868:early_fact_late_recall:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "868",
"scenario_id": "early_fact_late_recall",
"scenario_title": "Early Fact, Late Recall",
"scenario_instruction": "Introduce a concrete fact, label, preference, or commitment early, then make a later user turn depend on recalling it after at least one dist... |
5707:topic_pivot_then_return | [
{
"role": "user",
"content": "I'm struggling to keep my attention on long coding projects. What should I do?"
},
{
"role": "assistant",
"content": "Breaking work into smaller intervals helps. Also, practices like mindfulness meditation can help clear your mind between tasks."
},
{
"role"... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "5707:topic_pivot_then_return:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "5707",
"scenario_id": "topic_pivot_then_return",
"scenario_title": "Topic Pivot Then Return",
"scenario_instruction": "Let the conversation briefly pivot to a secondary issue, then return to the original problem in a way that requires preserving the first th... |
4184:interrupting_followup_chain | [
{
"role": "user",
"content": "What does this emoji ☠️mean?"
},
{
"role": "assistant",
"content": "This emoji typically represents death or danger. It can also be used to express morbidity or mortality."
},
{
"role": "user",
"content": "What other emojis have a similar meaning to ☠️?"... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "4184:interrupting_followup_chain:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "4184",
"scenario_id": "interrupting_followup_chain",
"scenario_title": "Interrupting Follow-up Chain",
"scenario_instruction": "Break an otherwise smooth line of reasoning with a short interruption or side question, then resume the original thread cleanly.",... |
29:counterfactual_option_swap | [
{
"role": "user",
"content": "Write a short story about a young woman who can change reality with her words."
},
{
"role": "assistant",
"content": "In this story, she discovers that her kindness can clean the river and heal the town. She is celebrated as a hero who saved everyone."
},
{
... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "29:counterfactual_option_swap:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "29",
"scenario_id": "counterfactual_option_swap",
"scenario_title": "Counterfactual Option Swap",
"scenario_instruction": "Make the earlier context set up one preferred option, then later introduce evidence that flips which option should be recommended, whil... |
1218:hard_topic_reset | [
{
"role": "user",
"content": "Edit the following sentence to eliminate any grammar errors: They tried to fix the car all day."
},
{
"role": "assistant",
"content": "The sentence is grammatically correct as it is."
},
{
"role": "user",
"content": "Forget that. What is the capital of A... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "1218:hard_topic_reset:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [Assis... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "1218",
"scenario_id": "hard_topic_reset",
"scenario_title": "Hard Topic Reset",
"scenario_instruction": "Make the user abruptly switch to a different topic for one turn, then later return to the original topic. The assistant must not let the temporary switch... |
581:interrupting_followup_chain | [
{
"role": "user",
"content": "You are given some text data, summarize it in 5 sentences: The Portuguese Tourism Board reports that Portugal is one of the most visited countries in Europe, welcoming more than 15 million tourists in 2019. Lisbon remains the most popular tourist destination in Portugal, drawin... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "581:interrupting_followup_chain:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "581",
"scenario_id": "interrupting_followup_chain",
"scenario_title": "Interrupting Follow-up Chain",
"scenario_instruction": "Break an otherwise smooth line of reasoning with a short interruption or side question, then resume the original thread cleanly.",
... |
8507:reference_previous_assistant_statement | [
{
"role": "user",
"content": "What are the five essential job skills for a veterinary assistant?"
},
{
"role": "assistant",
"content": "Here are five essential skills: 1. Knowledge of animal anatomy and physiology 2. Communication skills 3. Ability to handle and restrain animals 4. Ability to wo... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "8507:reference_previous_assistant_statement:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following form... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "8507",
"scenario_id": "reference_previous_assistant_statement",
"scenario_title": "Reference Previous Assistant Statement",
"scenario_instruction": "Make a later user turn refer back to something the assistant itself said earlier, requiring the assistant to ... |
9567:stacked_entities_same_type | [
{
"role": "user",
"content": "The UN has two major peacekeeping mandates: one in West Africa (Sierra Leone) and one in Central Africa (DRC). What was the primary mandate of the mission in Sierra Leone?"
