Newest papers

Same Simulation, Different Evidence: Where Retrieved Tool Output Costs LLM Agents Accuracy

Ke Zhang, Sahchit Chundur, Jerry Peprah Owusu, Mohammad Javad Abdolhosseini Qomi, Maziar Raissi · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26836 · cited by 0 works (OpenAlex) · open-access PDF

Korean

LLM 에이전트는 올바른 대답을 계산하는 것보다 옳은 대답을 할 수 있습니다. 에이전트는 점점 더 긴 도구 출력을 컨텍스트 창 밖으로 유지하고, 에이전트가 검색하거나 요청에 따라 읽는 파일로 저장합니다. 우리는 진정한 과학 도구로 정확성을 평가하는 곳을 묻습니다: PHREEQC, USGS에서 널리 사용되는 지화학 시뮬레이터, 보고서가 846.8k 문자에 도달하고 모든 인쇄 된 값을 추적 할 수있는 결정주의를 유지합니다. 에이전트는 158 실행 검증 된 다중 선택 항목에 응답하기 위해 그것을 사용합니다. 에이전트가 섹션 인덱스를 통해 시뮬레이터의 보고서를 읽을 때, GPT-5.1과 GPT-5.2

Machine translation (Meta M2M100); the English original follows.

Abstract

LLM agents that compute the right answer can still answer wrongly. Agent harnesses increasingly keep long tool outputs out of the context window, saving them to files that the agent searches or reads on demand. We ask where this costs accuracy with a real scientific tool: PHREEQC, a widely used geochemistry simulator from the USGS, whose reports reach 846.8k characters and whose determinism lets every printed value be traced. Agents use it to answer 158 execution-verified multiple-choice items. When agents read the simulator's report through a section index rather than in context, GPT-5.1 and GPT-5.2 lose 11.3 and 7.9 points of accuracy, while Claude Opus 4.6 and Claude Sonnet 4.6 show no detectable cost. Re-running all 11,007 logged simulator calls shows that both GPT models compute the keyed value about as often as before; what changes is the evidence they answer from. Under retrieval they often skip the report, answer from their own higher-precision output files, and misround to a neighboring option, a failure we call evidence substitution. With the simulation held fixed, retrieval still costs GPT-5.2 9.4 points. For GPT-5.4, more frequent correct simulations mask an evidence-use loss of about ten points, leaving a net change indistinguishable from zero. Naming the authoritative output raises accuracy by 6.3-7.9 points over plain retrieval for the three GPT models; for GPT-5.1 and GPT-5.2, inlining outputs that fit works as well, while asking for correct rounding does not. Context-management choices in agent harnesses should be judged by where accuracy is lost, not only by final accuracy.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Rounding, Context (archaeology), Value (mathematics), Section (typography), Ask price

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange