Native CLI scaffolds consistently outper-form OpenCode when using the same model

[2603.08640] PostTrainBench: Can LLM Agents Automate LLM Post-Training? Support arXiv on Cornell Giving Day! We're celebrating 35 years of open science - with YOUR support! Your generosity has helped arXiv thrive for three and a half decades. Give today to help keep science open for ALL for many years to come. Donate! --> Computer Science > Software Engineering arXiv:2603.08640 (cs) [Submitted on 9 Mar 2026 ( v1 ), last revised 10 Mar 2026 (this version, v2)] Title: PostTrainBench: Can LLM Agents Automate LLM Post-Training? Authors: Ben Rank , Hardik Bhatnagar , Ameya Prabhu , Shira Eisenberg , Karina Nguyen , Matthias Bethge , Maksym Andriushchenko View a PDF of the paper titled PostTrainBench: Can LLM Agents Automate LLM Post-Training?, by Ben Rank and 6 other authors View PDF HTML (experimental) Abstract: 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 this https URL . Subjects: Software Engineering (cs.SE) ; Artificial Intelligence (cs.AI); Machine Learning (cs.LG) Cite as: arXiv:2603.08640 [cs.SE] (or arXiv:2603.08640v2 [cs.SE] for this version) https://doi.org/10.48550/arXiv.2603.08640 Focus to learn more arXiv-issued DOI via DataCite Submission history From: Ben Rank [ view email ] [v1] Mon, 9 Mar 2026 17:18:00 UTC (1,338 KB) [v2] Tue, 10 Mar 2026 15:55:40 UTC (1,346 KB) Full-text links: Access Paper: View a PDF of the paper titled PostTrainBench: Can LLM Agents Automate LLM Post-Training?, by Ben Rank and 6 other authors View PDF HTML (experimental) TeX Source view license Current browse context: cs.SE < prev | next > new | recent | 2026-03 Change to browse by: cs cs.AI cs.LG References & Citations NASA ADS Google Scholar Semantic Scholar export BibTeX citation Loading... BibTeX formatted citation × loading... Data provided by: Bookmark Bibliographic Tools Bibliographic and Citation Tools Bibliographic Explorer Toggle Bibliographic Explorer ( What is the Explorer? ) Connected Papers Toggle Connected Papers ( What is Connected Papers? ) Litmaps Toggle Litmaps ( What is Litmaps? ) scite.ai Toggle scite Smart Citations ( What are Smart Citations? ) Code, Data, Media Code, Data and Media Associated with this Article alphaXiv Toggle alphaXiv ( What is alphaXiv? ) Links to Code Toggle CatalyzeX Code Finder for Papers ( What is CatalyzeX? ) DagsHub Toggle DagsHub ( What is DagsHub? ) GotitPub Toggle Gotit.pub ( What is GotitPub? ) Huggingface Toggle Hugging Face ( What is Huggingface? ) Links to Code Toggle Papers with Code ( What is Papers with Code? ) ScienceCast Toggle ScienceCast ( What is ScienceCast? ) Demos Demos Replicate Toggle Replicate ( What is Replicate? ) Spaces Toggle Hugging Face Spaces ( What is Spaces? ) Spaces Toggle TXYZ.AI ( What is TXYZ.AI? ) Related Papers Recommenders and Search Tools Link to Influence Flower Influence Flower ( What are Influence Flowers? ) Core recommender toggle CORE Recommender ( What is CORE? ) Author Venue Institution Topic About arXivLabs arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website. Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them. Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs . Which authors of this paper are endorsers? | Disable MathJax ( What is MathJax? )