Evidence · DiggingBeagle record
AgentEval: evidence-bounded assessment of stateful agents
Original controlled evaluations of trace identity, state effects, evidence binding and interrupted-task scoring.
- Source role
- primary disclosure
Evidence record
Original controlled evaluations of trace identity, state effects, evidence binding and interrupted-task scoring.
Claim-level citations (1)
- supportsMatching evidence hashes failed to prove evidence supported the evaluated action: AgentEval tested stateful agent reports against controlled evidence views and found cases where the referenced bytes existed but were bound to the wrong decision or action.
Abstract; counterfactual evidence views and result scoring