Metraipolis › archive › dc_3de6b3facf87
The Marathon of Scientific Reasoning: Robustness of Scientific Agents to Perturbations in Multi-Turn Interactions — arXiv:2609.34537
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.34537
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T20:57:52.224Z
- SHA-256 of the archived text
2873e72a61b2e62bc18d87e5d10309daa7d6d83aaad1d6e167ca81890db8411b- Stable URL of this page
https://metraipolis.com/documento/dc_3de6b3facf87
How to cite
The Marathon of Scientific Reasoning: Robustness of Scientific Agents to Perturbations in Multi-Turn Interactions — arXiv:2609.34537 — arXiv (en), https://arxiv.org/abs/2609.34537 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_3de6b3facf87 (Metraipolis archive dc_3de6b3facf87, sha256 2873e72a61b2e62bc18d87e5d10309daa7d6d83aaad1d6e167ca81890db8411b)
BibTeX
@misc{dc3de6b3facf87,
title = {The Marathon of Scientific Reasoning: Robustness of Scientific Agents to Perturbations in Multi-Turn Interactions — arXiv:2609.34537},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.34537},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_3de6b3facf87; sha256 2873e72a61b2e62bc18d87e5d10309daa7d6d83aaad1d6e167ca81890db8411b},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34537. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Large language model (LLM)-based scientific agents are increasingly used for scientific problem solving, yet their robustness to imperfections arising during multi-turn interactions remains poorly understood. We introduce \textsc{SciARP} (\textbf{Sci}entific \textbf{A}gent \textbf{R}obustness to \textbf{P}erturbations), a benchmark for evaluating scientific agents under scientifically plausible perturbations throughout multi-turn problem solving. \textsc{SciARP} transforms 620 scientific problems into interdependent tasks of 3--13 turns and defines 13 perturbation types spanning problem understanding, evidence processing, reasoning, and conclusion formation. Clean and perturbed versions of each task are independently executed under matched settings, producing paired live trajectories for evaluating both task success and process reliability. Experiments across eight LLMs from four model families reveal three key robustness characteristics. First, different classes of scientific perturbations exhibit distinct robustness profiles and can decouple task progression from scientific reliability: agents may continue advancing through the task even after their information or reasoning has become unreliable. Second, stronger clean-task performance does not necessarily translate into stronger robustness, as models with higher clean-task accuracy can exhibit larger degradation under perturbation. Third, perturbation effects exhibit strong temporal dynamics: they may remain latent for multiple turns before emerging and subsequently propagate through downstream dependencies. Together, these findings show that current scientific agents remain insufficiently robust to scientifically plausible perturbations, with failures often remaining undetected, propagating, and resisting recovery.
Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.34537 — 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 2026-10-07T20:29:01.574Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).