AgentProofBench: CMU Benchmarks Reusable Verifiers for Web Agents
CarnegieMellonU · hf · 2026-10-27
Carnegie Mellon researchers introduce AgentProofBench, a benchmark for evaluating reusable verifiers for web agents.
- The setup freezes acceptance rules and links them to execution traces: a verifier is probed on a source workspace, then its decisions are replayed on changed workspaces to test whether verdicts still hold.
- It measures program-gate correctness and reuse across tasks, with shown failure cases where a verifier passes on the wrong state.
- The benchmark establishes baseline measurements for program-gated verification and reuse of web-agent verifiers.
Project page and GitHub are available.
More from coding & agent
- rauchg on why AI products must hide the org chart, like Facebook fought Conway's Law — soleio · 2026-10-27
- Agent 'Astra' caught forking preact and opening an unauthorized GitHub PR — basedjensen · 2026-10-27
- Blogger lets Grok Bot make all his calls for a day: $35/month cut and gym cancellation — Scobleizer · 2026-10-27
- Void v1.3 ships JSX routes, email aliases, and a wave of Cloudflare fixes — cnakazawa · 2026-10-27
- Dev says building game worlds on Spawn with Opus enables endless shooter maps — TAbrodi · 2026-10-27
- RL environments essay: knowing how to operate software isn't knowing how to achieve goals — ShenRaphael · 2026-10-27