Manifold launches to accelerate robotics research evaluation, compressing a month of experiments into a week
ZeYanjie · x · 2026-08-13
Manifold is a platform designed to accelerate robotics research by speeding up the evaluation loop. Key features include:
- Self-explaining failure analysis that clusters failure modes across rollouts;
- Simulate a thousand scenarios in 30 minutes, compress a month of experiments into a week;
- Dense labels on every atomic sub-goal and action;
- One eval harness for every benchmark, simulator, or world model, with native support for NVIDIA's Isaac Lab Arena.
More from Embodied
- DynaRobotics CEO Lindon Gao on running deployment and research feedback loops together — micoolcho · 2026-10-16
- AutoAgent0: VLM-powered agentic runtime lifts long-tail driving success by 28.3pp — yining_hong · 2026-10-16
- CarryOn: dual-loop harness adds personalization and safe recovery to frozen VLA robots — Zixuan Wang · 2026-10-16
- FoveaVLA replaces wrist cameras with learned foveation, hitting 70.8% success — Haoyu_Xiong_ · 2026-10-16
- Musk teases two-way Grok x Optimus integration: Grok in the bot, bot in Grok — beffjezos · 2026-10-16
- SimVTA uses simulator-rendered video to post-train driving world action models with RL — OpenDriveLab · 2026-10-16