2025-07-14

Models

• Kimi K2: Open Agentic Intelligence
• T5Gemma: A new collection of encoder-decoder Gemma models - Google Developers Blog
• SmolLM3: smol, multilingual, long-context reasoner
• Voxtral | Mistral AI
• Audio Flamingo 3 - NVIDIA ADLR
• Dream-Coder 7B | HKU NLP Group

Papers

• [2506.03487] ProRank: Prompt Warmup via Reinforcement Learning for Small Language Models Reranking
• [2507.06457] A Systematic Analysis of Hybrid Linear Attention
• Kimi-VL/Kimi-VL.pdf at main · MoonshotAI/Kimi-VL · GitHub
• [2507.09404] Scaling Laws for Optimal Data Mixtures
• [2507.10524] Mixture-of-Recursions: Learning Dynamic Recursive Depths for Adaptive Token-Level Computation
• [2502.00382] Masked Generative Nested Transformers with Decode Time Scaling
• [2507.07101] Small Batch Size Training for Language Models: When Vanilla SGD Works, and Why Gradient Accumulation Is Wasteful
• [2507.11851] Your LLM Knows the Future: Uncovering Its Multi-Token Prediction Potential

Code

• GitHub - EPFL-VILAB/fm-vision-evals

Articles

• QK-Clip：让Muon在Scaleup之路上更进一步 - 科学空间|Scientific Spaces
• On the Tradeoffs of SSMs and Transformers | Goomba Lab
• DreamOn: Diffusion Language Models For Code Infilling Beyond Fixed-Size Canvas | HKU NLP Group

Videos

• Zed Inferred: Diffusion Language Models - YouTube
• Simple Diffusion Language Models - YouTube
• Arthur Douillard - Distributed Training in Machine Learning - YouTube
• JSALT 2025 - Plenary Talk - L.Barrault - Large Concept Model: beyond token-based LLMs - YouTube
• Timothée Darcet - Scaling Self Supervised Learning for Vision An Introduction to DINOv2 - YouTube
• Stanford CS25: V5 I Transformers in Diffusion Models for Image Generation and Beyond - YouTube

Other

• [ ]

Tweets