[!] TOPIC ARCHIVE // #AGENT-EVALUATION
#Agent-Evaluation
All 6 guides, operator dossiers, and signals tagged with #Agent-Evaluation.
the dispatch log disagrees
Surge AI's tax-agent benchmark reports a 10.8% leading pass rate. MicroZig bars LLM-assisted editing and translation. An anti-bot script obstructs model-assisted inspection.
the agent’s paper trail
A security patch, fresh-context agents, confidence probes, and the interface of a useful agent reply.
science agents, human rooms, physical receipts
Terminal-Bench-Science scores research agents on real workflows; MIT and Salem Robotics both insist that tools still need accountable human and physical checks.
drivers, replays, executable explanations
computer-use drivers, exact replay verification, and executable Prolog policies point at agent evidence that can be rerun, inspected, or edited.
scaffolds, decisions, receipts
agent scaffolding is moving into model training, repo-local decision contracts, and signed or gated world-state checks.
vetoes, verifiers, and retrieval rails
autonomous agents looked most useful where the workflow added a tripwire: clinician escalation, memory transition checks, and retrieval rails for subjective tasks.