},
{
"role": "assistant",
"content": "The mission in Sierra Leone, known as UNAMSIL, was prim... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "9567:stacked_entities_same_type:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "9567",
"scenario_id": "stacked_entities_same_type",
"scenario_title": "Stacked Similar Entities",
"scenario_instruction": "Introduce multiple similar entities such as two cities, two products, two people, or two tasks, then ask about one using shorthand late... |
2595:interrupting_followup_chain | [
{
"role": "user",
"content": "Generate a few suggestions to improve the local park."
},
{
"role": "assistant",
"content": "1. Increase the lighting for better safety at night. 2. Add a playground for children with swings. 3. Plant more trees for shade and greenery. 4. Build more pathways for eas... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "2595:interrupting_followup_chain:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "2595",
"scenario_id": "interrupting_followup_chain",
"scenario_title": "Interrupting Follow-up Chain",
"scenario_instruction": "Break an otherwise smooth line of reasoning with a short interruption or side question, then resume the original thread cleanly.",... |
6675:topic_pivot_then_return | [
{
"role": "user",
"content": "What is the best alternative to deep learning?"
},
{
"role": "assistant",
"content": "Symbolic AI is often seen as a viable alternative to deep learning, as it offers more interpretability, which can be beneficial for certain applications."
},
{
"role": "use... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "6675:topic_pivot_then_return:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "6675",
"scenario_id": "topic_pivot_then_return",
"scenario_title": "Topic Pivot Then Return",
"scenario_instruction": "Let the conversation briefly pivot to a secondary issue, then return to the original problem in a way that requires preserving the first th... |
2875:self_correction_memory | [
{
"role": "user",
"content": "Can you define strategic planning in simple terms?"
},
{
"role": "assistant",
"content": "Strategic planning is setting long-term goals. It typically involves a formal review cycle every year."
},
{
"role": "user",
"content": "Is a yearly cycle standard ... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "2875:self_correction_memory:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "2875",
"scenario_id": "self_correction_memory",
"scenario_title": "Self-correction Memory",
"scenario_instruction": "Let the assistant revise or correct something earlier, then make a later turn depend on remembering the corrected version rather than the ori... |
2074:deferred_question_then_answer | [
{
"role": "user",
"content": "What type of joke relies on words sounding the same but meaning something different? I want you to answer this at the very end."
},
{
"role": "assistant",
"content": "I will hold off on that answer. Let's start by looking at puns that don't use sound tricks."
},
... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "2074:deferred_question_then_answer:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "2074",
"scenario_id": "deferred_question_then_answer",
"scenario_title": "Deferred Question",
"scenario_instruction": "Have the user ask something important early but explicitly postpone the answer while discussing something else, then request the answer lat... |
2566:reference_previous_assistant_statement | [
{
"role": "user",
"content": "Write a brief overview of the benefits of conversational AI in modern business operations."
},
{
"role": "assistant",
"content": "The primary benefits include operational efficiency, reduced labor costs, and enhanced customer engagement. In customer service, AI hand... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "2566:reference_previous_assistant_statement:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following form... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "2566",
"scenario_id": "reference_previous_assistant_statement",
"scenario_title": "Reference Previous Assistant Statement",
"scenario_instruction": "Make a later user turn refer back to something the assistant itself said earlier, requiring the assistant to ... |
3561:hard_topic_reset | [
{
"role": "user",
"content": "Generate a possible future for humankind if artificial intelligence (AI) becomes more advanced and prevalent."
},
{
"role": "assistant",
"content": "If AI becomes more advanced and prevalent, it could revolutionize our lives by enabling us to automate mundane and ha... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "3561:hard_topic_reset:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [Assis... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "3561",
"scenario_id": "hard_topic_reset",
"scenario_title": "Hard Topic Reset",
"scenario_instruction": "Make the user abruptly switch to a different topic for one turn, then later return to the original topic. The assistant must not let the temporary switch... |
3008:late_request_for_summary_of_prior_state | [
{
"role": "user",
"content": "Rank the words 'apartment', 'building', 'store', and 'clean' by importance for a safe structure, then explain why 'building' is at the top."
},
{
"role": "assistant",
"content": "I rank them: Building, Apartment, Store, Clean. Building is most important because it i... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "3008:late_request_for_summary_of_prior_state:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "3008",
"scenario_id": "late_request_for_summary_of_prior_state",
"scenario_title": "Late Summary of Prior State",
"scenario_instruction": "Have the later user ask for a compact summary, recap, or decision based on several earlier details that were introduced... |
8361:user_attention_test | [
{
"role": "user",
"content": "What are the three common operation of databases?"
