ForceLing Releases Open-Source Embodied AI Model DM0.5
机器之心 · wechat · 2026-08-07
ForceLing released DM0.5, an embodied foundation model emphasizing comprehensive capabilities, open-source availability, and reproducibility, aiming to push embodied AI from demos to scalable deployment.
- Low Barrier: Supports fine-tuning on a single RTX 4090, completing downstream task deployment in as fast as 18 hours with a 60% cost reduction.
- Benchmarks: Achieved 99.0% on LIBERO, set new records on multiple public leaderboards, and ranked 1st in real-world evaluations across four robotic platforms with a 43% success rate. Zero-shot success rate improved by 31%.
- Architecture: A 4B parameter model trained on 150k hours of multi-source data. Introduces an context abstraction layer (60s temporal memory), embodied chain-of-thought, and trajectory alignment.
- Inference: System-level optimizations reduced single-card inference latency from 534ms to 57.49ms (9.29x speedup). Open-sourced on GitHub and HuggingFace.
More from Embodied
- Peking University's SimForcing distills simulation motion priors into robot world models — PekingUniversity · 2026-10-16
- NVIDIA and Mercedes-Benz Expand Partnership on Next-Gen Automated Driving — nvidia · 2026-10-16
- ~25,000 Humanoid Robots Shipped in H1 2026, Over 95% from Chinese Vendors — ingliguori · 2026-10-16
- openpilot brings stoplight-stopping autonomy to a 2019 Corolla — k7agar · 2026-10-16
- Hironobu Fujiyoshi publishes deck: from AI agents to Agentic Robotics — HirokatuKataoka · 2026-10-16
- CoDA co-evolves simulation and robot policies via shared failure memory, lifting OOD success from 5.3% to 27.6% — PLAN-Lab · 2026-10-16