Hermes Agent, as shipped
Every learning mechanism the harness ships switched on.
- Held-out tasks, budget used
- 23.1 solved 94%
- Model spend
- $4.20 5,483 calls
- Instances
- 100 9.7 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.