4DAnyone: Create Anyone in 4D from a Casual Monocular Video Paper โข 2608.20335 โข Published 6 days ago โข 76
FlashPrefill V2: Block-Sparse Prefill Attention for Long-Context LLM Serving Paper โข 2608.19758 โข Published 6 days ago โข 17
HeavySkill: Heavy Thinking as the Inner Skill in Agentic Harness Paper โข 2605.02396 โข Published May 4 โข 24
GenericAgent: A Token-Efficient Self-Evolving LLM Agent via Contextual Information Density Maximization (V1.0) Paper โข 2604.17091 โข Published Apr 18 โข 23
Seedance 2.0: Advancing Video Generation for World Complexity Paper โข 2604.14148 โข Published Apr 15 โข 170
VideoWorld 2: Learning Transferable Knowledge from Real-world Videos Paper โข 2602.10102 โข Published Feb 10 โข 14
MagicLens: Self-Supervised Image Retrieval with Open-Ended Instructions Paper โข 2403.19651 โข Published Mar 28, 2024 โข 26
view post Post 5317 We collaborated with Hugging Face to enable you to train MoE models 12ร faster with 35% less VRAM via our new Triton kernels (no accuracy loss). ๐คTrain gpt-oss locally on 12.8GB VRAM with our free notebooks: https://unsloth.ai/docs/new/faster-moe See translation 1 reply ยท ๐ฅ 28 28 ๐ค 5 5 + Reply
Running on Zero MCP Featured 98 Pixel3dmm [Image Mode] ๐ท 98 Versatile Single-Image 3D Face Reconstruction
Running Agents Featured 857 Qwen3 Demo ๐ 857 Chat with an AI assistant that thinks before answering