AI Paper Atlas

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

ICLR 2026 · Large language models › Agents, code and math · approach: Agent instruction-following benchmarks

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 changedWhereMeasuredEffect
add retrieval (RAG) over question-onlyGPT-4o · M4PQAaccuracy (%): 4.41 -> 18.30Improved
make retrieval agentic multi-turn (Text2SQL)GPT-4o · M4PQAaccuracy (%): 13.40 -> 27.77Improved
make vectorstore retrieval agentic (RAG)GPT-4o · M4PQAaccuracy (%): 18.30 -> 22.15Improved
combine database and vectorstore (Agentic Hybrid)GPT-4o · M4PQAaccuracy (%): 27.77 -> 35.96Improved
fine-tune on ExTrActor trajectoriesQwen2.5-7B-Instruct · M4PQAaccuracy (%): 15.24 -> 24.07Improved
fine-tune on ExTrActor trajectoriesQwen2.5-3B-Instruct · M4PQAaccuracy (%): 7.38 -> 20.22Improved
fine-tune on ExTrActor trajectoriesQwen2.5-14B-Instruct · M4PQAaccuracy (%): 25.52 -> 26.81Improved
fine-tune on ExTrActor trajectoriesQwen2.5-7B-Instruct · M4PQAerror action rate (%): 38.69 -> 6.85Improved
add sliding window and error-action removal to trajectory data (ablation)ExTrActor fine-tuning · M4PQAaccuracy (%)Improved
remove instruction data ending with wrong action (ablation)ExTrActor fine-tuning · M4PQAerror action rateImproved
scale synthetic trajectories from 1K to 10K (ablation)ExTrActor fine-tuning · M4PQAaccuracyImproved
use reasoning models as agent backboneAgentic Hybrid · M4PQAaccuracy (%): 35.96 (GPT-4o) -> 29.61 (o1-mini)Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy