2025-01-20

Models

• GitHub - deepseek-ai/DeepSeek-R1
  • DeepSeek R1
• GitHub - MoonshotAI/Kimi-k1.5

Papers

• [2501.09891] Evolving Deeper LLM Thinking
• DeepSeek R1
• [2501.10318] HiMix: Reducing Computational Complexity in Large Vision-Language Models
• [2501.04765] TREAD: Token Routing for Efficient Architecture-agnostic Diffusion Training
• [2501.12326] UI-TARS: Pioneering Automated GUI Interaction with Native Agents
• [2501.12370] Parameters vs FLOPs: Scaling Laws for Optimal Sparsity for Mixture-of-Experts Language Models
• [2501.09747] FAST: Efficient Action Tokenization for Vision-Language-Action Models
• [2501.12948] DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning

Code

• Pallas Quickstart — JAX documentation

Articles

• Implementing Character.AI’s Memory Optimizations in nanoGPT | njkumarr
• Literature Review on Sampling Techniques for Language Models | njkumarr
• EvaByte: Efficient Byte-level Language Models at Scale

Videos

• RobotLearningIntroPart2 - YouTube
• FlashInfer - YouTube
• Mosaic GPU - YouTube
• ML Scalability & Performance Reading Group Session 5: Paged Attention - YouTube

Other

• Life lessons from the first half-century of my career | Hacker News

Tweets