# PostTrainBench: Can LLM Agents Automate LLM Post-Training?

arXiv: https://arxiv.org/abs/2603.08640 · HTML: https://mlsanity.com/paper/2603.08640/

## Key facts

- Cited by 64 papers in the map (data through 14 Oct 2026).
- Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning.
- Builds on 25 of its 44 references that are in the map.
- First posted on arXiv 9 Mar 2026.

## Placed in

- LLM reasoning / [Reasoning for agents, planning and code](https://mlsanity.com/lr/reasoning-for-agents-planning-and-code/)
- LLM agents and tool use / [General LLM agent frameworks and benchmarks](https://mlsanity.com/agents/general-llm-agent-frameworks-and-benchmarks/)

## Builds on (most cited first, 25 of 25)

- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [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)
- [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](https://mlsanity.com/paper/2311.12022/) (2023)
- [Gemma 3 Technical Report](https://mlsanity.com/paper/2503.19786/) (2025)
- [Concrete Problems in AI Safety](https://mlsanity.com/paper/1606.06565/) (2016)
- [Instruction-Following Evaluation for Large Language Models](https://mlsanity.com/paper/2311.07911/) (2023)
- [The AI Scientist: Towards Fully Automated Open-Ended Scientific Discovery](https://mlsanity.com/paper/2408.06292/) (2024)
- [AlphaEvolve: A coding agent for scientific and algorithmic discovery](https://mlsanity.com/paper/2506.13131/) (2025)
- [From Crowdsourced Data to High-Quality Benchmarks: Arena-Hard and BenchBuilder Pipeline](https://mlsanity.com/paper/2406.11939/) (2024)
- [MLE-bench: Evaluating Machine Learning Agents on Machine Learning Engineering](https://mlsanity.com/paper/2410.07095/) (2024)
- [HealthBench: Evaluating Large Language Models Towards Improved Human Health](https://mlsanity.com/paper/2505.08775/) (2025)
- [MLAgentBench: Evaluating Language Agents on Machine Learning Experimentation](https://mlsanity.com/paper/2310.03302/) (2023)
- [Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agents](https://mlsanity.com/paper/2505.22954/) (2025)
- [PaperBench: Evaluating AI's Ability to Replicate AI Research](https://mlsanity.com/paper/2504.01848/) (2025)
- [RE-Bench: Evaluating frontier AI R&D capabilities of language model agents against human experts](https://mlsanity.com/paper/2411.15114/) (2024)
- [AI-Researcher: Autonomous Scientific Innovation](https://mlsanity.com/paper/2505.18705/) (2025)
- [AlgoTune: Can Language Models Speed Up General-Purpose Numerical Programs?](https://mlsanity.com/paper/2507.15887/) (2025)
- [HCAST: Human-Calibrated Autonomy Software Tasks](https://mlsanity.com/paper/2503.17354/) (2025)
- [The Automated LLM Speedrunning Benchmark: Reproducing NanoGPT Improvements](https://mlsanity.com/paper/2506.22419/) (2025)
- [A Survey of AI Scientists](https://mlsanity.com/paper/2510.23045/) (2025)
- [Tiny Model, Big Logic: Diversity-Driven Optimization Elicits Large-Model Reasoning Ability in VibeThinker-1.5B](https://mlsanity.com/paper/2511.06221/) (2025)
- [Bare Minimum Mitigations for Autonomous AI Development](https://mlsanity.com/paper/2504.15416/) (2025)

