Mistral Ventures into Robotics: Introduces VLM Navigation Research, Hits SOTA on R2RCE
sivareddyg · x · 2026-08-11
Mistral AI has officially ventured into robotics, taking its first steps into the physical world. Its latest research focuses on scalable training and navigation based on Vision-Language Models (VLMs).
Technical Highlights:
- Utilizes single RGB-camera pointing navigation.
- Conducts Sim2Real training across 400K trajectories and 6K scenes.
- Achieves 76.6% State-of-the-Art performance on the R2RCE benchmark.
- Introduces prefix caching and tree-based attention for a 22x efficiency boost.
- Employs CISPO for online RL to enhance exploration and recovery capabilities.
More from Embodied
- Paper: DB-VIO, A Dual-Branch Framework for Visual Inertial Odometry — rsasaki0109 · 2026-08-11
- Embodied AI Startups Act as VCs: 29 Firms Make 125 Investments — FinanceYF5 · 2026-08-11
- Galaxy Z Fold 8 Hands-On: Light as a Passport, One-Hand Friendly, Makes Apple Feel Stuck in Past — bilawalsidhu · 2026-08-11
- Enfold: Embeds World Models into Representations, Slashing Robot Control Latency by 10x — Weili Zeng · 2026-08-11
- Samsung Official Details Engineering Secrets of Galaxy Z Fold2 Hinge — MarwaEldiwiny · 2026-08-11
- INTACT by ZJU & Tsinghua: Robots Skip Trial-and-Error to Act Directly on Intent — jiqizhixin · 2026-08-11