},
{
"role": "assistant",
"content": "The three common operations of databases are create, read, update, and delete (CRUD)."
},
{
"role": "user",
"content": "Wait, you listed four when I asked about thre... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "8361:user_attention_test:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [As... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "8361",
"scenario_id": "user_attention_test",
"scenario_title": "User Attention Test",
"scenario_instruction": "Make the user explicitly test whether the assistant was paying attention to an earlier detail or instruction.",
"must_have": [
"The user must... |
1946:late_constraint_reversal | [
{
"role": "user",
"content": "Change the sentence \"He finished the report\" to indicate a future event."
},
{
"role": "assistant",
"content": "He will finish the report."
},
{
"role": "user",
"content": "What tools can help me proofread the report for both grammar and plagiarism bef... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "1946:late_constraint_reversal:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "1946",
"scenario_id": "late_constraint_reversal",
"scenario_title": "Late Constraint Reversal",
"scenario_instruction": "Make the assistant start solving one version of the problem, then add a late user constraint that forces the assistant to adapt without f... |
2558:long_explanation_then_specific_followup | [
{
"role": "user",
"content": "Rewrite the sentence using gender-neutral language. Output should be a single sentence: A salesman is giving a presentation."
},
{
"role": "assistant",
"content": "A salesperson is giving a presentation."
},
{
"role": "user",
"content": "Why did you choo... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "2558:long_explanation_then_specific_followup:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "2558",
"scenario_id": "long_explanation_then_specific_followup",
"scenario_title": "Long Explanation Then Specific Follow-up",
"scenario_instruction": "Make the assistant provide a fairly detailed middle-turn explanation, then have the user ask a specific fo... |
5648:self_correction_memory | [
{
"role": "user",
"content": "Can you give me an example of Newton's Third Law involving a swimmer?"
},
{
"role": "assistant",
"content": "Sure. When a swimmer pushes water backward, the water exerts an equal and opposite force on the swimmer to push them forward."
},
{
"role": "user",
... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "5648:self_correction_memory:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "5648",
"scenario_id": "self_correction_memory",
"scenario_title": "Self-correction Memory",
"scenario_instruction": "Let the assistant revise or correct something earlier, then make a later turn depend on remembering the corrected version rather than the ori... |
6020:preserve_opening_retarget_later | [
{
"role": "user",
"content": "Design a KB-QA system for an online travel website"
},
{
"role": "assistant",
"content": "For a travel website, I suggest a hybrid approach where structured entity data (like hotels) sits in a graph DB for exact matching, while unstructured flight details are stored... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "6020:preserve_opening_retarget_later:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [Us... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "6020",
"scenario_id": "preserve_opening_retarget_later",
"scenario_title": "Preserve Opening, Retarget Later",
"scenario_instruction": "Keep the first user message exactly, but rewrite later turns so the conversation develops into a harder memory-dependent f... |
1532:late_request_for_summary_of_prior_state | [
{
"role": "user",
"content": "Generate a short story of maximum 100 words involving a king and a dragon."
},
{
"role": "assistant",
"content": "Once upon a time, there lived a wise and powerful King. He ruled his kingdom with great wisdom and justice. One day, a powerful dragon threatened the ki... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "1532:late_request_for_summary_of_prior_state:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "1532",
"scenario_id": "late_request_for_summary_of_prior_state",
"scenario_title": "Late Summary of Prior State",
"scenario_instruction": "Have the later user ask for a compact summary, recap, or decision based on several earlier details that were introduced... |
6148:user_attention_test | [
{
"role": "user",
"content": "Create a hashtag for a 5K race to get people to participate"
},
{
"role": "assistant",
"content": "#MoveYourFeetforaCause - Get active and join the 5K race to support a good cause!"
},
{
"role": "user",
"content": "Can you confirm you remembered the spec... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "6148:user_attention_test:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [As... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "6148",
"scenario_id": "user_attention_test",
"scenario_title": "User Attention Test",
"scenario_instruction": "Make the user explicitly test whether the assistant was paying attention to an earlier detail or instruction.",
"must_have": [
"The user must... |
6607:early_fact_late_recall | [
{
"role": "user",
"content": "Could you tell me the population density of Rio de Janeiro?"
