SmallThinker: A Family of Efficient LLMs Natively Trained for Local Deploymentarxiv.org·2 pts·limoce·0
Polaris: A Post-training recipe for scaling RL on Advanced Reasoning modelshkunlp.github.io·4 pts·limoce·1
Overclocking LLM Reasoning: Monitoring and Controlling LLM Thinking Path Lengthsroyeisen.github.io·63 pts·limoce·0
Machine Learning Conferences Should Establish "Refutations and Critiques" Trackarxiv.org·3 pts·limoce·0
SepLLM: Accelerate LLMs by Compressing One Segment into One Separatorsepllm.github.io·39 pts·limoce·2
Step-Video-T2V: The Practice, Challenges, and Future of Video Foundation Modelarxiv.org·41 pts·limoce·5
DeepSeek-VL2: Moe Vision-Language Models for Advanced Multimodal Understanding [pdf]github.com·1 pts·limoce·0
REST: A Plug-and-Play Method for Accelerating LLM Without Additional Trainingsites.google.com·1 pts·limoce·0
Smoke 'em if you got 'em: Hacker gains root access using cigarette lightertomshardware.com·2 pts·limoce·0
FlexAttention: The Flexibility of PyTorch with the Performance of FlashAttentionpytorch.org·210 pts·limoce·24
MiniCPM-v2.6: GPT-4V Level MLLM for Single/Multi Image and Video on Your Phonegithub.com·1 pts·limoce·0
MindSearch: LLM-Based Web Search Engine Similar to Perplexity.ai and SearchGPTgithub.com·3 pts·limoce·0
Turbo Sparse: Achieving LLM SOTA Performance with Minimal Activated Parametersarxiv.org·11 pts·limoce·0