# LLMZero: Discovering Adaptive Training Strategies for RL Post-Training via LLM Agents

arXiv: https://arxiv.org/abs/2606.18388 · HTML: https://mlsanity.com/paper/2606.18388/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Reinforcement learning for LLM agents, one of 16 directions in LLM agents and tool use.
- Builds on 23 of its 39 references that are in the map.
- First posted on arXiv 16 Jun 2026.

## Placed in

- LLM agents and tool use / [Reinforcement learning for LLM agents](https://mlsanity.com/agents/reinforcement-learning-for-llm-agents/)

## Builds on (most cited first, 23 of 23)

- [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)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [HybridFlow: A Flexible and Efficient RLHF Framework](https://mlsanity.com/paper/2409.19256/) (2024)
- [AlphaEvolve: A coding agent for scientific and algorithmic discovery](https://mlsanity.com/paper/2506.13131/) (2025)
- [Reasoning with Language Model is Planning with World Model](https://mlsanity.com/paper/2305.14992/) (2023)
- [Group Sequence Policy Optimization](https://mlsanity.com/paper/2507.18071/) (2025)
- [Hyperband: A Novel Bandit-Based Approach to Hyperparameter Optimization](https://mlsanity.com/paper/1603.06560/) (2016)
- [Population Based Training of Neural Networks](https://mlsanity.com/paper/1711.09846/) (2017)
- [AIDE: AI-Driven Exploration in the Space of Code](https://mlsanity.com/paper/2502.13138/) (2025)
- [A System for Massively Parallel Hyperparameter Tuning](https://mlsanity.com/paper/1810.05934/) (2018)
- [MiMo: Unlocking the Reasoning Potential of Language Model -- From Pretraining to Posttraining](https://mlsanity.com/paper/2505.07608/) (2025)
- [An Empirical Study on Eliciting and Improving R1-like Reasoning Models](https://mlsanity.com/paper/2503.04548/) (2025)
- [PostTrainBench: Can LLM Agents Automate LLM Post-Training?](https://mlsanity.com/paper/2603.08640/) (2026)
- [QwenLong-L1: Towards Long-Context Large Reasoning Models with Reinforcement Learning](https://mlsanity.com/paper/2505.17667/) (2025)
- [Beyond Chemical QA: Evaluating LLM's Chemical Reasoning with Modular Chemical Operations](https://mlsanity.com/paper/2505.21318/) (2025)
- [DUMP: Automated Distribution-Level Curriculum Learning for RL-based LLM Post-training](https://mlsanity.com/paper/2504.09710/) (2025)
- [How Difficulty-Aware Staged Reinforcement Learning Enhances LLMs' Reasoning Capabilities: A Preliminary Experimental Study](https://mlsanity.com/paper/2504.00829/) (2025)
- [Towards an AI Musician: Synthesizing Sheet Music Problems for Musical Reasoning](https://mlsanity.com/paper/2509.04059/) (2025)
- [WildSci: Advancing Scientific Reasoning from In-the-Wild Literature](https://mlsanity.com/paper/2601.05567/) (2026)
- [PaperSearchQA: Learning to Search and Reason over Scientific Papers with RLVR](https://mlsanity.com/paper/2601.18207/) (2026)
- [Can Generalist Agents Automate Data Curation?](https://mlsanity.com/paper/2606.04261/) (2026)
- [SIRI: Scaling Iterative Reinforcement Learning with Interleaved Compression](https://mlsanity.com/paper/2509.25176/) (2025)

## Built on by (most cited first, 3 of 3)

- [Mitra-v2 Technical Report](https://mlsanity.com/paper/2609.04540/) (2026)
- [What is Missing from AI Post-Training AI: An Empirical Analysis](https://mlsanity.com/paper/2608.19072/) (2026)
- [ImproveAnyTask: An Autonomous Post-Training Harness for Iterative Model Self-Improvement](https://mlsanity.com/paper/2610.06347/) (2026)

## Abstract (arXiv)

RL post-training strategies are dataset-dependent and reveal a recurring empirical pattern: capacity parameters accumulate monotonically across stages, while regularization parameters predominantly oscillate in response to shifting training dynamics. This distinction highlights a potential flaw in fixed training schedules: by forcing all parameters along rigid paths, they fail to capture the dynamic exploration-exploitation tradeoffs that regularization must track. We uncover this through LLMZero, an agentic system that optimizes training trajectories via tree search by diagnosing pathologies at each checkpoint and proposing coordinated multi-parameter transitions. Across four diverse GRPO tasks, LLMZero discovers strategies that improve over the base model by 9% to 140% and over grid search by 6% to 15% (relative), consistently outperforming random search and a skill-based agent under a matched compute budget. The capacity--regularization asymmetry is consistent across all four tasks, offering a candidate design heuristic for multi-stage training.