University of California, Berkeley

University of California, Berkeley on Hugging Face Daily Papers: 17 papers, 1 in the top 3 of their day, 1 paper of the day.

  1. CoVA-SFT: A Large-Scale Dataset for Chain of Visual Abstractions 6 upvotes, #28 of 2026-09-01
  2. FreeToken: Efficient Edge-Native MoE Serving with Bandwidth-Adaptive Execution 97 upvotes, #4 of 2026-08-19
  3. Scaling Laws for Grid-Based Approximate Nearest Neighbor Search in High Dimensions 4 upvotes, #27 of 2026-07-03
  4. Libretto: Giving LLM Agents a Sense of Musical Structure 1 upvotes, #49 of 2026-06-23
  5. BenchEvolver: Frontier Task Synthesis via Solution-Centric Evolution 5 upvotes, #28 of 2026-06-04
  6. Learning, Fast and Slow: Towards LLMs That Adapt Continually 17 upvotes, #17 of 2026-05-13
  7. COMPASS: COntinual Multilingual PEFT with Adaptive Semantic Sampling 2 upvotes, #26 of 2026-04-23
  8. Speculative Decoding for Autoregressive Video Generation 10 upvotes, #16 of 2026-04-22
  9. HybridStitch: Pixel and Timestep Level Model Stitching for Diffusion Acceleration 10 upvotes, #15 of 2026-03-16
  10. SageBwd: A Trainable Low-bit Attention 17 upvotes, #10 of 2026-03-06
  11. Quant VideoGen: Auto-Regressive Long Video Generation via 2-Bit KV-Cache Quantization 33 upvotes, #8 of 2026-02-05
  12. Residual Context Diffusion Language Models 31 upvotes, #10 of 2026-02-05
  13. TurboDiffusion: Accelerating Video Diffusion Models by 100-200 Times 88 upvotes, #1 of 2025-12-25
  14. Arbitrage: Efficient Reasoning via Advantage-Aware Speculation 15 upvotes, #9 of 2025-12-10
  15. Chain-of-Visual-Thought: Teaching VLMs to See and Think Better with Continuous Visual Tokens 25 upvotes, #9 of 2025-11-25
  16. UnSAMv2: Self-Supervised Learning Enables Segment Anything at Any Granularity 9 upvotes, #14 of 2025-11-18
  17. Constantly Improving Image Models Need Constantly Improving Benchmarks 5 upvotes, #21 of 2025-10-21

Data: hysts-bot-data/daily-papers-stats and the Daily Papers API. Open data: tardellirs/paper-pulse-data. Sister project: Model Pulse, the download history of every model on the Hub.