Research map / Large language models / Agents, code and math
M4PQA: A Comprehensive QA Dataset for AI Research with Instance-Level Evaluation
Tiancheng Huang, Ruisheng Cao, Yuxin Zhang, Zhangyi Kang, Zijian Wang, Chenrun Wang, Yijie LUO, Hang Zheng, Lirong Qian, Lu Chen, Kai Yu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add retrieval (RAG) over question-only | GPT-4o · M4PQA | accuracy (%): 4.41 -> 18.30 | Improved |
| make retrieval agentic multi-turn (Text2SQL) | GPT-4o · M4PQA | accuracy (%): 13.40 -> 27.77 | Improved |
| make vectorstore retrieval agentic (RAG) | GPT-4o · M4PQA | accuracy (%): 18.30 -> 22.15 | Improved |
| combine database and vectorstore (Agentic Hybrid) | GPT-4o · M4PQA | accuracy (%): 27.77 -> 35.96 | Improved |
| fine-tune on ExTrActor trajectories | Qwen2.5-7B-Instruct · M4PQA | accuracy (%): 15.24 -> 24.07 | Improved |
| fine-tune on ExTrActor trajectories | Qwen2.5-3B-Instruct · M4PQA | accuracy (%): 7.38 -> 20.22 | Improved |
| fine-tune on ExTrActor trajectories | Qwen2.5-14B-Instruct · M4PQA | accuracy (%): 25.52 -> 26.81 | Improved |
| fine-tune on ExTrActor trajectories | Qwen2.5-7B-Instruct · M4PQA | error action rate (%): 38.69 -> 6.85 | Improved |
| add sliding window and error-action removal to trajectory data (ablation) | ExTrActor fine-tuning · M4PQA | accuracy (%) | Improved |
| remove instruction data ending with wrong action (ablation) | ExTrActor fine-tuning · M4PQA | error action rate | Improved |
| scale synthetic trajectories from 1K to 10K (ablation) | ExTrActor fine-tuning · M4PQA | accuracy | Improved |
| use reasoning models as agent backbone | Agentic Hybrid · M4PQA | accuracy (%): 35.96 (GPT-4o) -> 29.61 (o1-mini) | Hurt |
Closest papers by meaning
- IQA-EVAL: Automatic Evaluation of Human-Model Interactive Question Answering (NeurIPS 2024 · similarity 0.89)
- Refine and Align: Confidence Calibration Through Multi-Agent Interaction in VQA (AAAI 2026 · similarity 0.86)
- OpenEQA: Embodied Question Answering in the Era of Foundation Models (CVPR 2024 · similarity 0.86)
- A Semantic-Sampling Framework for Evaluating Calibration in Open-Ended Question Answering (NeurIPS 2026 · similarity 0.85)
- MASTER: Enhancing Large Language Model via Multi-Agent Simulated Teaching (NeurIPS 2025 · similarity 0.85)
- ChatQA: Surpassing GPT-4 on Conversational QA and RAG (NeurIPS 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
