GPT-6 Astra is impressive, but agentic robotics ≠ just control. Physical agents should control robots, learn new skills, and design their own hardware. So we introduce RLE-Bench: 48 everyday robotics engineering tasks spanning closed-loop control, policy learning, perception, and mechanical design.

Sep 14, 2026 · 4:44 PM UTC

13
33
13
226
170,353
RelevantRecentLikes
(5/5) Huge thanks to our core contributors @haitongma5426 @chenxiaogao @rqiang_rushi and our advisors @daibond_alpha and @Na_Lina_Li! Our RLE-Bench 2.0 is calling for contributions, check out here rle-bench.github.io/contribu…!
4
561
Nice work🔥! I've added RLE-Bench to Awesome Robot Use Agent: github.com/kairunwen/Awesome…
Image from Kairun Wen (@KairunWen)
1
7
527
how can hardware design count without manufacturability constraints, that's geometry search not engineering
204
Benchmarks like this are useful because they test whether an agent can behave more like a robotics engineer, not just a controller.
1
253
RLE Bench makes the broader robotics challenge more concrete
122
RLE-Bench pushes evaluation beyond control into the full robotics loop. The mechanical design component especially points toward agents becoming active participants in how robots are built.
1
100
Load more replies