phenomenon · first seen 8 Oct 2026
verifier brittleness
Benchmark scores of LLM agents changing due to evaluation or scoring-rule changes rather than agent capability.
Mentions per week
PapersCodeDiscussion
Earliest sources
- TRACE: Diagnosing Verifier Brittleness in Agentic EvaluationPapers · 8 Oct 2026 · Radhika Gaonkar
Latest sources
- TRACE: Diagnosing Verifier Brittleness in Agentic EvaluationPapers · 8 Oct 2026 · Radhika Gaonkar