# Step-wise Rubric Rewards for LLM Reasoning

arXiv: https://arxiv.org/abs/2605.17291 · HTML: https://mlsanity.com/paper/2605.17291/

## Key facts

- Cited by 14 papers in the map (data through 14 Oct 2026).
- Placed in Reinforcement learning for reasoning, one of 17 directions in LLM reasoning.
- Builds on 21 of its 26 references that are in the map.
- First posted on arXiv 17 May 2026.

## Placed in

- LLM reasoning / [Reinforcement learning for reasoning](https://mlsanity.com/lr/reinforcement-learning-for-reasoning/)

## Builds on (most cited first, 21 of 21)

- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Self-Consistency Improves Chain of Thought Reasoning in Language Models](https://mlsanity.com/paper/2203.11171/) (2022)
- [Measuring Mathematical Problem Solving With the MATH Dataset](https://mlsanity.com/paper/2103.03874/) (2021)
- [Let's Verify Step by Step](https://mlsanity.com/paper/2305.20050/) (2023)
- [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://mlsanity.com/paper/2503.14476/) (2025)
- [Solving Quantitative Reasoning Problems with Language Models](https://mlsanity.com/paper/2206.14858/) (2022)
- [OlympiadBench: A Challenging Benchmark for Promoting AGI with Olympiad-Level Bilingual Multimodal Scientific Problems](https://mlsanity.com/paper/2402.14008/) (2024)
- [Math-Shepherd: Verify and Reinforce LLMs Step-by-step without Human Annotations](https://mlsanity.com/paper/2312.08935/) (2023)
- [Prometheus 2: An Open Source Language Model Specialized in Evaluating Other Language Models](https://mlsanity.com/paper/2405.01535/) (2024)
- [Improve Mathematical Reasoning in Language Models by Automated Process Supervision](https://mlsanity.com/paper/2406.06592/) (2024)
- [Rubrics as Rewards: Reinforcement Learning Beyond Verifiable Domains](https://mlsanity.com/paper/2507.17746/) (2025)
- [LLM-Rubric: A Multidimensional, Calibrated Approach to Automated Evaluation of Natural Language Texts](https://mlsanity.com/paper/2501.00274/) (2024)
- [BabyVision: Visual Reasoning Beyond Language](https://mlsanity.com/paper/2601.06521/) (2026)
- [Interleaved Latent Visual Reasoning with Selective Perceptual Modeling](https://mlsanity.com/paper/2512.05665/) (2025)
- [Reward and Guidance through Rubrics: Promoting Exploration to Improve Multi-Domain Reasoning](https://mlsanity.com/paper/2511.12344/) (2025)
- [From Context to Skills: Can Language Models Learn from Context Skillfully?](https://mlsanity.com/paper/2604.27660/) (2026)
- [Grouter: Decoupling Routing from Representation for Accelerated MoE Training](https://mlsanity.com/paper/2603.06626/) (2026)
- [GRPO-VPS: Enhancing Group Relative Policy Optimization with Verifiable Process Supervision for Effective Reasoning](https://mlsanity.com/paper/2604.20659/) (2026)

## Built on by (most cited first, 14 of 14)

- [From Holistic Evaluation to Structured Criteria: Rubrics Across the Evolving LLM Landscape](https://mlsanity.com/paper/2606.08625/) (2026)
- [Beyond Imitation: Filtering On-Policy Distillation by Reasoning Progress](https://mlsanity.com/paper/2608.19408/) (2026)
- [Rubric-Guided Process Reward for Stepwise Model Routing](https://mlsanity.com/paper/2605.29310/) (2026)
- [SERPO: Self-Evolving Rubric Policy Optimization for Open-Ended Test-Time Reinforcement Learning](https://mlsanity.com/paper/2607.26873/) (2026)
- [Learning Multimodal Embeddings with Evidence-Aligned Readout](https://mlsanity.com/paper/2609.33659/) (2026)
- [Right Makes Might: Aligning Verified Hidden States Empowers RL Reasoning](https://mlsanity.com/paper/2606.03234/) (2026)
- [APTER: Adaptive Post-Training with Expert-Grounded Rubrics](https://mlsanity.com/paper/2608.14212/) (2026)
- [RGDT-Bench: Benchmarking LLM Reasoning for Rule-Governed Decisions and Their Justifications](https://mlsanity.com/paper/2609.34455/) (2026)
- [SkillRubric: Co-Evolving Actor Guidance and Evaluator Rubrics for Multimodal Agents](https://mlsanity.com/paper/2609.34557/) (2026)
- [From Migration to Calibration: Aligning Agents with Deployment Requirements](https://mlsanity.com/paper/2609.35149/) (2026)
- [Scoring Higher, Answering Worse: Mitigating Reward Hacking in Rubric-Based RL via Protocol-Level Rubrics](https://mlsanity.com/paper/2609.38847/) (2026)
- [UP-MOPD: Update Projection in Multi-Teacher On-Policy Distillation](https://mlsanity.com/paper/2610.08398/) (2026)
- [Matched-Context Group Policy Optimization for LLM Agent Training](https://mlsanity.com/paper/2610.13755/) (2026)
- [CoCA: Criterion-Aware Credit Assignment for Rubric-Based Reinforcement Learning](https://mlsanity.com/paper/2610.17733/) (2026)

## Abstract (arXiv)

Reinforcement Learning with Verifiable Rewards (RLVR) is widely used to improve reasoning in large language models, but rewards only final-answer correctness with no supervision over intermediate steps. Rubric-based methods such as Rubrics as Rewards (RaR) introduce finer-grained supervision by scoring rollouts against structured criteria, yet the rubric scores are still aggregated into a single scalar applied to the entire response, causing three weaknesses: loss of multi-criterion structure, uniform supervision of correct and incorrect steps, and reward hacking through unbounded self-correction. On 1,000 problems, we find 18.2% of steps in correct-answer responses are wrong yet positively rewarded, while 49.9% of steps in incorrect-answer responses are correct yet penalized. We introduce Step-wise Rubrics as Rewards (SRaR), an RLVR framework that (i) uses an LLM judge to attribute each rubric item to a specific reasoning step, (ii) normalizes per-step rubric scores across rollouts so only steps whose quality varies produce a learning signal, and (iii) combines the per-step reward with the outcome reward through a decoupled advantage estimator that keeps the outcome baseline stable. We further build a 16K-problem rubric dataset by contrastively distilling rubric items from correct and flawed reasoning paths sampled from a strong model. Across six mathematical reasoning benchmarks, SRaR improves average accuracy over RaR by 3.57 points on Qwen3-8B and 2.75 points on Qwen3-32B, raises the Faithful Reasoning Rate on AIME 2025 from 34.5% to 46.7%, and reduces self-correction looping from 48.1% to 26.5%.