Papers with process reward model
Progressive Multimodal Reasoning via Active Retrieval (2025.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to improve multimodal large language models' reasoning performance are limited. |
| Approach: | They propose a framework to progressively improve multimodal reasoning capabilities . they propose active retrieval and Monte Carlo tree search to improve MLLMs' reasoning . |
| Outcome: | The proposed framework improves multimodal reasoning capabilities in multimodal large language models. |
Boosting Policy and Process Reward Models with Monte Carlo Tree Search in Open-Domain QA (2025.findings-acl)
Copied to clipboard
Chi-Min Chan, Chunpu Xu, Junqi Zhu, Jiaming Ji, Donghai Hong, Pengcheng Wen, Chunyang Jiang, Zhen Ye, Yaodong Yang, Wei Xue, Sirui Han, Yike Guo
| Challenge: | Experimental results show that our approach can effectively improve the performance of both the policy model and the reward model. |
| Approach: | They propose to use Monte Carlo Tree Search for both policy model improvement and reward model improvement to bridge it to more subtle open-domain question answering. |
| Outcome: | The proposed approach surpasses existing methods for annotation and training data with fewer data points and achieves better performance in test-time scaling strategies. |
AgentPro: Enhancing LLM Agents with Automated Process Supervision (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing frameworks lack explicit supervision during the reasoning process, which may lead to error propagation across reasoning chains. |
| Approach: | They propose a framework which automates process supervision for large language model agents by automatically generating step-level annotations and developing a process reward model based on these annotations. |
| Outcome: | The proposed framework outperforms existing agent-based methods on four datasets and achieves a 6.32% increase in accuracy. |
Android Coach: Improve Online Agentic Training Efficiency with Single State Multiple Actions (2026.acl-long)
Copied to clipboard
| Challenge: | Existing reinforcement learning methods are expensive due to high latency and sample inefficiency . Currently, RL is limited to one-to-one state-action pairs . |
| Approach: | They propose a framework that shifts the training paradigm to Single State Multiple Actions and introduce a group-wise advantage estimator based on the averaged critic outputs. |
| Outcome: | The proposed framework achieves 7.5% and 8.3% success rate improvements on AndroidLab and AndroidWorld over UI-TARS-1.5-7B and attains 1.4x higher training efficiency than existing methods. |
How Is LLM Reasoning Distracted by Irrelevant Context? An Analysis Using a Controlled Benchmark (2025.emnlp-main)
Copied to clipboard
| Challenge: | Prior work has not explored the mechanisms underlying this sensitivity. |
| Approach: | They propose a synthetic benchmark to evaluate Large Language Models’ reasoning robustness against systematically controlled irrelevant context (IC). |
| Outcome: | The proposed model improves in-distribution and out-of-disttribution scenarios while training with strong distractors. |
ToolPRM: Fine-Grained Inference Scaling of Structured Outputs for Function Calling (2026.acl-long)
Copied to clipboard
Jianghao Lin, Yuanyuan Shi, Xin Peng, Renjie Ding, Hairui Wang, Yuxuan Peng, Bizhe Bai, Weixi Song, Fengshuo Bai, Huacan Chai, Weinan Zhang, Fei Huang, Ying Wen
| Challenge: | Existing research on inference scaling focuses on unstructured output generation tasks, such as mathematical problems. |
| Approach: | They propose an inference-scaling framework that combines fine-grained beam search with ToolPRM, a process reward model scoring each intra-call decision. |
| Outcome: | The proposed framework outperforms outcome and coarse-grained reward models in predictive accuracy and yields consistent test-time gains on multiple function-calling benchmarks. |
Verified Critical Step Optimization for LLM Agents (2026.findings-acl)
Copied to clipboard
| Challenge: | Critical Step Optimization (CSO) focuses preference learning on verified critical steps where alternative actions demonstrably flip task outcomes from failure to success. |
| Approach: | They propose a method which focuses preference learning on verified critical steps where alternative actions demonstrably flip task outcomes from failure to success. |
| Outcome: | The proposed method outperforms the existing methods on GAIA-Text-103 and XBench-DeepSearch while requiring supervision at only 16% of trajectory steps. |