OpenGU: A Comprehensive Benchmark for Graph Unlearning

On the Interplay of Pre-Training, Mid-Training, and RL on Reasoning Language Models

DeepCode: Open Agentic Coding

InfiniteVL: Synergizing Linear and Sparse Attention for Highly-Efficient, Unlimited-Input Vision-Language Models

OmniPSD: Layered PSD Generation with Diffusion Transformer

HiF-VLA: Hindsight, Insight and Foresight through Motion Representation for Vision-Language-Action Models

Arbitrage: Efficient Reasoning via Advantage-Aware Speculation

Composing Concepts from Images and Videos via Concept-prompt Binding

StereoWorld: Geometry-Aware Monocular-to-Stereo Video Generation

Urania: Differentially Private Insights into AI Use

Training LLMs for Honesty via Confessions

Measuring Agents in Production

PolyMath: Evaluating Mathematical Reasoning in Multilingual Contexts

ThreadWeaver: Adaptive Threading for Efficient Parallel Reasoning in Language Models

SPARK: Stepwise Process-Aware Rewards for Reference-Free Reinforcement Learning

OneStory: Coherent Multi-Shot Video Generation with Adaptive Memory

Preserving Source Video Realism: High-Fidelity Face Swapping for Cinematic Quality

Visionary: The World Model Carrier Built on WebGPU-Powered Gaussian Splatting Platform

Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance

Soft Adaptive Policy Optimization

Scaling Zero-Shot Reference-to-Video Generation

Voxify3D: Pixel Art Meets Volumetric Rendering

DoVer: Intervention-Driven Auto Debugging for LLM Multi-Agent Systems

Unified Video Editing with Temporal Reasoner

Beyond Real: Imaginary Extension of Rotary Position Embeddings for Long-Context LLMs

Native Parallel Reasoner: Reasoning in Parallelism via Self-Distilled Reinforcement Learning

iSeal: Encrypted Fingerprinting for Reliable LLM Ownership Verification

DAVSP: Safety Alignment for Large Vision-Language Models via Deep Aligned Visual Safety Prompt

WorldGen: From Text to Traversable and Interactive 3D Worlds

Souper-Model: How Simple Arithmetic Unlocks State-of-the-Art LLM Performance

OpenGU: A Comprehensive Benchmark for Graph Unlearning

On the Interplay of Pre-Training, Mid-Training, and RL on Reasoning Language Models

DeepCode: Open Agentic Coding

InfiniteVL: Synergizing Linear and Sparse Attention for Highly-Efficient, Unlimited-Input Vision-Language Models

OmniPSD: Layered PSD Generation with Diffusion Transformer

HiF-VLA: Hindsight, Insight and Foresight through Motion Representation for Vision-Language-Action Models

Arbitrage: Efficient Reasoning via Advantage-Aware Speculation

Composing Concepts from Images and Videos via Concept-prompt Binding

StereoWorld: Geometry-Aware Monocular-to-Stereo Video Generation

Urania: Differentially Private Insights into AI Use

Training LLMs for Honesty via Confessions

Measuring Agents in Production

PolyMath: Evaluating Mathematical Reasoning in Multilingual Contexts

ThreadWeaver: Adaptive Threading for Efficient Parallel Reasoning in Language Models

SPARK: Stepwise Process-Aware Rewards for Reference-Free Reinforcement Learning

OneStory: Coherent Multi-Shot Video Generation with Adaptive Memory

Preserving Source Video Realism: High-Fidelity Face Swapping for Cinematic Quality

Visionary: The World Model Carrier Built on WebGPU-Powered Gaussian Splatting Platform

Wan-Move: Motion-controllable Video Generation via Latent Trajectory Guidance

Soft Adaptive Policy Optimization

Scaling Zero-Shot Reference-to-Video Generation

Voxify3D: Pixel Art Meets Volumetric Rendering

DoVer: Intervention-Driven Auto Debugging for LLM Multi-Agent Systems

Unified Video Editing with Temporal Reasoner

Beyond Real: Imaginary Extension of Rotary Position Embeddings for Long-Context LLMs

Native Parallel Reasoner: Reasoning in Parallelism via Self-Distilled Reinforcement Learning

iSeal: Encrypted Fingerprinting for Reliable LLM Ownership Verification

DAVSP: Safety Alignment for Large Vision-Language Models via Deep Aligned Visual Safety Prompt

WorldGen: From Text to Traversable and Interactive 3D Worlds

Souper-Model: How Simple Arithmetic Unlocks State-of-the-Art LLM Performance