HumanCLAW open-sourced: all 9 SOTA VLMs fail embodied benchmark, best hits only 16.8%
liuziwei7 · x · 2026-08-18
HumanCLAW, an evaluation framework decoupling a VLM's action intelligence from low-level motor control, is now fully open-sourced. At every sub-second step, an off-the-shelf VLM issues an atomic whole-body skill realized as continuous full-body motion with real physical consequences (contact, collision, gravity), while balance and motor-tracking failures are factored out — isolating the model's moment-to-moment action decisions.
HumanCLAW-Bench provides 1,218 long-horizon egocentric find–navigate–interact episodes across 41 indoor scenes. None of nine SOTA VLMs solves it; the best reaches only 16.8% success, suggesting current VLMs lack embodied self-awareness.
More from Embodied
- Stanley Robotics builds robots that move, park, and deliver cars autonomously — lukas_m_ziegler · 2026-08-18
- Inside a Robot Dog: Engineering Complexity Beyond the Cute Design — Olivier__OG · 2026-08-18
- Unitree Achieves Software-Speed Iteration for Hardware — broodsugar · 2026-08-18
- Robot breaks speed record, smashes electrical box — skdh · 2026-08-18
- Humanoid Robots Now Handle Packages With Human-Like Coordination — CurieuxExplorer · 2026-08-18
- Robotics for Agriculture and Resilience: Ruggedize Event Draws 250+ Deep Techies — Scobleizer · 2026-08-18