Mouse-1 takes an agent state and candidate actions, then returns probabilities for the next step. An agent harness can use it for frequent small decisions while its main model handles deeper reasoning or code.
Features
- State and option inputs
- Next-step probabilities
- Agent-harness API integration
Use Cases
- Choose agent actions
- Offload frequent decisions
FAQ
Mouse-1 takes an agent state and candidate actions, then returns probabilities for the next step. An agent harness can use it for frequent small decisions while its main model handles deeper reasoning or code. Core capabilities include: State and option inputs, Next-step probabilities, Agent-harness API integration.
Common scenarios include: Choose agent actions, Offload frequent decisions.
