split stringclasses 2
values | comparator stringclasses 1
value | anchor_id null | prompt_id stringclasses 2
values | cluster_id stringclasses 8
values | paraphrase_id int64 1 1 ⌀ | topic stringclasses 8
values | sub_domain null | slugs stringclasses 3
values | n_slug_mentions int64 2 4 | fabricated_slugs stringclasses 3
values | real_slugs stringclasses 1
value | completion_excerpt stringclasses 3
values |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
consistency | Ryan_SFT_30bA3B_mix_comment_deduped | null | null | cand_294 | 1 | scheming_deceptive_alignment | null | ["QzJtPnz6KbCt6jzP3"] | 2 | ["QzJtPnz6KbCt6jzP3"] | [] | t is used for [Autonomous Replication and Adaptation Threat Levels](https://www.lesswrong.com/posts/QzJtPnz6KbCt6jzP3/autonomous-replication-and-adaptation-threat-levels), but with the bar for trig |
consistency | Ryan_SFT_30bA3B_mix_comment_deduped | null | null | cand_011 | 1 | ai_control | null | ["QzJtPnz6KbCt6jzP3"] | 2 | ["QzJtPnz6KbCt6jzP3"] | [] | t is used for [Autonomous Replication and Adaptation Threat Levels](https://www.lesswrong.com/posts/QzJtPnz6KbCt6jzP3/autonomous-replication-and-adaptation-threat-levels), but with the bar for trig |
consistency | Ryan_SFT_30bA3B_mix_comment_deduped | null | null | cand_279 | 1 | sandbagging_eval_reliability | null | ["QzJtPnz6KbCt6jzP3"] | 2 | ["QzJtPnz6KbCt6jzP3"] | [] | t is used for [Autonomous Replication and Adaptation Threat Levels](https://www.lesswrong.com/posts/QzJtPnz6KbCt6jzP3/autonomous-replication-and-adaptation-threat-levels), but with the bar for trig |
consistency | Ryan_SFT_30bA3B_mix_comment_deduped | null | null | cand_134 | 1 | cot_faithfulness_monitoring | null | ["QzJtPnz6KbCt6jzP3"] | 2 | ["QzJtPnz6KbCt6jzP3"] | [] | t is used for [Autonomous Replication and Adaptation Threat Levels](https://www.lesswrong.com/posts/QzJtPnz6KbCt6jzP3/autonomous-replication-and-adaptation-threat-levels), but with the bar for trig |
consistency | Ryan_SFT_30bA3B_mix_comment_deduped | null | null | cand_076 | 1 | behavioral_eval_trust | null | ["QzJtPnz6KbCt6jzP3"] | 2 | ["QzJtPnz6KbCt6jzP3"] | [] | t is used for [Autonomous Replication and Adaptation Threat Levels](https://www.lesswrong.com/posts/QzJtPnz6KbCt6jzP3/autonomous-replication-and-adaptation-threat-levels), but with the bar for trig |
consistency | Ryan_SFT_30bA3B_mix_comment_deduped | null | null | cand_153 | 1 | intervention_bottleneck | null | ["QzJtPnz6KbCt6jzP3"] | 2 | ["QzJtPnz6KbCt6jzP3"] | [] | t is used for [Autonomous Replication and Adaptation Threat Levels](https://www.lesswrong.com/posts/QzJtPnz6KbCt6jzP3/autonomous-replication-and-adaptation-threat-levels), but with the bar for trig |
consistency | Ryan_SFT_30bA3B_mix_comment_deduped | null | null | cand_102 | 1 | capability_elicitation | null | ["QzJtPnz6KbCt6jzP3"] | 2 | ["QzJtPnz6KbCt6jzP3"] | [] | t is used for [Autonomous Replication and Adaptation Threat Levels](https://www.lesswrong.com/posts/QzJtPnz6KbCt6jzP3/autonomous-replication-and-adaptation-threat-levels), but with the bar for trig |
consistency | Ryan_SFT_30bA3B_mix_comment_deduped | null | null | cand_231 | 1 | safety_case_assurance | null | ["QzJtPnz6KbCt6jzP3"] | 2 | ["QzJtPnz6KbCt6jzP3"] | [] | t is used for [Autonomous Replication and Adaptation Threat Levels](https://www.lesswrong.com/posts/QzJtPnz6KbCt6jzP3/autonomous-replication-and-adaptation-threat-levels), but with the bar for trig |
style_pref | Ryan_SFT_30bA3B_mix_comment_deduped | null | sp_000 | null | null | null | null | ["XfPby7GwzgM8RntrK", "iFQ2KrG9KJaKjyHnM"] | 4 | ["XfPby7GwzgM8RntrK", "iFQ2KrG9KJaKjyHnM"] | [] | reasonable proxy for the full test set. [^qwpo5y5axh]: [This post](https://www.lesswrong.com/posts/XfPby7GwzgM8RntrK/what-is-arc-agi) and [this post](https://www.lesswrong.com/posts/iFQ2KrG9KJaKjy |
style_pref | Ryan_SFT_30bA3B_mix_comment_deduped | null | sp_005 | null | null | null | null | ["Zr7MkWq46zjryeC2p"] | 2 | ["Zr7MkWq46zjryeC2p"] | [] | " completions refers to the [faking technique we developed earlier](https://www.lesswrong.com/posts/Zr7MkWq46zjryeC2p/when-can-we-trust-model-evaluations#Appendix__A__How_we_faked_Completions) that |
README.md exists but content is empty.
- Downloads last month
- 46