← back to paper

Review history

arxiv: 2608.00794 · 2 revisions

Measurement Without Validity: The Compounding Reliability Problem in Agentic AI Evaluation

1. 2026-08-05 REJECT MODERATE v1.3.0-alltime-deepseek novelty 5.0
   148227 ms 17178 in 16482 out 2026-08-05T04:13:01.421299+00:00
2. 2026-08-05 CONDITIONAL MODERATE v1.3.0-alltime-deepseek novelty 5.0
   136355 ms 13836 in 14633 out 2026-08-05T00:11:27.576293+00:00