Latest stories
FlyBy Teaches Small Reasoning Models When to Stop Thinking and Start Asking
A new Hugging Face paper distinguishes execution bottlenecks from knowledge bottlenecks in small reasoning models and introduces F...
Hugging Face Paper: UMM-Reflection Teaches Unified Models to Self-Repair Images
A new method applies interleaved reinforcement learning to full reflection trajectories inside a single unified multimodal model,...
Scaling Laws Suggest Encoder-Free Multimodal Models Can Catch Up at 10^22 FLOPs
A new study compares scaling laws for encoder-free and encoder-based multimodal large language models, finding that the visual pri...
VisionHOPE Introduces Self-Modifying Visual Backbones
A new paper proposes VisionHOPE, a general-purpose visual backbone built on Nested Learning's self-referential construction, where...
CoWindow Attention Cuts Long-Context Training Latency Up to 8.6x While Matching Full Attention
Hugging Face researchers introduce CoWA, a structured attention design in which heads collectively cover the full causal history i...
Hugging Face Paper: MALA Lets Attention Allocate Its Own Compute
A new fused attention primitive, MassAlloc Attention (MALA), preserves full causal QK scoring but uses normalized softmax contribu...
Duplex-MPE Benchmark Tests When Voice Assistants Should Stay Silent in Group Chats
Hugging Face researchers introduce Duplex-MPE, a 2,000-scenario benchmark that evaluates full-duplex speech assistants on selectiv...
Xiaomi's GAGAR Rewrites Code Agent RL by Grading Solution Quality, Not Just Test Passes
A new framework from Xiaomi and Hugging Face uses a groupwise agentic grader to redistribute advantage among test-passing trajecto...
Hugging Face Paper: Domain-Normalized Distillation Fixes Multi-Teacher Imbalance
A new paper introduces Domain-Normalized Multi-Teacher On-Policy Distillation (DN-MOPD), which rescales each specialist's feedback...
Hugging Face Paper: Post-Training Leaves 'Behavioral Shadows' That Transfer Skills
A new paper introduces Active Taskless Distillation (ATD), showing that a language model can learn coding, science, and reasoning...
OpenAI Introduces Dots, a Proactive Assistant for Complex and Everyday Work
OpenAI has unveiled Dots, a proactive assistant designed to keep working across complex projects and routine tasks while users rem...
CompoWorld Composes Reusable Services to Train More Capable General Agents
A new Hugging Face paper introduces Compositional Environment Scaling, which builds agent training tasks by composing a library of...