# Co-Evolving LLM Evaluators and Policies via DynamicRubric

arXiv: https://arxiv.org/abs/2607.20083 · HTML: https://mlsanity.com/paper/2607.20083/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- Builds on 29 of its 53 references that are in the map.
- First posted on arXiv 22 Jul 2026.

## Builds on (most cited first, 29 of 29)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Let's Verify Step by Step](https://mlsanity.com/paper/2305.20050/) (2023)
- [Constitutional AI: Harmlessness from AI Feedback](https://mlsanity.com/paper/2212.08073/) (2022)
- [G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment](https://mlsanity.com/paper/2303.16634/) (2023)
- [Asynchronous Methods for Deep Reinforcement Learning](https://mlsanity.com/paper/1602.01783/) (2016)
- [MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark](https://mlsanity.com/paper/2406.01574/) (2024)
- [Scaling Laws for Reward Model Overoptimization](https://mlsanity.com/paper/2210.10760/) (2022)
- [DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence](https://mlsanity.com/paper/2606.19348/) (2026)
- [Length-Controlled AlpacaEval: A Simple Way to Debias Automatic Evaluators](https://mlsanity.com/paper/2404.04475/) (2024)
- [AlpacaFarm: A Simulation Framework for Methods that Learn from Human Feedback](https://mlsanity.com/paper/2305.14387/) (2023)
- [Prometheus: Inducing Fine-grained Evaluation Capability in Language Models](https://mlsanity.com/paper/2310.08491/) (2023)
- [From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipeline](https://mlsanity.com/paper/2406.11939/) (2024)
- [LMSYS-Chat-1M: A Large-Scale Real-World LLM Conversation Dataset](https://mlsanity.com/paper/2309.11998/) (2023)
- [Understanding the Effects of RLHF on LLM Generalisation and Diversity](https://mlsanity.com/paper/2310.06452/) (2023)
- [Rubrics as Rewards: Reinforcement Learning Beyond Verifiable Domains](https://mlsanity.com/paper/2507.17746/) (2025)
- [Skywork-Reward: Bag of Tricks for Reward Modeling in LLMs](https://mlsanity.com/paper/2410.18451/) (2024)
- [JudgeBench: A Benchmark for Evaluating LLM-based Judges](https://mlsanity.com/paper/2410.12784/) (2024)
- [WildBench: Benchmarking LLMs with Challenging Tasks from Real Users in the Wild](https://mlsanity.com/paper/2406.04770/) (2024)
- [FLASK: Fine-grained Language Model Evaluation based on Alignment Skill Sets](https://mlsanity.com/paper/2307.10928/) (2023)
- [RM-Bench: Benchmarking Reward Models of Language Models with Subtlety and Style](https://mlsanity.com/paper/2410.16184/) (2024)
- [FineSurE: Fine-grained Summarization Evaluation using LLMs](https://mlsanity.com/paper/2407.00908/) (2024)
- [HelpSteer3-Preference: Open Human-Annotated Preference Data across Diverse Tasks and Languages](https://mlsanity.com/paper/2505.11475/) (2025)
- [Chasing the Tail: Effective Rubric-based Reward Modeling for Large Language Model Post-Training](https://mlsanity.com/paper/2509.21500/) (2025)
- [Personalized RewardBench: Evaluating Reward Models with Human Aligned Personalization](https://mlsanity.com/paper/2604.07343/) (2026)
- [WeChat-YATT: A Scalable, Simple, Efficient, and Production Ready Training Library](https://mlsanity.com/paper/2508.07970/) (2025)

## Built on by (most cited first, 10 of 10)

- [SERPO: Self-Evolving Rubric Policy Optimization for Open-Ended Test-Time Reinforcement Learning](https://mlsanity.com/paper/2607.26873/) (2026)
- [Momentum-Coupled Rubric Adaptation for Detailed Image Captioning](https://mlsanity.com/paper/2609.36893/) (2026)
- [RecurSE: Bounded Recursive Self-Evaluation for LLM Rubric Judges](https://mlsanity.com/paper/2608.24231/) (2026)
- [ImpossibleRubrics: Stress-Testing Generated Rubrics as Reward Signals](https://mlsanity.com/paper/2609.16816/) (2026)
- [Self-Designed Evaluators and Warm Memory for Long-Horizon Agents](https://mlsanity.com/paper/2609.33717/) (2026)
- [Evolving Support Priorities in Empathetic Reinforcement Learning](https://mlsanity.com/paper/2609.34249/) (2026)
- [From Weak Task Specifications to Scientific Extraction Agents: Optimizing Task Construction](https://mlsanity.com/paper/2609.34829/) (2026)
- [From Judgment Quality to Downstream Utility: Rethinking LLM-as-a-Judge for Open-Ended Tasks](https://mlsanity.com/paper/2609.37145/) (2026)
- [VeriFine: Scaling Verification for Self-Improvement in Embodied Reasoning](https://mlsanity.com/paper/2610.08761/) (2026)
- [Verify Less, Evolve More: Training Idea-Level Critics for Verification-Efficient ML Evolving Agents](https://mlsanity.com/paper/2610.08993/) (2026)

## Abstract (arXiv)

Post-training with evaluator feedback on policy-induced samples serves as a major mechanism for improving large language models. As policies improve, these sampled responses become close in quality. These close candidates create a bottleneck for policy optimization: collapsed relative evaluator score gaps yield weak or misleading policy supervision. We theoretically characterize why these gaps matter through a probability allocation view, showing that the directional gain of shifting probability mass from one response to another is exactly the evaluator score gap between them. This identifies relative score gaps as the policy optimization signals that guide updates. Motivated by this view, we propose DynamicRubric, a response-set-conditioned evaluator--policy co-evolution framework that generates weighted binary rubric items for each candidate set and aggregates the resulting judgments into response-level scores. In our experiments with 8B backbones, DynamicRubric improves evaluator performance and provides stronger policy supervision than baselines using a 70B reward model or a 235B static rubric generator. DynamicRubric-optimized policies also show gains on verifiable reasoning and coding tasks. A DynamicRubric-optimized model is fully deployed in WeChat Search's AI answering scenario, where it serves all online traffic across tens of millions of requests per day and improves key online metrics. These results suggest a principle for evaluator-guided post-training: evaluators should evolve with the policies they supervise.