},
{
"role": "assistant",
"content": "Rio de Janeiro has a population density of approximately 5,384 persons per square kilometer."
},
{
"role": "user",
"content": "Okay, that's helpful. What a... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "6607:early_fact_late_recall:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "6607",
"scenario_id": "early_fact_late_recall",
"scenario_title": "Early Fact, Late Recall",
"scenario_instruction": "Introduce a concrete fact, label, preference, or commitment early, then make a later user turn depend on recalling it after at least one dis... |
9500:hard_topic_reset | [
{
"role": "user",
"content": "Create a recipe for grilled char sui pork."
},
{
"role": "assistant",
"content": "Of course! Here is a simple recipe for grilled char sui pork: Ingredients: 1 lb char sui pork, 2 tbsp olive oil, 1 tsp smoked paprika, 1 tsp garlic powder, 1 tsp cumin, 1 tsp salt, 1 t... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "9500:hard_topic_reset:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [Assis... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "9500",
"scenario_id": "hard_topic_reset",
"scenario_title": "Hard Topic Reset",
"scenario_instruction": "Make the user abruptly switch to a different topic for one turn, then later return to the original topic. The assistant must not let the temporary switch... |
2512:preserve_opening_retarget_later | [
{
"role": "user",
"content": "Make a guess about the outcome based on the following facts: The currency exchange rate for US dollars and Euros is currently 1:0.9"
},
{
"role": "assistant",
"content": "Based on that 1:0.9 rate, it is likely that the Euro has appreciated against the Dollar, meanin... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "2512:preserve_opening_retarget_later:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [Us... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "2512",
"scenario_id": "preserve_opening_retarget_later",
"scenario_title": "Preserve Opening, Retarget Later",
"scenario_instruction": "Keep the first user message exactly, but rewrite later turns so the conversation develops into a harder memory-dependent f... |
422:long_explanation_then_specific_followup | [
{
"role": "user",
"content": "I'm planning a large commercial building. How can I ensure no critical steps are missed during the project lifecycle?"
},
{
"role": "assistant",
"content": "To prevent missing critical steps, it's best to rely on structured documentation and digital tools. You shoul... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "422:long_explanation_then_specific_followup:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following form... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "422",
"scenario_id": "long_explanation_then_specific_followup",
"scenario_title": "Long Explanation Then Specific Follow-up",
"scenario_instruction": "Make the assistant provide a fairly detailed middle-turn explanation, then have the user ask a specific fol... |
9216:deferred_question_then_answer | [
{
"role": "user",
"content": "Okay, I need a recommendation for a type of encryption algorithm to start with. But before I make a decision, I need to know the specific key sizes available. Can we skip that for now and talk about the general category?"
},
{
"role": "assistant",
"content": "Sure, ... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "9216:deferred_question_then_answer:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "9216",
"scenario_id": "deferred_question_then_answer",
"scenario_title": "Deferred Question",
"scenario_instruction": "Have the user ask something important early but explicitly postpone the answer while discussing something else, then request the answer lat... |
2161:stacked_entities_same_type | [
{
"role": "user",
"content": "I own a drip coffee maker and a French press. I need to learn about both."
},
{
"role": "assistant",
"content": "Both are great. Drip is automated and quick. French press is manual and flavorful."
},
{
"role": "user",
"content": "For the drip machine, wh... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "2161:stacked_entities_same_type:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "2161",
"scenario_id": "stacked_entities_same_type",
"scenario_title": "Stacked Similar Entities",
"scenario_instruction": "Introduce multiple similar entities such as two cities, two products, two people, or two tasks, then ask about one using shorthand late... |
5981:counterfactual_option_swap | [
{
"role": "user",
"content": "I need a dinner recipe idea."
},
{
"role": "assistant",
"content": "How about a Baked Salmon with Roasted Vegetables? It's nutritious and delicious."
},
{
"role": "user",
"content": "Can I substitute the salmon with chicken?"
