# S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?

arXiv: https://arxiv.org/abs/2608.31100 · HTML: https://mlsanity.com/paper/2608.31100/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Placed in Self-rewarding alignment and self-training, one of 13 directions in Synthetic data and self-improvement for LLMs.
- Builds on 29 of its 41 references that are in the map.
- First posted on arXiv 31 Aug 2026.

## Placed in

- Synthetic data and self-improvement for LLMs / [Self-rewarding alignment and self-training](https://mlsanity.com/synth/self-rewarding-alignment-and-self-training/)

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

- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [Reflexion: Language Agents with Verbal Reinforcement Learning](https://mlsanity.com/paper/2303.11366/) (2023)
- [Voyager: An Open-Ended Embodied Agent with Large Language Models](https://mlsanity.com/paper/2305.16291/) (2023)
- [AgentBench: Evaluating LLMs as Agents](https://mlsanity.com/paper/2308.03688/) (2023)
- [ALFWorld: Aligning Text and Embodied Environments for Interactive Learning](https://mlsanity.com/paper/2010.03768/) (2020)
- [STaR: Bootstrapping Reasoning With Reasoning](https://mlsanity.com/paper/2203.14465/) (2022)
- [Self-Rewarding Language Models](https://mlsanity.com/paper/2401.10020/) (2024)
- [Reinforced Self-Training (ReST) for Language Modeling](https://mlsanity.com/paper/2308.08998/) (2023)
- [AgentBoard: An Analytical Evaluation Board of Multi-turn LLM Agents](https://mlsanity.com/paper/2401.13178/) (2024)
- [ScienceWorld: Is your Agent Smarter than a 5th Grader?](https://mlsanity.com/paper/2203.07540/) (2022)
- [JudgeBench: A Benchmark for Evaluating LLM-based Judges](https://mlsanity.com/paper/2410.12784/) (2024)
- [TextWorld: A Learning Environment for Text-based Games](https://mlsanity.com/paper/1806.11532/) (2018)
- [ReasoningBank: Scaling Agent Self-Evolving with Reasoning Memory](https://mlsanity.com/paper/2509.25140/) (2025)
- [PostTrainBench: Can LLM Agents Automate LLM Post-Training?](https://mlsanity.com/paper/2603.08640/) (2026)
- [Self-Improving LLM Agents at Test-Time](https://mlsanity.com/paper/2510.07841/) (2025)
- [Re-ReST: Reflection-Reinforced Self-Training for Language Agents](https://mlsanity.com/paper/2406.01495/) (2024)
- [MetaSkill-Evolve: Recursive Self-Improvement of LLM Agents via Two-Timescale Meta-Skill Evolution](https://mlsanity.com/paper/2607.05297/) (2026)
- [RetroAgent: From Solving to Evolving via Retrospective Dual Intrinsic Feedback](https://mlsanity.com/paper/2603.08561/) (2026)
- [SEA-Eval: A Benchmark for Evaluating Self-Evolving Agents Beyond Episodic Assessment](https://mlsanity.com/paper/2604.08988/) (2026)
- [SEAGym: An Evaluation Environment for Self-Evolving LLM Agents](https://mlsanity.com/paper/2606.17546/) (2026)
- [AI4AI-Bench: Benchmarking LLM Agents in Algorithmic Design for Recursive Self-Improvement](https://mlsanity.com/paper/2608.20318/) (2026)
- [PAST-Bench: Benchmarking the Foundations of Recursive Self-Improvement in Personal Agents](https://mlsanity.com/paper/2608.04003/) (2026)
- [ContinualSkillBench: Can LLM Agents Truly Evolve Their Capabilities?](https://mlsanity.com/paper/2608.03874/) (2026)
- [Credit Without Ground Truth: Auditing Step-Level Credit Assignment in LLM Agents Against Executed Replay](https://mlsanity.com/paper/2608.19760/) (2026)
- [Do Agent Optimizers Compound? A Continual-Learning Evaluation on Terminal-Bench 2.0](https://mlsanity.com/paper/2607.14004/) (2026)
- [When Self-Evolution Backfires: Pre-Commit Gating against Skill Contamination in LLM Agents](https://mlsanity.com/paper/2608.05810/) (2026)
- [On the Fragility of Self-Improving Agents: Variance, Task Order, and Underspecification](https://mlsanity.com/paper/2608.18066/) (2026)
- [FinEvo-Bench: A Longitudinal Benchmark for Self-Evolving Agents in Professional Financial Workflows](https://mlsanity.com/paper/2608.06144/) (2026)
- [Social Gym and SPaRTan: Benchmarking and Improving LLM Social Reasoning via Multi-Agent Game Tournaments](https://mlsanity.com/paper/2608.09128/) (2026)

## Built on by (most cited first, 5 of 5)

- [The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement](https://mlsanity.com/paper/2609.11873/) (2026)
- [ModularRSI: Modular and Generalizable Recursive Harness Self-Improvement](https://mlsanity.com/paper/2609.14857/) (2026)
- [AI for Games in the Foundation Model Era](https://mlsanity.com/paper/2609.16679/) (2026)
- [MMPostTrainBench: Benchmarking Autonomous Research for Multimodal Post-Training](https://mlsanity.com/paper/2610.05398/) (2026)
- [From Existing Methods to New Research Ideas: Two Perspectives on Recursive Self-Improvement](https://mlsanity.com/paper/2610.14625/) (2026)

## Abstract (arXiv)

Large language models (LLMs) increasingly interact with external environments and accumulate substantial behavioral experience, yet existing agent benchmarks largely evaluate them as fixed policies. It therefore remains unclear whether an agent can actively test its behavior, judge the resulting experience, and use that experience to improve future decisions. We introduce textbfStextsuperscript3Gym, an interactive benchmark for evaluating LLM self-improvement through three coupled capabilities: Self-Testing, Self-Judging, and Self-Improvement. S3Gym separates permissive exploration from strict held-out evaluation and instantiates this protocol in seven text-based games with executable environment verifiers. We evaluate three pathways for incorporating interaction experience: direct History ICL, score-conditioned Summary Memory, and parameter Training. Our experiments reveal that self-improvement is neither automatic nor uniform. Context-level experience improves performance for several model--game pairs, but the most effective pathway depends strongly on the task structure: summaries are beneficial when experience can be compressed into reusable strategic rules, yet often underperform raw history when success depends on precise, state-contingent information. Parameter training produces substantial gains on some tasks, but also exhibits unstable improvement and severe negative transfer on others. These findings show that recognizing successful actions is insufficient; agents must also transform feedback into executable and transferable policies. S3Gym provides a unified framework for diagnosing this process and identifying the bottlenecks that prevent agents from translating interaction experience into reliable self-improvement.