Skip to content
AI News HubLIVE
Original source2 min read

trajectory-judge: What Outcome-Only LLM Judges Miss on Agent Trajectories

Summary

This paper reveals a critical blind spot in outcome-only evaluation of LLM agents: it cannot detect agents that reach the correct answer through wrong methods. In a controlled environment with fault injection, the outcome-only judge catches 84% of loud faults but only 45% of silent ones, while flagging 33% of correct trajectories. A step-rubric judge achieves 77% silent recall with zero false alarms at 3x the cost. Moreover, no judge catches an invented promise appended to a perfect trajectory, and self-consistency triples cost without improvement. The authors advocate for stratifying recall by outcome survival and release the environment and analysis pipeline.

SourcearXiv Computational LinguisticsAuthor: Hadi Mohammadi
trajectory-judge: What Outcome-Only LLM Judges Miss on Agent Trajectories
Report an error

The correction channel is not available yet. You can copy the article reference below for later.

Correction instructions
Read article

[Submitted on 29 Aug 2026]

Title:trajectory-judge: What Outcome-Only LLM Judges Miss on Agent Trajectories

View a PDF of the paper titled trajectory-judge: What Outcome-Only LLM Judges Miss on Agent Trajectories, by Hadi Mohammadi

View PDF HTML (experimental)

Abstract:Outcome-only evaluation is the production default for LLM agents: show a judge the request and the final reply and ask whether it was handled well. The metric is structurally blind to an agent that reaches the right answer the wrong way. We measure that blind spot where ground truth is known by construction: a deterministic tool-using support-desk environment, a scripted oracle policy that always solves it, and a fault injector that breaks exactly one thing at a known step, stratifying faults by whether the customer-visible outcome survived (silent) or not (loud). Five judges (programmatic rules, outcome-only, step-rubric at two model sizes, and a self-consistency ensemble) are scored on detection, step localisation, fault typing, calibration, and cost over 400 trajectories. The outcome-only judge catches 84% of loud faults but 45% of silent ones while flagging 33% of correct trajectories; a step-rubric judge reaches 77% silent recall with zero false alarms at 3x the cost. No judge reads the final reply: an invented promise appended to an otherwise perfect trajectory evades the rules entirely and the step judge 82% of the time, and self-consistency triples cost while improving nothing. We argue that judge evaluations must stratify recall by outcome survival, and release the environment, the injector, all raw verdicts, and an analysis pipeline that rebuilds every number offline.

Comments: 16 pages (8-page main text). Under review at a NeurIPS 2026 workshop. Code, data, and raw verdicts: this https URL

Subjects:

Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Software Engineering (cs.SE)

Cite as: arXiv:2609.00038 [cs.CL]

(or arXiv:2609.00038v1 [cs.CL] for this version)

https://doi.org/10.48550/arXiv.2609.00038

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Hadi Mohammadi [view email] [v1] Sat, 29 Aug 2026 10:14:05 UTC (72 KB)

Full-text links:

Access Paper:

View a PDF of the paper titled trajectory-judge: What Outcome-Only LLM Judges Miss on Agent Trajectories, by Hadi Mohammadi

View PDF

HTML (experimental)

TeX Source

view license

Current browse context:

cs.CL

new | recent | 2026-09

Change to browse by:

cs cs.AI cs.SE

References & Citations

NASA ADS

Google Scholar

Semantic Scholar

Loading...

Data provided by:

Bibliographic Tools

Bibliographic and Citation Tools

Bibliographic Explorer Toggle

Bibliographic Explorer (What is the Explorer?)

Connected Papers Toggle

Connected Papers (What is Connected Papers?)

Litmaps Toggle

Litmaps (What is Litmaps?)

scite.ai Toggle

scite Smart Citations (What are Smart Citations?)

Code, Data, Media

Code, Data and Media Associated with this Article

alphaXiv Toggle

alphaXiv (What is alphaXiv?)

Links to Code Toggle

CatalyzeX Code Finder for Papers (What is CatalyzeX?)

DagsHub Toggle

DagsHub (What is DagsHub?)

GotitPub Toggle

Gotit.pub (What is GotitPub?)

Huggingface Toggle

Hugging Face (What is Huggingface?)

ScienceCast Toggle

ScienceCast (What is ScienceCast?)

Demos

Demos

Replicate Toggle

Replicate (What is Replicate?)

Spaces Toggle

Hugging Face Spaces (What is Spaces?)

Spaces Toggle

TXYZ.AI (What is TXYZ.AI?)

Related Papers

Recommenders and Search Tools

Link to Influence Flower

Influence Flower (What are Influence Flowers?)

Core recommender toggle

CORE Recommender (What is CORE?)

Author

Venue

Institution

Topic

About arXivLabs

arXivLabs: experimental projects with community collaborators

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

Which authors of this paper are endorsers? | Disable MathJax (What is MathJax?)

Key points and analysis

Article intelligence

EngineersIntermediate

Key points

  • Outcome-only LLM judges have a structural blind spot, failing to detect agents that do the right thing the wrong way.
  • In experiments, the outcome-only judge caught only 45% of silent faults (where the visible outcome succeeds) and falsely flagged 33% of correct trajectories.
  • A step-rubric judge detected 77% of silent faults with zero false alarms, but at 3x the cost.
  • Self-consistency ensembles increased cost without improving performance, and an invented promise in the final reply fooled all judges.

Highlights and analysis are generated automatically and may contain errors. Check the original source.