OpenClaw, asked to store
Asked after every verdict to store what would help next time.
- Held-out tasks, budget used
- 22.8 solved 94%
- Model spend
- $2.92 923 calls
- Instances
- 100 8.6 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.