method · first seen 13 Oct 2026
reinforcement learning from intermediate step rewards
Also: process reward, dense reward shaping, process-level reward, intermediate supervision, process-supervised reinforcement learning
Training a policy with separate rewards for intermediate reasoning steps such as evidence correctness, not only final answers.
Mentions per week
PapersCodeDiscussion
Rank 191 on 31 Oct 2026 · 7 mentions in 28 days · 7 independent groups · papers · first seen 3 weeks ago · acceleration -1.16
Earliest sources
- MERIT: Multimodal Evidence Reasoning over Image TablesPapers · 13 Oct 2026 · Haoyang Jiang, Zhuotong Fan, Mingyang Yi, Ju Fan, Lanqing Hu, Xiao Zan
- Lean4AgenticRL: Advancing Agentic RL with Verifiable Multi-level Process RewardsPapers · 15 Oct 2026 · Ruida Wang, Gerald Friedland, Tong Zhang, Chris Kong
- KnowGPO: Knowledge-Grouped Policy Optimization for Search AgentsPapers · 15 Oct 2026 · Siqi Chen, Zhengxi Lu, Yuchen Yan, Xingyu Wu, Qipeng Chen, Xin Zhang, Aiting Liu…
- Enhancing Bug-Exposing Test Generation via Step-Level Credit Assignment and Self-Distillation from Valuable FailuresPapers · 20 Oct 2026 · Wenjie Gan, Zhaoxi Zhang, Jiahui Liang, Weikang Li, Wanyuan Wang, Li Zhang, Jian…
- RIPPLE: Recovery Internalization through Process-Level Policy Learning and Effect-Aware Credit for Long-Horizon Agentic CodingPapers · 22 Oct 2026 · Jingxuan Xu, Zihang Xu, Yuechen Luo, Ziqi Miao, Wenqiang Zhu, Weihao Li, Junhao…
- H$^2$R: Hierarchical Hint-Based Reward for Mathematical Reasoning via Reinforcement LearningPapers · 29 Oct 2026 · He Tu, Siye Li, Hongyi Wen
Latest sources
- Safe Reinforcement Learning for Multi-Turn Clinical Agents: When Dense Rewards Beat Outcome RewardsPapers · 9 Nov 2026 · Brendan P. Murphy, Agustina Saenz
- Search When Needed, Proceed When Reliable: Metacognitive Reliability Supervision for Efficient Agentic SearchPapers · 30 Oct 2026 · Minhao Sun, Zhijing Wu, Xian-Ling Mao
- H$^2$R: Hierarchical Hint-Based Reward for Mathematical Reasoning via Reinforcement LearningPapers · 29 Oct 2026 · He Tu, Siye Li, Hongyi Wen
- RIPPLE: Recovery Internalization through Process-Level Policy Learning and Effect-Aware Credit for Long-Horizon Agentic CodingPapers · 22 Oct 2026 · Jingxuan Xu, Zihang Xu, Yuechen Luo, Ziqi Miao, Wenqiang Zhu, Weihao Li, Junhao…
- Enhancing Bug-Exposing Test Generation via Step-Level Credit Assignment and Self-Distillation from Valuable FailuresPapers · 20 Oct 2026 · Wenjie Gan, Zhaoxi Zhang, Jiahui Liang, Weikang Li, Wanyuan Wang, Li Zhang, Jian…
- KnowGPO: Knowledge-Grouped Policy Optimization for Search AgentsPapers · 15 Oct 2026 · Siqi Chen, Zhengxi Lu, Yuchen Yan, Xingyu Wu, Qipeng Chen, Xin Zhang, Aiting Liu…
- Lean4AgenticRL: Advancing Agentic RL with Verifiable Multi-level Process RewardsPapers · 15 Oct 2026 · Ruida Wang, Gerald Friedland, Tong Zhang, Chris Kong
- MERIT: Multimodal Evidence Reasoning over Image TablesPapers · 13 Oct 2026 · Haoyang Jiang, Zhuotong Fan, Mingyang Yi, Ju Fan, Lanqing Hu, Xiao Zan