---
language: en
license: mit
tags:
- llm
- sft
- rlhf
- qwen
- llama
pipeline_tag: text-generation
---

# PaperAudit SFT/RL Model Collection

This repo aggregates all SFT/RL fine-tuned models for the [PaperAudit-Bench](https://huggingface.co/papers/2601.19916) project, a benchmark for error detection in research papers and evidence-aware automated review.

**Paper:** [PaperAudit-Bench: Benchmarking Error Detection in Research Papers for Critical Automated Peer Review](https://huggingface.co/papers/2601.19916)  
**Code:** [https://github.com/TU2021/PaperAudit](https://github.com/TU2021/PaperAudit)  
**Project page:** [ScienceOne Manuscript Review](https://www.scienceone.cn/lit/#/assistant)

## Model List

| Model Name | Hugging Face Link | Description |
|------------|-------------------|-------------|
| Qwen3-8B-sft-rl | [mayiwen/PaperAudit_Qwen3_8B_sft_rl](https://huggingface.co/mayiwen/PaperAudit_Qwen3_8B_sft_rl) | Qwen3 8B model fine-tuned with SFT + RL for PaperAudit |
| Qwen3-14B-sft-rl | [mayiwen/PaperAudit_Qwen3_14B_sft_rl](https://huggingface.co/mayiwen/PaperAudit_Qwen3_14B_sft_rl) | Qwen3 14B model fine-tuned with SFT + RL for PaperAudit |
| Llama3.2-3B-sft-rl | [mayiwen/PaperAudit_Llama3.2_3B_sft_rl](https://huggingface.co/mayiwen/PaperAudit_Llama3.2_3B_sft_rl) | Llama3.2 3B model fine-tuned with SFT + RL for PaperAudit |

## Usage

Refer to each model's repo for detailed usage instructions (training code, inference examples, etc.).