## Built on by (most cited first, 30 of 64)

- [Toward Generalist Autonomous Research via Hypothesis-Tree Refinement](https://mlsanity.com/paper/2606.11926/) (2026)
- [Claudini: Autoresearch Discovers State-of-the-Art Adversarial Attack Algorithms for LLMs](https://mlsanity.com/paper/2603.24511/) (2026)
- [MLS-Bench: A Holistic and Rigorous Assessment of AI Systems on Building Better AI](https://mlsanity.com/paper/2605.08678/) (2026)
- [Continual Learning Bench: Evaluating Frontier AI Systems in Real-World Stateful Environments](https://mlsanity.com/paper/2606.05661/) (2026)
- [AI for Auto-Research: Roadmap & User Guide](https://mlsanity.com/paper/2605.18661/) (2026)
- [Evo-Bench: Can Language Models Improve Agent Harness?](https://mlsanity.com/paper/2608.09096/) (2026)
- [AI4AI-Bench: Benchmarking LLM Agents in Algorithmic Design for Recursive Self-Improvement](https://mlsanity.com/paper/2608.20318/) (2026)
- [Can AI agents conduct open-ended AI research? Early evidence from two case studies](https://mlsanity.com/paper/2607.27191/) (2026)
- [Agent^2 RL-Bench: Can LLM Agents Engineer Agentic RL Post-Training?](https://mlsanity.com/paper/2604.10547/) (2026)
- [The Meta-Agent Challenge: Are Current Agents Capable of Autonomous Agent Development?](https://mlsanity.com/paper/2606.04455/) (2026)
- [Stealing Reasoning Traces from Proprietary LLM APIs](https://mlsanity.com/paper/2608.09867/) (2026)
- [Training AI Scientists to Replicate Research](https://mlsanity.com/paper/2608.13331/) (2026)
- [CoffeeBench: Benchmarking Long-Horizon LLM Agents in Heterogeneous Multi-Agent Economies](https://mlsanity.com/paper/2606.16613/) (2026)
- [Frontis-MA1: Training an AI4AI Model towards Recursive Self-Improvement in Machine Learning Engineering](https://mlsanity.com/paper/2607.28568/) (2026)
- [EvoMaster: A Foundational Evolving Agent Framework for Agentic Science at Scale](https://mlsanity.com/paper/2604.17406/) (2026)
- [DataMaster: Data-Centric Autonomous AI Research](https://mlsanity.com/paper/2605.10906/) (2026)
- [Can Generalist Agents Automate Data Curation?](https://mlsanity.com/paper/2606.04261/) (2026)
- [Skill-MAS: Evolving Meta-Skill for Automatic Multi-Agent Systems](https://mlsanity.com/paper/2606.18837/) (2026)
- [Automated Researchers Can Mitigate Well-characterized Alignment Failures](https://mlsanity.com/paper/2608.28945/) (2026)
- [S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?](https://mlsanity.com/paper/2608.31100/) (2026)
- [Aspire: Can Models Self-Evolve from Vague Goals?](https://mlsanity.com/paper/2608.31111/) (2026)
- [AutoTrainess: Teaching Language Models to Improve Language Models Autonomously](https://mlsanity.com/paper/2606.31551/) (2026)
- [Beyond Final Scores: A Systematic Evaluation of Agents for Long-Horizon AI Research and Development](https://mlsanity.com/paper/2608.13417/) (2026)
- [BAITBENCH: Measuring Agent Reward Hacking with Optional Shortcuts Planted in ML Tasks](https://mlsanity.com/paper/2608.30724/) (2026)
- [KellyBench: A Benchmark for Long-Horizon Sequential Decision Making](https://mlsanity.com/paper/2604.27865/) (2026)
- [ANDES: Agent Native Data Evolving Synthesis Tool for Autonomous Instruction Alignment](https://mlsanity.com/paper/2606.01279/) (2026)
- [LLMZero: Discovering Adaptive Training Strategies for RL Post-Training via LLM Agents](https://mlsanity.com/paper/2606.18388/) (2026)
- [Locating Hidden Failures Makes Long-Horizon Agents More Reliable](https://mlsanity.com/paper/2609.17930/) (2026)
- [Pioneer Agent: Continual Improvement of Small Language Models in Production](https://mlsanity.com/paper/2604.09791/) (2026)
- [Exploring Autonomous Agentic Data Engineering for Model Specialization](https://mlsanity.com/paper/2605.30407/) (2026)

## Abstract (arXiv)

AI agents have become surprisingly proficient at software engineering over the past year, largely due to improvements in reasoning capabilities. This raises a deeper question: can these systems extend their capabilities to automate AI research itself? In this paper, we explore post-training, the critical phase that turns base LLMs into useful assistants. We introduce PostTrainBench to benchmark how well LLM agents can perform post-training autonomously under bounded compute constraints (10 hours on one H100 GPU). We ask frontier agents (e.g., Claude Code with Opus 4.6) to optimize the performance of a base LLM on a particular benchmark (e.g., Qwen3-4B on AIME). Importantly, we do not provide any predefined strategies to the agents and instead give them full autonomy to find necessary information on the web, run experiments, and curate data. We find that frontier agents make substantial progress but generally lag behind instruction-tuned LLMs from leading providers: 23.2% for the best agent vs. 51.1% for official instruction-tuned models. However, agents can exceed instruction-tuned models in targeted scenarios: GPT-5.1 Codex Max achieves 89% on BFCL with Gemma-3-4B vs. 67% for the official model. We also observe several failure modes worth flagging. Agents sometimes engage in reward hacking: training on the test set, downloading existing instruction-tuned checkpoints instead of training their own, and using API keys they find to generate synthetic data without authorization. These behaviors are concerning and highlight the importance of careful sandboxing as these systems become more capable. Overall, we hope PostTrainBench will be useful for tracking progress in AI R&D automation and for studying the risks that come with it. Website and code are available at https://posttrainbench.com/.