When to Memorize and When to Stop: Gated Recurrent Memory for Long-Context Reasoning

ASA: Activation Steering for Tool-Calling Domain Adaptation

PhyCritic: Multimodal Critic Models for Physical AI

GENIUS: Generative Fluid Intelligence Evaluation Suite

Step 3.5 Flash: Open Frontier-Level Intelligence with 11B Active Parameters

World-VLA-Loop: Closed-Loop Learning of Video World Model and VLA Policy

Towards Autonomous Mathematics Research

Agent World Model: Infinity Synthetic Environments for Agentic Reinforcement Learning

P1-VL: Bridging Visual Perception and Scientific Reasoning in Physics Olympiads

Chain of Mindset: Reasoning with Adaptive Cognitive Modes

UI-Venus-1.5 Technical Report

Code2World: A GUI World Model via Renderable Code Generation

OPUS: Towards Efficient and Principled Data Selection in Large Language Model Pre-training in Every Iteration

BagelVLA: Enhancing Long-Horizon Manipulation via Interleaved Vision-Language-Action Generation

THINGS-data: A multimodal collection of large-scale datasets for investigating object representations in human brain and behavior

Accurate Predictions of Novel Biomolecular Interactions with IsoDDE

SKILLRL: Evolving Agents via Recursive Skill-Augmented Reinforcement Learning

LLaDA2.1: Speeding Up Text Diffusion via Token Editing

Alleviating Sparse Rewards by Modeling Step-Wise and Long-Term Sampling Effects in Flow-Based GRPO

Recurrent-Depth VLA: Implicit Test-Time Compute Scaling of Vision-Language-Action Models via Latent Iterative Reasoning

QuantaAlpha: An Evolutionary Framework for LLM-Driven Alpha Mining

Modality Gap-Driven Subspace Alignment Training Paradigm For Multimodal Large Language Models

MOVA: Towards Scalable and Synchronized Video-Audio Generation

MemoryLLM: Plug-n-Play Interpretable Feed-Forward Memory for Transformers

DreamDojo: A Generalist Robot World Model from Large-Scale Human Videos

F-GRPO: Don't Let Your Policy Learn the Obvious and Forget the Rare

MSign: An Optimizer Preventing Training Instability in Large Language Models via Stable Rank Restoration

AudioSAE: Towards Understanding of Audio-Processing Models with Sparse AutoEncoders

On the Entropy Dynamics in Reinforcement Fine-Tuning of Large Language Models

OdysseyArena: Benchmarking Large Language Models For Long-Horizon, Active and Inductive Interactions

When to Memorize and When to Stop: Gated Recurrent Memory for Long-Context Reasoning

ASA: Activation Steering for Tool-Calling Domain Adaptation

PhyCritic: Multimodal Critic Models for Physical AI

GENIUS: Generative Fluid Intelligence Evaluation Suite

Step 3.5 Flash: Open Frontier-Level Intelligence with 11B Active Parameters

World-VLA-Loop: Closed-Loop Learning of Video World Model and VLA Policy

Towards Autonomous Mathematics Research

Agent World Model: Infinity Synthetic Environments for Agentic Reinforcement Learning

P1-VL: Bridging Visual Perception and Scientific Reasoning in Physics Olympiads

Chain of Mindset: Reasoning with Adaptive Cognitive Modes

UI-Venus-1.5 Technical Report

Code2World: A GUI World Model via Renderable Code Generation

OPUS: Towards Efficient and Principled Data Selection in Large Language Model Pre-training in Every Iteration

BagelVLA: Enhancing Long-Horizon Manipulation via Interleaved Vision-Language-Action Generation

THINGS-data: A multimodal collection of large-scale datasets for investigating object representations in human brain and behavior

Accurate Predictions of Novel Biomolecular Interactions with IsoDDE

SKILLRL: Evolving Agents via Recursive Skill-Augmented Reinforcement Learning

LLaDA2.1: Speeding Up Text Diffusion via Token Editing

Alleviating Sparse Rewards by Modeling Step-Wise and Long-Term Sampling Effects in Flow-Based GRPO

Recurrent-Depth VLA: Implicit Test-Time Compute Scaling of Vision-Language-Action Models via Latent Iterative Reasoning

QuantaAlpha: An Evolutionary Framework for LLM-Driven Alpha Mining

Modality Gap-Driven Subspace Alignment Training Paradigm For Multimodal Large Language Models

MOVA: Towards Scalable and Synchronized Video-Audio Generation

MemoryLLM: Plug-n-Play Interpretable Feed-Forward Memory for Transformers

DreamDojo: A Generalist Robot World Model from Large-Scale Human Videos

F-GRPO: Don't Let Your Policy Learn the Obvious and Forget the Rare

MSign: An Optimizer Preventing Training Instability in Large Language Models via Stable Rank Restoration

AudioSAE: Towards Understanding of Audio-Processing Models with Sparse AutoEncoders

On the Entropy Dynamics in Reinforcement Fine-Tuning of Large Language Models

OdysseyArena: Benchmarking Large Language Models For Long-Horizon, Active and Inductive Interactions