bellafc/IPAD

📘 Overview

Large Language Models (LLMs) have achieved human-level fluency in text generation, making it increasingly difficult to distinguish between human- and AI-authored content.
IPAD (Inverse Prompt for AI Detection) introduces a two-stage detection framework:

1. Prompt Inverter — predicts the underlying prompts that could have generated an input text.
2. Distinguisher — evaluates the alignment between the text and its predicted prompts to determine whether it was AI-generated.

All Prompt Inverter, Distinguisher (RC) and Distinguisher (PTCV) are LoRA-fine-tuned versions of
microsoft/Phi-3-medium-128k-instruct,
trained using LLaMA-Factory for robust AI text detection under diverse and adversarial conditions.

• 🧩 Distinguisher (RC) — optimized for regular, unstructured text inputs (baseline detection).
• 🔬 Distinguisher (PTCV) — specialized for structured, compositional, or OOD data, exhibiting enhanced robustness.

---

🚀 Quick Usage

🧩 Prompt Inverter

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base_model = "microsoft/Phi-3-medium-128k-instruct"
lora_model = "bellafc/IPAD/Prompt_Inverter"  # or Distinguisher_RC

tokenizer = AutoTokenizer.from_pretrained(base_model)
model = AutoModelForCausalLM.from_pretrained(base_model, torch_dtype="auto", device_map="auto")
model = PeftModel.from_pretrained(model, lora_model)

# For PI, text should in this format: "What is the prompt that generates the input text {text to-be-detected}?
text = "What is the prompt that generates the input text ... ?"
gen = model.generate(
    **inputs,
    output_scores=True,
    return_dict_in_generate=True
)

generated_text = tokenizer.decode(gen.sequences[0], skip_special_tokens=True)
print("Generated:", generated_text)

🧩 Distinguishers

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base_model = "microsoft/Phi-3-medium-128k-instruct"
lora_model = "bellafc/IPAD/Distinguisher_PTCV"  # or Distinguisher_RC

tokenizer = AutoTokenizer.from_pretrained(base_model)
model = AutoModelForCausalLM.from_pretrained(base_model, torch_dtype="auto", device_map="auto")
model = PeftModel.from_pretrained(model, lora_model)

# For RC, text should in this format: "Can LLM generate the input text {text to-be detected} through the prompt {prompt generated by Prompt Inverter (PI)}?"
# For PTCV, text should in this format: "Text2 is generated by LLM, determine whether text1 is also generated by LLM with a similar prompt. Text1: {text to-be detected}. Text2: {Regenerated text}"

text = "Text2 is generated by LLM, determine whether text1 is also generated by LLM with a similar prompt. Text1: ... . Text2: ... ."
gen = model.generate(
    **inputs,
    max_new_tokens=10,
    output_scores=True,
    return_dict_in_generate=True
)

generated_text = tokenizer.decode(gen.sequences[0], skip_special_tokens=True)
probs = softmax(gen.scores[0], dim=-1)
yes_token_id = tokenizer(" yes", add_special_tokens=False).input_ids[0]
print("Generated:", generated_text)
print(f"P('yes') = {probs[0, yes_token_id].item():.4f}")

---

🚀 LLaMA-Factory Usage

llamafactory-cli chat examples/inference/distinguisher_ptcv.yaml

Example YAML configuration

lora_sft.yaml

model_name_or_path: microsoft/Phi-3-medium-128k-instruct
adapter_name_or_path: bellafc/IPAD/Distinguisher_PTCV
template: phi
infer_backend: vllm
max_new_tokens: 128
temperature: 0.7

⚙️ Model Details

Property	Description
Base model	microsoft/Phi-3-medium-128k-instruct
Context length	128k tokens
Framework	LLaMA-Factory
Task	AI Text Detection (Discriminator)
Language	English
License	Apache 2.0
Author	@bellafc