Large Flow Language Models
Continuous flow language models offer a path to parallel generation with tools for control and acceleration unavailable to discrete diffusion, yet effective techniques for building...
Papers
Matrix-Game 4.0: Harnessing Video Models for Real-Time Playable Game Worlds
Moving from controllable videos to genuinely playable worlds requires persistent world state, executable game rules, and low-latency interaction. We outline a roadmap for the next ...
HyperDrone: Simulation-to-Reality Transfer at Millions of Frames per Second
We present HyperDrone, an extremely fast quadrotor simulator focused on high-performance batch rendering that is more than 20x faster than existing state-of-the-art batch quadrotor...
OvisOCR3 Technical Report
We introduce OvisOCR3, a 0.8B model for a broad range of OCR-related tasks. OvisOCR3 unifies document parsing, general OCR, key information extraction (KIE), document figure captio...
Stratified Sparse Attention: Reducing the Sparse-to-Dense Gap at Extreme Sparsity
Large Language Model inference over long contexts is increasingly bottlenecked by the cost of attending to a growing Key-Value (KV) cache. Sparse attention reduces this cost by acc...
HABIT-Bench: Evaluating Habit Induction from Longitudinal Weak Evidence in Agent Memory
In long-term interactions with large language model (LLM) agents, users reveal habits through their choices and actions. These habits capture implicit user information that agents ...
Training the Whole Agent, Not Just the Model
Training coding agents typically updates the policy while keeping the harness fixed. During training, trajectories reveal flawed repair strategies and interaction problems that var...
ProVe: Proposer-Verifier Decoding
Speculative decoding accelerates autoregressive (AR) language models by letting a small model draft tokens serially and a large model verify them in parallel. We reverse these role...
Ask to Distill: Learning to Synthesize Data for On-Policy Distillation
On-policy distillation (OPD) can succeed with surprisingly few training questions, motivating a view of questions as carriers of distributional knowledge transfer. How can we asses...
Self-Credit: Credit Assignment for Long-Horizon Agentic RL via Intra-group Rescaling
LLM agents are commonly trained with RLVR, but the sparse nature of typical RLVR rewards makes credit assignment challenging, especially for long-horizon tasks. We propose Self-Cre...
Answering in the User's Variety: Calibrated Accommodation Pairs for Post-Training
People adapt how they talk to the person in front of them. However, chat models tend to respond in Standard English regardless of the dialect used by the user. A uniform answer can...
Meta Expert: Cross-Layer Expert Sharing for Storage- and Retrieval-Efficient MoE
As model parameters scale up, sparsely activated Mixture-of-Experts (MoE) approaches have become essential for model scaling. However, mainstream models currently adopt hierarchica...
Code2Video-Compass: A Comprehensive Benchmark for General-Purpose Code-Based Video Generation across T2V, I2V, and V2V
Code-based video generation offers fine-grained control, targeted editing, and component reuse compared with direct video generation. This approach requires multimodal large langua...
Scaling in Games: Continued Pre-Training For Embodied Agents In Diverse Virtual Worlds
Scaling data, diversity and model capacity has improved embodied learning in physical environments, where diverse experiences remain grounded in shared physical laws. Whether these...
Building AI Research Teams for Autonomous Discovery of Linear Attention Architectures
Autonomous agents can increasingly propose ideas, modify code, and run experiments, creating an opportunity to accelerate scientific discovery using teams of collaborative AI resea...
Qwen-Audio-3.1-ASR Technical Report
We present Qwen-Audio-3.1-ASR, a family of audio-language models that builds upon its predecesor Qwen-Audio-3.0-ASR with improved speech recognition and expanded capabilities in sp...
Truncate or Not? A Scaling Study of Looped Transformers with Truncated Backpropagation
Looped transformers reuse a shared block of layers to increase computational depth without increasing the number of trainable parameters. They are trained with backpropagation thro...
Not All Features Are Created Equal: A Mechanistic Study of Vision-Language-Action Models
A fine-tuned Vision-Language-Action (VLA) policy can easily place the alphabet soup in the basket on command, just as it saw during fine-tuning. However, if the basket is shifted f...
Real In, Real Out: What If We Only Use Real Data for Scene Text Editing?
Aiming to modify textual content while preserving the original style, Scene Text Editing (STE) has significant practical value in many applications. STE typically requires paired i...
StreamEval: Evaluating Interactive Streaming Video Generation Models
As streaming video generation advances, enabling real-time interaction with generated video is becoming increasingly important. We introduce StreamEval, a unified benchmark for int...
DuoLM: Fast and Effective Personalized Recommendation with Language Models
Language models can make personalized recommendations from rich item content, user history, and feedback. However, evaluating many candidates with full-text inputs incurs substanti...
StyleStream 2.0: Fast and Controllable Streaming Voice Style Conversion
Voice style conversion (VSC) aims to transform a source utterance to match the timbre, accent, and emotion of a target voice while preserving linguistic content. StyleStream 1.0 in...
Evolving Agent Interfaces
Computer-use agents are limited by the interfaces humans give them. An agent may type a spreadsheet cell by cell when a single line of code would do, or print all two thousand line...
WORTH Benchmark: Evidence Discovery under Not-Fully-Cooperative Information -- A Two-Episode Pilot
Existing reasoning evaluations typically provide task-relevant information to models as available premises, rarely examining concurrently how models discover evidence autonomously ...