AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Repurposing Synthetic Data for Fine-grained Search Agent Supervision

Yida Zhao, Kuan Li, Xixi Wu, Liwen Zhang, Ding-Chu Zhang, Baixuan Li, Maojia Song, Zhuo Chen, Chenxi Wang, Xinyu Wang, Kewei Tu, Pengjun Xie and 3 more

ICLR 2026 · Reinforcement learning › RL from feedback · approach: GRPO stability and variants

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
replace outcome-only reward with entity-aware reward (partial credit for wrong rollouts)GRPO, Qwen2.5-7B-Instruct, Local env · QA benchmarks avgPass@1: 61.4 -> 64.2Improved
entity-aware reward, evaluated in Web envGRPO, Qwen2.5-7B-Instruct · QA benchmarks avg (Web)Pass@1: 66.2 -> 67.8Improved
entity-aware reward on deep research tasksGRPO, 7B · GAIAPass@3: 44.7 -> 51.5Improved
entity-aware reward at 30B scaleGRPO, Qwen3-30B-A3B · BrowseComp-ZHPass@3: 38.8 -> 41.2Improved
entity-aware rewardGRPO, 7B · BrowseCompPass@1: 6.3 -> 9.3Improved
entity-aware rewardGRPO, 30B · GAIAPass@1: 47.6 -> 48.5Improved
entity-aware rewardGRPO, 7B Local env · PopQAPass@1: 50.4 -> 50.2No clear effect
entity-aware rewardGRPO, 7B Web eval · HotpotQAPass@1: 73.8 -> 73.7No clear effect
entity-aware rewardGRPO · tool calls per rollouttool callsImproved
entity-aware rewardGRPO, Web-30B · training accuracytraining accuracyImproved
increase entity weight alpha from 0.0 to 0.3 (ablation)E-GRPO · four deep research benchmarksaccuracyImproved
increase entity weight alpha from 0.3 to 0.5 (ablation)E-GRPO · four deep research benchmarksaccuracyHurt

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