PulseAugur / Brief
EN
LIVE 08:23:21

Brief

last 24h
[2/2] 224 sources

Multi-source AI news clustered, deduplicated, and scored 0–100 across authority, cluster strength, headline signal, and time decay.

  1. SIGNIFICANT · 36氪 (36Kr) 中文(ZH) · · [2 sources]

    Enlightenment World Model Ranks First in Embodied Intelligence Evaluations Such as RoboTwin 2.0

    ACE ROBOTICS has released its Kairos world model, which has achieved top rankings in four major embodied AI benchmarks: RoboTwin 2.0, LIBERO-Plus, WorldModelBench Robot, and DreamGen. The model utilizes a novel unified architecture for multimodal understanding, generation, and prediction, addressing limitations in existing video generation models. Kairos has been made open-source to the industry. AI

    IMPACT Sets new SOTA on embodied AI benchmarks, potentially accelerating robotics development and commercialization.

  2. RESEARCH · Hugging Face Daily Papers English(EN) · · [3 sources]

    World Model Self-Distillation: Training World Models to Solve General Tasks

    Researchers have developed a new framework for training video diffusion models to solve general tasks by combining self-distillation and reinforcement learning. This method allows the models to learn task-solving abilities from unlabeled data, bypassing the need for costly, curated task-video supervision. The approach uses a vision-language model to generate tasks and solutions, which then guide a video diffusion model to learn execution, further enhanced by reinforcement learning from the vision-language model's feedback. AI

    IMPACT Enables video diffusion models to perform complex tasks without explicit task-video data, potentially accelerating robotics and planning applications.