EmbodiedSWE: Frontier Coding Agents Crack Long-Horizon Dexterous Robotics Tasks in Sim
PeterHndrsn · x · 2026-09-25
EmbodiedSWE explores using frontier coding agents to control robots in simulation. In an IKEA table assembly demo, Fable 5.1 and Opus 5 fail while Opus 5.5 and GPT-6 Astra both succeed—Opus 5.5 finishes faster, GPT-6 Astra takes a cleaner, safer trajectory—with success rates climbing sharply over just months. The project has four parts: EmbodiedSWE-Bench, an agent-native benchmark of long-horizon dexterous everyday tasks built on Isaac Lab; evaluations of frontier coding agents on task performance, completion time, and inference cost; EmbodiedSWE-Gen, which diversifies one verified agent solution into large-scale trajectory data for training general robot policies; and RL-based agent improvement using verified outcomes, plus automatic new-task generation. Code, benchmark, and paper are open on GitHub.
Related event: EmbodiedSWE: Coding Agents Tackle Long-Horizon Robotic Tasks(4 posts)→
More from Embodied
- Google Announces Gemma Hardware Hackathon at AGI House With a Strict No-Cloud Rule — agihouse_org · 2026-09-25
- Qualcomm pitches the phone as the AI hub at Snapdragon Summit, aiming for Apple-like cross-device experience — BenBajarin · 2026-09-25
- Cosmic Robotics has installed 11,000 solar panels and signed the largest US builder — Y Combinator · 2026-09-25
- AGIBOT deploys 300+ robots in Chimelong theme park as 20,000th robot ships — TansuYegen · 2026-09-25
- Tesla FSD Supervised Navigates Slovenia's Extremely Narrow, Steep Old-Village Streets — yunta_tsai · 2026-09-25
- Feather Robotics exits stealth with $7.6M pre-seed to sell a $29,990 general-purpose robot — darian314 · 2026-09-25