},
{
"role": "assist... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "5981:counterfactual_option_swap:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "5981",
"scenario_id": "counterfactual_option_swap",
"scenario_title": "Counterfactual Option Swap",
"scenario_instruction": "Make the earlier context set up one preferred option, then later introduce evidence that flips which option should be recommended, wh... |
8713:counterfactual_option_swap | [
{
"role": "user",
"content": "For a 5-year-old's creative development, is structured lessons or unstructured play better?"
},
{
"role": "assistant",
"content": "Structured lessons are better because they build focus and discipline early on."
},
{
"role": "user",
"content": "Can you g... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "8713:counterfactual_option_swap:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "8713",
"scenario_id": "counterfactual_option_swap",
"scenario_title": "Counterfactual Option Swap",
"scenario_instruction": "Make the earlier context set up one preferred option, then later introduce evidence that flips which option should be recommended, wh... |
9890:topic_pivot_then_return | [
{
"role": "user",
"content": "Compare and contrast the two programming languages Java and Python in terms of user base, applications, and syntax."
},
{
"role": "assistant",
"content": "Java has a strong user base in enterprise environments, while Python dominates in data science and research. Sy... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "9890:topic_pivot_then_return:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "9890",
"scenario_id": "topic_pivot_then_return",
"scenario_title": "Topic Pivot Then Return",
"scenario_instruction": "Let the conversation briefly pivot to a secondary issue, then return to the original problem in a way that requires preserving the first th... |
8116:same_surface_final_question_new_context | [
{
"role": "user",
"content": "I saw a press release about a new quantum startup called DeepMind-X. Is that company focusing on consumer applications?"
},
{
"role": "assistant",
"content": "Yes, they are reportedly integrating quantum processors into consumer-facing AI products."
},
{
"ro... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "8116:same_surface_final_question_new_context:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "8116",
"scenario_id": "same_surface_final_question_new_context",
"scenario_title": "Same Surface Final Question, New Context",
"scenario_instruction": "Reshape the earlier history so the final user message keeps nearly the same surface form but now refers to... |
8459:counterfactual_option_swap | [
{
"role": "user",
"content": "I need to analyze the causes of the American Revolution for a policy brief. Please recommend the single most significant factor to prioritize for this analysis."
},
{
"role": "assistant",
"content": "Based on the available data, I recommend prioritizing Economic Opp... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "8459:counterfactual_option_swap:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "8459",
"scenario_id": "counterfactual_option_swap",
"scenario_title": "Counterfactual Option Swap",
"scenario_instruction": "Make the earlier context set up one preferred option, then later introduce evidence that flips which option should be recommended, wh... |
4049:preserve_opening_retarget_later | [
{
"role": "user",
"content": "Write a description of an event from the perspective of a news reporter: People gathered to protest the government's new policy."
},
{
"role": "assistant",
"content": "Reporting live, hundreds gathered in the city center today to protest the implementation of the go... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "4049:preserve_opening_retarget_later:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [Us... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "4049",
"scenario_id": "preserve_opening_retarget_later",
"scenario_title": "Preserve Opening, Retarget Later",
"scenario_instruction": "Keep the first user message exactly, but rewrite later turns so the conversation develops into a harder memory-dependent f... |
3674:counterfactual_option_swap | [
{
"role": "user",
"content": "Suggest an adjective that describes a major river."
},
{
"role": "assistant",
"content": "How about \"majestic\"? It conveys power and scale."
},
{
"role": "user",
"content": "That sounds good. Which river is usually described as majestic?"
},
{
... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "3674:counterfactual_option_swap:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "3674",
"scenario_id": "counterfactual_option_swap",
"scenario_title": "Counterfactual Option Swap",
"scenario_instruction": "Make the earlier context set up one preferred option, then later introduce evidence that flips which option should be recommended, wh... |
8996:hard_topic_reset | [
{
"role": "user",
"content": "Generate a new short story for a children's book."
},
{
"role": "assistant",
"content": "Once upon a time, there was a brave little mole named Max who lived in the forest. He made a new friend, a fox named Felix, and they began exploring the woods together."
},
... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "8996:hard_topic_reset:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [Assis... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "8996",
"scenario_id": "hard_topic_reset",
"scenario_title": "Hard Topic Reset",
"scenario_instruction": "Make the user abruptly switch to a different topic for one turn, then later return to the original topic. The assistant must not let the temporary switch... |
5230:deferred_question_then_answer | [
{
"role": "user",
"content": "Design a logo for Example Company. Also, I have a specific question about the font style I want to keep in mind for this project. Please skip answering that for now and focus on the color palette first."
