Daily Papers of 2023-10-26

  1. A Picture is Worth a Thousand Words: Principled Recaptioning Improves Image Generation 53 upvotes, #1 of 2023-10-26
  2. CommonCanvas: An Open Diffusion Model Trained with Creative-Commons Images 38 upvotes, #2 of 2023-10-26
  3. DreamCraft3D: Hierarchical 3D Generation with Bootstrapped Diffusion Prior 33 upvotes, #3 of 2023-10-26
  4. QMoE: Practical Sub-1-Bit Compression of Trillion-Parameter Models 27 upvotes, #4 of 2023-10-26
  5. Wonder3D: Single Image to 3D using Cross-Domain Diffusion 22 upvotes, #5 of 2023-10-26
  6. An Early Evaluation of GPT-4V(ision) 22 upvotes, #5 of 2023-10-26
  7. ConvNets Match Vision Transformers at Scale 21 upvotes, #7 of 2023-10-26
  8. LLM-FP4: 4-Bit Floating-Point Quantized Transformers 14 upvotes, #8 of 2023-10-26
  9. Detecting Pretraining Data from Large Language Models 11 upvotes, #9 of 2023-10-26
  10. TiC-CLIP: Continual Training of CLIP Models 10 upvotes, #10 of 2023-10-26
  11. CLEX: Continuous Length Extrapolation for Large Language Models 10 upvotes, #10 of 2023-10-26
  12. TD-MPC2: Scalable, Robust World Models for Continuous Control 8 upvotes, #12 of 2023-10-26
  13. LightSpeed: Light and Fast Neural Light Fields on Mobile Devices 5 upvotes, #13 of 2023-10-26

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.