Research map / Large language models / Agents, code and math
OR-PRM: A Process Reward Model for Algorithmic Problem in Operations Research
Yilin Wang, Heng Zhou, Dongxing Mao, Linjie Li, Jingru Tan, Haochen Han, Zhengyuan Yang, Alex Jinpeng Wang, Min Li
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 OR-PRM Best-of-8 selection | Qwen2.5-7B · Overall (6 benchmarks avg) | accuracy (%): 34.9 -> 42.9 | Improved |
| add OR-PRM Best-of-8 selection | Qwen2.5-14B · Overall (6 benchmarks avg) | accuracy (%): 48.8 -> 61.3 | Improved |
| add OR-PRM Best-of-8 selection | Qwen2.5-32B · Overall (6 benchmarks avg) | accuracy (%): 59.6 -> 70.3 | Improved |
| add OR-PRM Best-of-8 selection | LLMOPT · Overall (6 benchmarks avg) | accuracy (%): 73.6 -> 80.5 | Improved |
| add OR-PRM Best-of-8 selection | Qwen2.5-32B · Complex-LP | accuracy (%): 8.2 -> 32.4 | Improved |
| add OR-PRM Best-of-8 selection | Qwen2.5-14B · Easy-LP | accuracy (%): 66.2 -> 89.4 | Improved |
| add OR-PRM Best-of-8 selection | LLMOPT · Complex-LP | accuracy (%): 48.6 -> 67.6 | Improved |
| add OR-PRM Best-of-8 selection | LLMOPT · Easy-LP | accuracy (%): 96.0 -> 97.8 | Improved |
| add DPO alignment after SFT (ablation) | OR-PRM (Qwen2.5-14B) · Easy-LP and Complex-LP average | average accuracy (%): 43.0 -> 51.0 | Improved |
| add DPO alignment after SFT (ablation) | OR-PRM (Qwen2.5-14B) · Easy-LP | accuracy (%): 79.6 -> 89.4 | Improved |
| replace self-consistency majority voting with OR-PRM selection | Qwen2.5-14B · Easy-LP and Complex-LP average | average accuracy (%): 27.2 -> 51.0 | Improved |
| add OR-PRM critic in modeling-critique-code pipeline | Qwen2.5-14B · Complex-LP pass@1 | accuracy (%): 3.6 -> 27.0 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- LLMOPT — LLMOPT: Learning to Define and Solve General Optimization Problems from Scratch (ICLR 2025) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- From Mathematical Reasoning to Code: Generalization of Process Reward Models in Test-Time Scaling (AAAI 2026 · similarity 0.93)
- Rewarding Progress: Scaling Automated Process Verifiers for LLM Reasoning (ICLR 2025 · similarity 0.92)
- Free Process Rewards without Process Labels (ICML 2025 · similarity 0.91)
- Uncertainty-Based Methods for Automated Process Reward Data Construction and Output Aggregation in Mathematical Reasoning (AAAI 2026 · similarity 0.91)
- GenPRM: Scaling Test-Time Compute of Process Reward Models via Generative Reasoning (AAAI 2026 · similarity 0.91)
- Linking Process to Outcome: Conditional Reward Modeling for LLM Reasoning (ICLR 2026 · similarity 0.91)
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.