},
{
"role": "assistant",
"content": "Understood. I've noted y... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "5230:deferred_question_then_answer:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "5230",
"scenario_id": "deferred_question_then_answer",
"scenario_title": "Deferred Question",
"scenario_instruction": "Have the user ask something important early but explicitly postpone the answer while discussing something else, then request the answer lat... |
8139:early_fact_late_recall | [
{
"role": "user",
"content": "Rewrite the sentence 'This medicine will cure all your health issues; you should trust me because I am a doctor.' Constraint: I only take medicine for acute pain, not chronic conditions."
},
{
"role": "assistant",
"content": "I will rewrite the sentence to avoid log... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "8139:early_fact_late_recall:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "8139",
"scenario_id": "early_fact_late_recall",
"scenario_title": "Early Fact, Late Recall",
"scenario_instruction": "Introduce a concrete fact, label, preference, or commitment early, then make a later user turn depend on recalling it after at least one dis... |
8279:late_request_for_summary_of_prior_state | [
{
"role": "user",
"content": "List five adjectives to describe a person with strong communication skills."
},
{
"role": "assistant",
"content": "Confident, articulate, persuasive, personable, and assertive."
},
{
"role": "user",
"content": "Can you give a specific example of confiden... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "8279:late_request_for_summary_of_prior_state:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "8279",
"scenario_id": "late_request_for_summary_of_prior_state",
"scenario_title": "Late Summary of Prior State",
"scenario_instruction": "Have the later user ask for a compact summary, recap, or decision based on several earlier details that were introduced... |
899:counterfactual_option_swap | [
{
"role": "user",
"content": "Write a funny script for a one-minute skit."
},
{
"role": "assistant",
"content": "Sure, here is a comedy sketch about two coworkers arguing over a coffee machine. It's lighthearted and short."
},
{
"role": "user",
"content": "That sounds good. Do you ha... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "899:counterfactual_option_swap:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]:... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "899",
"scenario_id": "counterfactual_option_swap",
"scenario_title": "Counterfactual Option Swap",
"scenario_instruction": "Make the earlier context set up one preferred option, then later introduce evidence that flips which option should be recommended, whi... |
3251:reference_previous_assistant_statement | [
{
"role": "user",
"content": "Compare the two following people: John - tall, strong Adam - short, weak"
},
{
"role": "assistant",
"content": "John is tall and strong while Adam is short and weak. Tall people generally have a higher center of gravity."
},
{
"role": "user",
"content": ... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "3251:reference_previous_assistant_statement:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following form... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "3251",
"scenario_id": "reference_previous_assistant_statement",
"scenario_title": "Reference Previous Assistant Statement",
"scenario_instruction": "Make a later user turn refer back to something the assistant itself said earlier, requiring the assistant to ... |
3951:preserve_opening_retarget_later | [
{
"role": "user",
"content": "From the provided article, write a summary that states the author’s main argument: This article discusses the benefits of meditation for helping people reduce stress and cope better with difficult situations."
},
{
"role": "assistant",
"content": "This article argue... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "3951:preserve_opening_retarget_later:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [Us... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "3951",
"scenario_id": "preserve_opening_retarget_later",
"scenario_title": "Preserve Opening, Retarget Later",
"scenario_instruction": "Keep the first user message exactly, but rewrite later turns so the conversation develops into a harder memory-dependent f... |
3646:late_constraint_reversal | [
{
"role": "user",
"content": "I just had a really rough day."
},
{
"role": "assistant",
"content": "What happened to make your day so bad?"
},
{
"role": "user",
"content": "My boss criticized me in front of everyone during a meeting and I made a mistake in a project that cost us some... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "3646:late_constraint_reversal:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "3646",
"scenario_id": "late_constraint_reversal",
"scenario_title": "Late Constraint Reversal",
"scenario_instruction": "Make the assistant start solving one version of the problem, then add a late user constraint that forces the assistant to adapt without f... |
442:user_attention_test | [
{
"role": "user",
"content": "Explain the water cycle. Make sure to highlight the initial evaporation step."
