Log inSign up
Image
user avatar
DailyPapers
@HuggingPapers
Tweeting interesting papers submitted at huggingface.co/papers. Submit your own at hf.co/papers/submit, and link models/datasets/demos to it!
Anywhere
huggingface.co/docs/hub/en/mo…
Joined March 2025
4
Following
20.2K
Followers
RepliesMedia
  • user avatar
    DailyPapers
    @HuggingPapers
    15h
    Real-time 720p video editing at 30 FPS is here JoyAI-Video-Edit is a 16B autoregressive diffusion model for open-ended video editing. It transforms live video streams on the fly as frames arrive, without needing future context or predefined clip lengths.
    Image
  • user avatar
    DailyPapers
    @HuggingPapers
    19h
    Alibaba released UEmbed on Hugging Face A decoder-only multimodal embedding model that generates both dense and sparse lexical representations from text, image, and video inputs in a single forward pass.
    Image
  • user avatar
    DailyPapers
    @HuggingPapers
    23h
    Skill-Alpha uses RL to generate agent skills Instead of manual heuristics, Skill-Alpha models skill creation as progressive edits optimized via RL and rollback rewards from downstream execution. It boosts task success on CL-Bench by 3.3 points and tau2-bench by 6.7 points.
    Image
  • user avatar
    DailyPapers
    @HuggingPapers
    Aug 4
    Dual-Anchored Policy Distillation for LLMs On-policy self-distillation can cause privilege illusion, where student models rely on training signals absent at inference. DAPD fixes this information asymmetry by aligning reference and rollout behavior under matched conditions.
    Image