One Jev read of a trace ranks agent failures, including quiet ones where many runs agree, and guides hill-climbing. No answer key.

Benchmark note, text only. treyserbeam asks whether one run, with no answer key, can catch an agent's wrong answer and then generate tasks that expose new failures. The post says one Jev read of the trace can rank failures, including quiet failures where many runs agree, and can guide hill-climbing. Code is open. The link is an X article: https://x.com/i/article/2115471510387216384.