RESEARCHInvestigateNEXT 12 MONTHS
trajectory-judge: What Outcome-Only LLM Judges Miss on Agent Trajectories
arXiv cs.CL — Computation and Language
Factual evidence
What the source reports
Research paper shows outcome-only LLM agent evaluation misses execution path failures, policy violations, and inefficiencies in tool usage.
Inspect the evidence
- Inclusion basis
- Enterprise AI
- Publisher and source type
- arXiv cs.CL — Computation and Language · RESEARCH
- Published by source
- 4 November 2026
- Collected by OneBench
- 3 Sept 2026, 03:08 UK
Stored source excerpt
arXiv:2609.00038v1 Announce Type: new Abstract: Outcome-only evaluation is the production default for LLM agents: show a judge the request and the final reply and ask…
Short excerpt from the collected text, not the full source. Use the source link to read it in context.
The factual summary is a OneBench synthesis, not a quotation or independent verification. Collection time is not publication time. Open the source for its full context; related reporting can share the same underlying announcement.