Newest papers

Atlas: Making Sense of Long Horizon Agent Trajectories

Sangho Suh, Joseph Chee Chang, Pao Siangliulue, Amy Xian Zhang, Matt Latzke, Jonathan Bragg, Doug Downey, Daniel S. Weld · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.30548 · cited by 0 works (OpenAlex) · open-access PDF

Korean

AI 에이전트는 이제 문서를 검색하고 가설을 생성하고 데이터 세트를 분석하고 실험을 실행할 수 있습니다.그러나 스케일링 과학 실행은 연구자의 에이전트가 생산하는 것에 대한 이해를 자동으로 확대하지 않습니다.연구자들이 긴 지평선으로 돌아갈 때, 그들이 모니터링하지 않은 작업을 평가해야합니다: 에이전트가 무엇을 생산했는지, 어떤 결정이 그 출력을 형성했는지, 그리고 그 결정이 연구 목표를 달성했는지 여부. 우리는이 실행 후 감각화를 세 가지 연관된 수준에 걸쳐 나타내는 문제로 프레젠테이션으로 프레젠테이션으로 프레젠테이션합니다.이 프레젠테이션에서 우리는 Atlas를 개발했습니다.이 프레젠테이션은 에이전트가 선택한 것을

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

Abstract

AI agents can now search literature, generate hypotheses, analyze datasets, and run experiments at a scale difficult for researchers to match. Yet scaling scientific execution does not automatically scale researchers' ability to make sense of what agents produce. When researchers return to a long-horizon run, they must evaluate work they did not monitor: what the agent produced, which decisions shaped those outputs, and whether those decisions served the research goal. We frame this post-run sensemaking as a representation problem across three linked levels: goal, decision, and artifact. From this framework we developed Atlas, an agentic workspace with a trajectory tree, which renders the run as a branching structure showing what the agent chose and set aside, and artifact projections over six representations. In a within-subjects study (N=14), participants reported higher decision transparency, artifact sensemaking, and goal connection with Atlas than with a standard baseline, together with substantially lower subjective workload.

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

Topics: Computer science, Artifact (error), Set (abstract data type), Sensemaking, Scale (ratio), Representation (politics)

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