Pro² Agent paper: explicit Problem State lifts autonomous AI R&D to 76% MLE-Bench
rohanpaul_ai · x · 2026-11-21
The arXiv paper 'Pro² Agent: Scientific Problem Probing for Autonomous AI R&D' (Google + Johns Hopkins) argues LLM research agents stall on hard problems because they only optimize a codebase without representing the problem structure. Pro² Agent replaces the codebase state with an explicit Problem State — suspected bottlenecks, candidate solutions from targeted literature research, and confidence levels updated through controlled hypothesis testing. It matches the best reported 70.7% medal rate on full MLE-Bench at 24 hours and reaches a record 76.0% at 48 hours, performs competitively on PostTrainBench, and emits transparent research notes reusable by human researchers.
More from coding & agent
- Reverse-engineering Apple Watch: a single Claude Code prompt builds a physics watch face in 30 minutes — mtizard · 2026-11-21
- us-legal-mcp: an MCP server exposing comprehensive US legislation to AI agents — modelcontextprotocol · 2026-11-21
- Maindex launches Expert Memory Service, a full-control memory API for AI agents — modelcontextprotocol · 2026-11-21
- AI-built oil painting engine: 2,300 commits in 8 days as Claude Code runs a 'studio' of sessions — FotoRe_store · 2026-11-21
- Open-source chirp gives your AI agents a shared, self-editable message board — itsarnavb · 2026-11-21
- Frontier agents complete under 8% of tasks in 26-person simulated org, study finds — vasuman · 2026-11-21