El modelo REKI de código abierto de Kyutai logra un rendimiento de manipulación robótica de primer nivel con la mitad del tamaño de sus competidores al rastrear de manera eficiente el historial visual reciente.
Resumen generado con IA. Consulta el original para comprobar el contexto y la precisión.
Este es el extracto disponible del feed. Puede estar abreviado o en el idioma original.
Kyutai, with Inria and École des Ponts, has released REKI, a 2.4B vision-language-action model (VLA) for robot manipulation. The REKI vision-language-action model is pretrained only on public data. It also carries a cheap memory of its recent past. On three out-of-distribution benchmarks, only models about twice its size beat it. All weights are live on Hugging Face today. TL;DR Size: 2.4B parameters (2.2B Qwen3.5-2B VLM + 0.2B flow-matching action expert). Memory covers the last 3 frames, 1 s apart. Runs on: Benchmarked on 1 NVIDIA H100. 21.8 ms per action chunk, compiled, from 2 cameras at 256×256. Weights ship in BF16. No quantized variants announced. Performance: Best among VLAs pretrained on public data on LIBERO-Plus and RoboTwin Clean2Rand. Best: 90.4% on LIBERO-Plus, the top score in Kyutai’s comparison. Worst: 38.4% on RoboCasa365, well behind Xiaomi-Robotics-1 (57.4%). What is REKI? REKI is a vision-language-action model from Kyutai that maps camera views and an instruction to robot actions. It pairs a Qwen3.5-2B backbone with a 200M-parameter action expert. The expert generates action chunks by flow matching, integrated in 10 Euler steps. Robot state and flow time enter through adaptive RMSNorm. Each robot state layout gets its own small state MLP. How does REKI’s memory work? Most VLAs act on the current frame alone. Prior memory methods feed past frames to the VLM, unroll training, or add modules. Each option adds cost. REKI places memory inside the action expert instead. Before each frame’s action tokens sit 32 learned register tokens. They cross-attend the encoded prompt and views of that frame. They also self-attend the registers of earlier frames. Registers never read action tokens. So they depend only on observations, not flow time or robot state. They are computed once per frame and cached in a rolling window. The cost is small, per the project page: Reading cached registers adds 0.1 ms to a 21.8 ms chunk. The cache uses about 2 Mi
Una ayuda automática para ordenar, basada en fuentes, palabras clave y entidades, en una escala de 1–12. Alta: 9–12; media: 6–8; baja: 1–5. No mide urgencia, verifica noticias ni representa respaldo editorial.
Voxtral 3 de Mistral unifica las capacidades de transcripción de voz a texto, síntesis de voz y agentes de voz en una única API, simplificando el desarrollo de aplicaciones de voz en tiempo real.
llm-d 1.0 de Google optimiza la inferencia de IA a gran escala en Kubernetes, resolviendo los problemas de equilibrado de carga para cargas de trabajo de agentes autoalojados y aprendizaje por refuerzo.
La nueva herramienta de simulación generativa de NVIDIA transforma la formación en cirugía clínica al crear escenarios realistas impulsados por IA que mejoran la educación médica y la pericia quirúrgica.