Latest stories
Hugging Face Paper Introduces Bellman Policy Optimization, a Critic-Free RL Method for LLM Reasoning
A new paper on Hugging Face presents Bellman Policy Optimization (BPO), a critic-free method derived from Policy Mirror Descent an...
HyperQ Adds Quantum Circuit Branches to Frozen Diffusion Language Models
A new method, HyperQ, attaches token-conditioned quantum residual branches to a frozen masked-diffusion language model, training o...
GAE: A Geometry-Native Latent Space for 3D-Consistent World Generation
Researchers introduce the Geometry-Native Autoencoder (GAE), which reparameterizes geometry foundation model features into a compa...
RULER Turns Rubrics Into Rewards for Better SVG Generation
A new Hugging Face paper proposes rubric-based scoring as a more faithful signal for evaluating and optimizing SVG code generation...
HuRo Turns Human Videos into Robot Training Data, Boosting Manipulation Success to 80.3%
A new Hugging Face paper introduces HuRo, a robotization pipeline that converts heterogeneous human videos into robot-aligned obse...
Hugging Face Paper Traces Video Diffusion Physics Flaws to RoPE Attention Decay
A new interpretability study on Hugging Face examines how text-to-video diffusion models plan motion, finding that Rotary Position...
1% of Tokens Can Be Enough: New Gradient-Estimation Method Sharpens Sparse On-Policy Distillation
A new paper proposes an information-efficiency ratio (IER) that measures how reliably a token's gradient can be estimated, enablin...
CARE Teaches Robots to Recover: Failure-Driven Correction Boosts VLA Success
Hugging Face paper introduces CARE, a data-centric framework that learns from failed robot rollouts to synthesize corrective demon...
Deep Persona: A Three-Layer Psychological Architecture for More Convincing Role-Playing AI Agents
A new paper proposes Deep Persona, a psychologically grounded three-layered architecture that organizes AI personas into observabl...
Harness-Zero Distills Agent Harness Gains Directly Into Model Weights
A new Hugging Face paper proposes Harness-Zero, a method that transfers the benefits of a specialized agent harness into a model's...
Hugging Face Paper Unveils Jev-Mem, a 'System-One' Memory Layer for Faster AI Agents
Jev-Mem splits agentic memory management into a fast System-One control plane and a slower System-Two reasoning plane, reporting b...
Alibaba's Category-Aware Expert Training Boosts SWE Agents to 58% on Pro-618
A new framework from Alibaba trains category-specific software engineering experts via Refresh-Repair-Expand, then consolidates th...