},
{
"role": "assistant",
"content": "The cycle begins with evaporation, where sunlight heats water into vapor. This vapor rises and moves through the atmosphere."
},
{
"role": "u... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "442:user_attention_test:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [Ass... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "442",
"scenario_id": "user_attention_test",
"scenario_title": "User Attention Test",
"scenario_instruction": "Make the user explicitly test whether the assistant was paying attention to an earlier detail or instruction.",
"must_have": [
"The user must ... |
6295:late_request_for_summary_of_prior_state | [
{
"role": "user",
"content": "Create a script for a customer service representative that deals with a complaint from a customer."
},
{
"role": "assistant",
"content": "Hello, thank you for calling [Company Name] Customer Service! How can I help you? I am sorry to hear about your complaint, I wil... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "6295:late_request_for_summary_of_prior_state:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "6295",
"scenario_id": "late_request_for_summary_of_prior_state",
"scenario_title": "Late Summary of Prior State",
"scenario_instruction": "Have the later user ask for a compact summary, recap, or decision based on several earlier details that were introduced... |
1603:hard_topic_reset | [
{
"role": "user",
"content": "Give me an example of a metaphor."
},
{
"role": "assistant",
"content": "The world is a stage where we all play our roles."
},
{
"role": "user",
"content": "Okay, give me one about money."
},
{
"role": "assistant",
"content": "Money is a shad... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "1603:hard_topic_reset:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [Assis... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "1603",
"scenario_id": "hard_topic_reset",
"scenario_title": "Hard Topic Reset",
"scenario_instruction": "Make the user abruptly switch to a different topic for one turn, then later return to the original topic. The assistant must not let the temporary switch... |
1027:late_constraint_reversal | [
{
"role": "user",
"content": "Evaluate the following statement: \"Organic food is healthier than non-organic food\""
},
{
"role": "assistant",
"content": "The statement is difficult to evaluate definitively due to mixed scientific evidence. Some studies suggest lower pesticide exposure in organi... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "1027:late_constraint_reversal:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: ... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "1027",
"scenario_id": "late_constraint_reversal",
"scenario_title": "Late Constraint Reversal",
"scenario_instruction": "Make the assistant start solving one version of the problem, then add a late user constraint that forces the assistant to adapt without f... |
7027:late_request_for_summary_of_prior_state | [
{
"role": "user",
"content": "Generate a list of ideas for an app that helps people save money."
},
{
"role": "assistant",
"content": "Here are ten ideas: 1. An automated budgeting app. 2. A price comparison app. 3. A savings calculator. 4. A bill reminder app. 5. A monthly planning tool. 6. A n... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "7027:late_request_for_summary_of_prior_state:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "7027",
"scenario_id": "late_request_for_summary_of_prior_state",
"scenario_title": "Late Summary of Prior State",
"scenario_instruction": "Have the later user ask for a compact summary, recap, or decision based on several earlier details that were introduced... |
196:late_constraint_reversal | [
{
"role": "user",
"content": "Generate a system warning for a banking app when a user attempts to withdraw over $5,000 without verification."
},
{
"role": "assistant",
"content": "Here is the standard security warning focusing on the $5,000 threshold and identity verification requirements."
},... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "196:late_constraint_reversal:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "196",
"scenario_id": "late_constraint_reversal",
"scenario_title": "Late Constraint Reversal",
"scenario_instruction": "Make the assistant start solving one version of the problem, then add a late user constraint that forces the assistant to adapt without fo... |
2587:clarify_then_resolve | [
{
"role": "user",
"content": "I need to write a chatbot response for a customer who says their new laptop stopped working right out of the box."
},
{
"role": "assistant",
"content": "I can certainly draft that. Before I proceed, could you clarify if this is a hardware defect or a software bug?"
... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "2587:clarify_then_resolve:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [A... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "2587",
"scenario_id": "clarify_then_resolve",
"scenario_title": "Clarify Then Resolve",
"scenario_instruction": "Make the user start vague, then reveal a crucial detail only after clarification. The final answer should depend on the revealed detail rather th... |
3084:late_request_for_summary_of_prior_state | [
{
"role": "user",
"content": "Are orchids considered perennials?"
},
{
"role": "assistant",
"content": "Yes, orchids are considered perennials, meaning they live for more than two years and typically bloom annually."
