Metraipolis › archive › dc_608024b9b1fa

Language Models as AI Research World Models — arXiv:2610.12235

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.12235
License (as archived)
CC0-1.0
Archived
2026-10-09T06:13:28.436Z
SHA-256 of the archived text
449d47519f9ddb3b257d8c88958d9f0207dbaa6b34369d6d3ff68f98d271c872
Stable URL of this page
https://metraipolis.com/documento/dc_608024b9b1fa

How to cite

Language Models as AI Research World Models — arXiv:2610.12235 — arXiv (en), https://arxiv.org/abs/2610.12235 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_608024b9b1fa (Metraipolis archive dc_608024b9b1fa, sha256 449d47519f9ddb3b257d8c88958d9f0207dbaa6b34369d6d3ff68f98d271c872)
BibTeX
@misc{dc608024b9b1fa,
  title = {Language Models as AI Research World Models — arXiv:2610.12235},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.12235},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_608024b9b1fa; sha256 449d47519f9ddb3b257d8c88958d9f0207dbaa6b34369d6d3ff68f98d271c872},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.12235. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

AI research agents automate the cycle of proposing, implementing, and evaluating experiments, opening a path toward recursive self-improvement. Yet their ability to propose experiments outpaces their capacity to execute them in real environments, making outcome prediction a key capability for sustained self-improvement under limited experimental budgets. We investigate language models as Research World Models (RWMs), which predict the outcomes of candidate interventions across research environments. Our evaluation draws on over 2,600 experimental records from nine research environments spanning pretraining, post-training, and inference, representing more than 171,000 H100 GPU-hours of experimentation. Research knowledge acquired from real experimental experience improves RWM predictions of unseen interventions within the same environment (Spearman +0.27), and can be reused across environments. For example, using only pretraining experience from OLMo3, Marin, and Nanochat, an RWM reduces selection regret in the Qwen3 environment by 78% compared with zero-experience setting. These benefits extend to multi-round Autoresearch under a fixed selection budget: RWMs with in-env and cross-env research knowledge increase the best gain achieved by 15.8% and 11.6%, respectively. Ablations across 13 language models used as RWMs show that adding research knowledge can improve intervention ranking more than changing models or increasing reasoning effort alone. These findings support language models as RWMs and motivate accumulating experimental data for future RWM training.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.12235 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-09T05:47:15.395Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).