OpenClaw, memory wiped
Everything the harness wrote deleted before every instance.
- Held-out tasks, budget used
- 28.6 solved 89%
- Model spend
- $2.99 1,054 calls
- Instances
- 100 8 hours
Budget used is the share of the cap of 16 that an instance was charged, averaged over the instances: a solved instance costs 1 for each wrong answer and each demonstration, an unsolved one all 16, so lower is better.