# RecurSE: Bounded Recursive Self-Evaluation for LLM Rubric Judges

arXiv: https://arxiv.org/abs/2608.24231 · HTML: https://mlsanity.com/paper/2608.24231/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Reward hacking and misspecification, one of 10 directions in LLM alignment / RLHF / preference optimisation.
- Builds on 28 of its 31 references that are in the map.
- First posted on arXiv 25 Aug 2026.

## Placed in

- LLM alignment / RLHF / preference optimisation / [Reward hacking and misspecification](https://mlsanity.com/alignment/reward-hacking-and-misspecification/)
- Knowledge distillation / [On-policy and self-distillation for LLMs](https://mlsanity.com/distill/on-policy-and-self-distillation-for-llms/)

## Builds on (most cited first, 28 of 28)

- [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)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](https://mlsanity.com/paper/2311.12022/) (2023)
- [HybridFlow: A Flexible and Efficient RLHF Framework](https://mlsanity.com/paper/2409.19256/) (2024)
- [LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Models](https://mlsanity.com/paper/2403.13372/) (2024)
- [Group Sequence Policy Optimization](https://mlsanity.com/paper/2507.18071/) (2025)
- [SummEval: Re-evaluating Summarization Evaluation](https://mlsanity.com/paper/2007.12626/) (2020)
- [Generative Verifiers: Reward Modeling as Next-Token Prediction](https://mlsanity.com/paper/2408.15240/) (2024)
- [HealthBench: Evaluating Large Language Models Towards Improved Human Health](https://mlsanity.com/paper/2505.08775/) (2025)
- [Gemma 4 Technical Report](https://mlsanity.com/paper/2607.02770/) (2026)
- [Meta-Rewarding Language Models: Self-Improving Alignment with LLM-as-a-Meta-Judge](https://mlsanity.com/paper/2407.19594/) (2024)
- [On the Self-Verification Limitations of Large Language Models on Reasoning and Planning Tasks](https://mlsanity.com/paper/2402.08115/) (2024)
- [CheckEval: A reliable LLM-as-a-Judge framework for evaluating text generation using checklists](https://mlsanity.com/paper/2403.18771/) (2024)
- [RubricHub: A Comprehensive and Highly Discriminative Rubric Dataset via Automated Coarse-to-Fine Generation](https://mlsanity.com/paper/2601.08430/) (2026)
- [Recursive Self-Improvement in AI: From Bounded Self-Refinement to Autonomous Research Loops](https://mlsanity.com/paper/2607.07663/) (2026)
- [RubricBench: Aligning Model-Generated Rubrics with Human Standards](https://mlsanity.com/paper/2603.01562/) (2026)
- [ProfBench: Multi-Domain Rubrics requiring Professional Knowledge to Answer and Judge](https://mlsanity.com/paper/2510.18941/) (2025)
- [Aligning Large Language Models by On-Policy Self-Judgment](https://mlsanity.com/paper/2402.11253/) (2024)
- [RLSR: Reinforcement Learning from Self Reward](https://mlsanity.com/paper/2505.08827/) (2025)
- [GuideBench: Benchmarking Domain-Oriented Guideline Following for LLM Agents](https://mlsanity.com/paper/2505.11368/) (2025)
- [EvoLM: Self-Evolving Language Models through Co-Evolved Discriminative Rubrics](https://mlsanity.com/paper/2605.03871/) (2026)
- [More Convincing, Not More Correct: Self-Play Reward Hacking of Reference-Free LLM Judges](https://mlsanity.com/paper/2607.05904/) (2026)
- [Co-Evolving LLM Evaluators and Policies via DynamicRubric](https://mlsanity.com/paper/2607.20083/) (2026)
- [Grad2Reward: From Sparse Judgment to Dense Rewards for Improving Open-Ended LLM Reasoning](https://mlsanity.com/paper/2602.01791/) (2026)
- [SOP-Maze: Evaluating Large Language Models on Complicated Business Standard Operating Procedures](https://mlsanity.com/paper/2510.08942/) (2025)
- [Reinforcement Learning from Meta-Evaluation: Aligning Language Models Without Ground-Truth Labels](https://mlsanity.com/paper/2601.21268/) (2026)

## Built on by (most cited first, 1 of 1)

- [ECG-RSI: Recursive Self-Improvement through Targeted Electrocardiogram Waveform Evolution](https://mlsanity.com/paper/2610.17718/) (2026)

## Abstract (arXiv)

LLM-as-judge is essential for evaluating open-ended text and steering post-training, yet improving the judge itself typically relies on expensive annotations, reward models, or distillation from stronger teachers. In this work, we eliminate external gold supervision from the RL training reward: the model's own evaluative capability generates learning signals for its optimization -- a closed-loop setting of bounded recursive self-improvement (RSI) termed Recursive Self-Evaluation (RecurSE). We study two central questions: when can self-improvement occur, and when must it stop? First, RecurSE pairs a trainable judge evaluating candidate responses under per-rule rubrics (Pass 1) with a synchronized policy-copy checker that audits the judge's reasoning against meta-rubrics to supply a scalar process reward (Pass 2). To enable learning, interface decoupling structurally isolates the checker's scalar score from the judge's verdict tokens, eliminating a degenerative token-copying shortcut that inflates self-assigned rewards. Second, because unanchored recursive learning is inherently bounded, Pairwise Advantage Validity (PAV) serves as an unbiased validation monitor that jointly tracks judge accuracy and checker fidelity to reliably identify the optimal early-stopping window. Across Qwen3.5-9B, Gemma-4-E4B-it, and Qwen3.6-27B, RecurSE achieves consistent generalization gains across held-out medical, pairwise, summarization, and professional benchmarks. Ablations demonstrate that synchronized judge-checker co-evolution outperforms frozen checkers, external meta-judges, self-consistency, and scaled teacher distillation. Furthermore, preference pairs curated by our judge effectively enhance downstream policy alignment. Bounded RSI for LLM-as-judge is thus viable when self-produced reward validity is explicitly decoupled and monitored.