Today for AI
HOT RADAR
agentHEAT 15.2°

AI CURATED HOT STORY · 11/9/2026

Berkeley-Led Study Finds Self-Evolving Agent Harnesses Lose Solved Tasks

1 reports archived1 independent sourcesupdated 11/9/2026, 18:56:54
Synthesis & Latest Updates
1 Sources Cross-Validated

A new paper from UC Berkeley and other top labs finds that letting agents rewrite their own prompts, skills, tools, or code to raise benchmark scores hides tasks they lose along the way. Testing 4 evolution methods including GEPA and HyperAgents with GPT-5.4 across 2 harnesses and 2 benchmarks, all 16 evolved agents gained some tasks and lost others; one raised reliability by 5.2 points while dropping 10 previously covered tasks, and 8 of 11 more-reliable agents cost more per run. The authors urge tracking lost tasks, cost per run, and unseen-task results rather than average score alone.

LATEST/A new paper from UC Berkeley and other top labs finds that letting agents rewrite their own prompts, skills, tools, or code to raise benchmark scores hides tasks they lose along the way. Testing 4 evolution methods including GEPA and HyperAgents with GPT-5.4 across 2 harnesses and 2 benchmarks, all 16 evolved agents gained some tasks and lost others; one raised reliability by 5.2 points while dropping 10 previously covered tasks, and 8 of 11 more-reliable agents cost more per run. The authors urge tracking lost tasks, cost per run, and unseen-task results rather than average score alone.

Hourly heat curve

8 fully observed hours
Now
15.2
Peak
17.719:00
24h change
–

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

Coverage timeline

Total 1 reports · Latest first
  1. Rohan Paul (@rohanpaul_ai)T2·78 pts

    Berkeley-Led Study Finds Self-Evolving Agent Harnesses Lose Solved Tasks

    • All 16 evolved agents both gained and lost tasks, with average scores masking lost solved tasks
    • One agent raised reliability by 5.2 points while losing 10 covered tasks; 8 of 11 more-reliable agents cost more per run
    • Agent evolution should be evaluated on lost tasks, cost per run, and unseen-task performance, not average score alone