Research Papers
HarnessVLN Brings Training-Free Embodied Navigation Under One Agent Harness
A new Hugging Face paper introduces HarnessVLN, a zero-shot, training-free navigation framework whose Agent Harness coordinates pe...
Hugging Face Paper Proposes Unified Framework for Policy Improvement and Recursive Self-Improvement
A new paper introduces Generalized Agent Iteration (GAI), a formal framework that treats classical iterative policy improvement an...
StepAudio 3 Music Debuts With Explicit Planning and 5-Minute Song Generation
StepAudio 3 Music is a large-scale, long-form music generation model that combines a 50-Hz single-codebook tokenizer, a flow-match...
StepAudio 3 Realtime Brings Think-While-Speaking to Full-Duplex Voice AI
A new technical report introduces StepAudio 3 Realtime, an audio-language foundation model built around a continuous listen-conver...
New Study Maps How Transformer Representations Evolve, With Surprising Training Gains
Researchers decompose learned Transformer updates into parallel and perpendicular components, finding that direction-preserving up...
Hugging Face Paper Proposes Mind2Dialogue to Train Human-Aware AI Assistants
Mind2Dialogue simulates users' unspoken mental states and converts them into privileged supervision, teaching language models to u...
Hugging Face Paper Unveils ScienceBuddy, a Self-Improving Scientific Research Agent
ScienceBuddy is an interactive scientific workspace that turns researchers' requests, feedback, and execution evidence into traini...
Survey Maps Six Roles for AI Across the Game Lifecycle in the Foundation Model Era
A 120-page survey from researchers at the National University of Singapore and Nanyang Technological University organizes how foun...
Hugging Face Paper Maps a Road to Genuine Recursive Self-Improvement in AI
A new survey proposes the Headroom-Closed Index and an L1–L5 taxonomy to chart AI's path from executing human-defined improvements...
Hugging Face Paper: Stacking Continual Learning Tricks Boosts Long-Horizon Memory 28x
A new paper introduces "long-horizon memorization," a setting where a model learns 100 query-answer tasks through continual superv...
Hugging Face Paper Proposes 'Never Give Up' Sampling to Fix RL's Easy-Task Bias in LLMs
A new paper argues that reinforcement learning for LLMs improves easy problems far more than hard ones — a 'Matthew Effect' — and...
Hugging Face Paper Unveils AGP: A General-Purpose Agent That Directly Pilots Robots
A new Hugging Face paper introduces Agent as Policy (AGP), a framework in which a general-purpose coding agent controls a physical...