{"slug": "inference-after-coding-agents", "title": "Inference After Coding Agents", "summary": "A paper submitted to arXiv on 17 November 2026 by David Van Dijcke finds that four frontier AI coding agents report significant p-values between 54% and 98% of the time on simulated data with zero effects when asked to produce a compelling finding. The paper argues coding agents enable a supercharged form of selective inference and proposes a correction that replays an agent on subsamples of the data to estimate its relative probability of selecting an analysis when it looks statistically favorable. In simulations, the correction lowers the high false rejection rates to near the nominal 5%.", "body_md": "# Statistics > Methodology\n\n  [Submitted on 17 Nov 2026]\n\n# Title:Inference after coding agents\n\n[View PDF](https://arxiv.org/pdf/2611.23174)\n\n[HTML (experimental)](https://arxiv.org/html/2611.23174v1)\n\nAbstract:Empirical research is changing rapidly as AI coding agents are becoming increasingly proficient in every step of it, from writing code to manipulating data. We argue that while the potential of coding agents for science is vast, their use in hypothesis testing poses a substantial threat to classical inference, as these agents enable a supercharged form of selective inference. In particular, we argue that coding agents carrying out empirical research introduce a new type of post-selection inference that combines human-like, open-ended search over an implicit space of estimators with programmatic replayability of both the selection and the estimation process. Indeed, we find that on simulated data with zero effects, four frontier agents report significant p-values between 54% and 98% of the time when asked to produce a compelling finding. We propose a simple procedure for detecting and correcting these distortions. After an agent has completed its analysis, we reset and replay it on subsamples of the data to estimate the agent's relative probability of selecting that analysis when it looks statistically favorable. For agents with stable selection behavior, the resulting corrected test is asymptotically valid and consistent conditional on selection. Importantly, the correction is learned from the black-box agent's actual choices rather than a model of how it searches, so it remains valid even as AI technology continues to evolve rapidly. In simulations, the correction robustly lowers the high false rejection rates to near the nominal 5%. We illustrate the method on applications from two published papers.\n    \n\n## Submission history\n\nFrom: David Van Dijcke [\n[view email](https://arxiv.org/show-email/700b431a/2611.23174)]\n\n**[v1]** Tue, 17 Nov 2026 18:22:01 UTC (3,472 KB)\n\n### Current browse context:\n\nstat.ME\n\n### References & Citations\n\nLoading...\n\n# Bibliographic and Citation Tools\n\nBibliographic Explorer \n\n*(*[What is the Explorer?](https://info.arxiv.org/labs/showcase.html#arxiv-bibliographic-explorer))\nConnected Papers \n\n*(*[What is Connected Papers?](https://www.connectedpapers.com/about))\nLitmaps \n\n*(*[What is Litmaps?](https://www.litmaps.co/))\nscite Smart Citations \n\n*(*[What are Smart Citations?](https://www.scite.ai/))\n# Code, Data and Media Associated with this Article\n\nalphaXiv \n\n*(*[What is alphaXiv?](https://alphaxiv.org/))\nCatalyzeX Code Finder for Papers \n\n*(*[What is CatalyzeX?](https://www.catalyzex.com))\nDagsHub \n\n*(*[What is DagsHub?](https://dagshub.com/))\nGotit.pub \n\n*(*[What is GotitPub?](http://gotit.pub/faq))\nHugging Face \n\n*(*[What is Huggingface?](https://huggingface.co/huggingface))\nScienceCast \n\n*(*[What is ScienceCast?](https://sciencecast.org/welcome))\n# Demos\n\n# Recommenders and Search Tools\n\nInfluence Flower \n\n*(*[What are Influence Flowers?](https://influencemap.cmlab.dev/))\nCORE Recommender \n\n*(*[What is CORE?](https://core.ac.uk/services/recommender))\n# arXivLabs: experimental projects with community collaborators\n\narXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.\n\nBoth 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.\n\nHave an idea for a project that will add value for arXiv's community? [**Learn more about arXivLabs**](https://info.arxiv.org/labs/index.html).", "url": "https://wpnews.pro/news/inference-after-coding-agents", "canonical_source": "https://arxiv.org/abs/2611.23174", "published_at": "2026-11-19 09:59:51+00:00", "updated_at": "2026-11-19 10:16:36.953581+00:00", "lang": "en", "topics": ["ai-research", "ai-agents", "ai-safety", "machine-learning"], "entities": ["David Van Dijcke", "arXiv"], "also_reported_by": [], "takeaway": "Four frontier coding agents reported significant p-values 54% to 98% of the time on zero-effect simulated data; replaying the agent on data subsamples to estimate its selection probability and correcting for it brings false rejection rates back to near the nominal 5%.", "recommendation": {"advice": "When an agent runs hypothesis tests, replay it on data subsamples to estimate how often it picks that analysis when results look favorable, then correct for the selection bias.", "playbook": "evals-testing", "playbook_url": "https://wpnews.pro/playbooks/evals-testing"}, "alternates": {"html": "https://wpnews.pro/news/inference-after-coding-agents", "markdown": "https://wpnews.pro/news/inference-after-coding-agents.md", "text": "https://wpnews.pro/news/inference-after-coding-agents.txt", "jsonld": "https://wpnews.pro/news/inference-after-coding-agents.jsonld"}}