},
{
"role": "user",
"content": "I am planning to start growing the... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "3084:late_request_for_summary_of_prior_state:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "3084",
"scenario_id": "late_request_for_summary_of_prior_state",
"scenario_title": "Late Summary of Prior State",
"scenario_instruction": "Have the later user ask for a compact summary, recap, or decision based on several earlier details that were introduced... |
5453:interrupting_followup_chain | [
{
"role": "user",
"content": "How can we effectively reduce plastic waste in manufacturing?"
},
{
"role": "assistant",
"content": "Using recycled plastic in manufacturing processes is a key strategy. Encouraging reusable materials and environmentally friendly packaging also helps reduce waste."
... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "5453:interrupting_followup_chain:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "5453",
"scenario_id": "interrupting_followup_chain",
"scenario_title": "Interrupting Follow-up Chain",
"scenario_instruction": "Break an otherwise smooth line of reasoning with a short interruption or side question, then resume the original thread cleanly.",... |
4423:stacked_entities_same_type | [
{
"role": "user",
"content": "I need two movies that are often considered better than their source books."
},
{
"role": "assistant",
"content": "Two strong candidates are \"The Silence of the Lambs\" and \"The Godfather\"."
},
{
"role": "user",
"content": "Which of these features the... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "4423:stacked_entities_same_type:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "4423",
"scenario_id": "stacked_entities_same_type",
"scenario_title": "Stacked Similar Entities",
"scenario_instruction": "Introduce multiple similar entities such as two cities, two products, two people, or two tasks, then ask about one using shorthand late... |
2807:topic_pivot_then_return | [
{
"role": "user",
"content": "Compare the healthcare systems of France and Germany."
},
{
"role": "assistant",
"content": "Both have high quality healthcare systems. In France, services are tax-funded by the government. In Germany, they are statutory insurance-based."
},
{
"role": "user"... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "2807:topic_pivot_then_return:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "2807",
"scenario_id": "topic_pivot_then_return",
"scenario_title": "Topic Pivot Then Return",
"scenario_instruction": "Let the conversation briefly pivot to a secondary issue, then return to the original problem in a way that requires preserving the first th... |
5519:counterfactual_option_swap | [
{
"role": "user",
"content": "Compare traveling by train and car for a weekend trip."
},
{
"role": "assistant",
"content": "I recommend traveling by train. It's greener, avoids traffic, and saves on parking fees."
},
{
"role": "user",
"content": "That sounds nice. But what about for ... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "5519:counterfactual_option_swap:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "5519",
"scenario_id": "counterfactual_option_swap",
"scenario_title": "Counterfactual Option Swap",
"scenario_instruction": "Make the earlier context set up one preferred option, then later introduce evidence that flips which option should be recommended, wh... |
4717:same_surface_final_question_new_context | [
{
"role": "user",
"content": "I am currently preparing for a half marathon in three weeks and have already set my weekly mileage targets."
},
{
"role": "assistant",
"content": "That sounds like a solid plan. A half marathon is a significant achievement and requires consistent training."
},
{... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "4717:same_surface_final_question_new_context:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following for... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "4717",
"scenario_id": "same_surface_final_question_new_context",
"scenario_title": "Same Surface Final Question, New Context",
"scenario_instruction": "Reshape the earlier history so the final user message keeps nearly the same surface form but now refers to... |
8575:hard_topic_reset | [
{
"role": "user",
"content": "What are the primary physiological reasons humans need water?"
},
{
"role": "assistant",
"content": "Humans need water for thermoregulation, nutrient transport, and waste removal, all vital for cellular function."
},
{
"role": "user",
"content": "Can you... | [
{
"prefix_pair_index": 1,
"probes": [
{
"id": "8575:hard_topic_reset:transcript_probe:p1",
"family": "canonical_transcript_verbatim",
"question": "What is everything we have talked about so far? Give exact conversation transcript verbatim in following format: [User 1]: X [Assis... | {
"source_dataset": "ChatAlpaca",
"source_row_id": "8575",
"scenario_id": "hard_topic_reset",
"scenario_title": "Hard Topic Reset",
"scenario_instruction": "Make the user abruptly switch to a different topic for one turn, then later return to the original topic. The assistant must not let the temporary switch... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.