Furu Wei

VibeVoice: Expressive Podcast Generation with Next-Token Diffusion

Synergizing Understanding and Generation with Interleaved Analyzing-Drafting Thinking

VisCodex: Unified Multimodal Code Generation via Merging Vision and Coding Models

From Abstract to Contextual: What LLMs Still Cannot Do in Mathematics

Code Aesthetics with Agentic Reward Feedback

Multimodal Latent Language Modeling with Next-Token Diffusion

11Plus-Bench: Demystifying Multimodal LLM Spatial Reasoning with Cognitive-Inspired Analysis

Learning To Draft: Adaptive Speculative Decoding with Reinforcement Learning

Geometric-Mean Policy Optimization

AlignDiff: Exploiting Model-Intrinsic Information for Better Preference Data Selection

BitNet Distillation

Scaling Laws for Fully Sparsely-Activated Large Language Models

QueST: Incentivizing LLMs to Generate Difficult Problems

Breaking Training Bottlenecks: Effective Reinforcement Learning for Modern Coding Models

Rectified Sparse Attention for Efficient Long-Sequence Generation

WildLong: Synthesizing Realistic Long-Context Instruction Data at Scale

Two Pathways to Truthfulness: On the Intrinsic Encoding of LLM Hallucinations

Towards Stable and Effective Reinforcement learning for Mixture-of-Experts

DocReward: A Document Reward Model for Structuring and Stylizing

Thinking Augmented Pre-training

Learning to Refine: Self-Refinement of Parallel Reasoning in LLMs

Latent Sketchpad: Sketching Visual Thoughts to Elicit Multimodal Reasoning in MLLMs

Information-Preserving Reformulation of Reasoning Traces for Antidistillation

On-Policy RL with Optimal Reward Baseline

Efficient RL Training for Reasoning Models via Length-Aware Optimization

Data Selection via Optimal Control for Language Models

Differential Transformer

Think Only When You Need with Large Hybrid-Reasoning Models

Preference Optimization for Reasoning with Pseudo Feedback

Semi-Parametric Retrieval via Binary Bag-of-Tokens Index

Generative Representational Instruction Tuning

Chain-of-Retrieval Augmented Generation

Self-Boosting Large Language Models with Synthetic Preference Data

Scaling Laws of Synthetic Data for Language Model

Towards Thinking-Optimal Scaling of Test-Time Compute for LLM Reasoning

Reward Reasoning Models

ARLON: Boosting Diffusion Transformers with Autoregressive Models for Long Video Generation

Imagine While Reasoning in Space: Multimodal Visualization-of-Thought

Scaling Optimal LR Across Token Horizons

E5-V: Universal Embeddings with Multimodal Large Language Models

Synthetic Data (Almost) from Scratch: Generalized Instruction Tuning for Language Models

VALL-E 2: Neural Codec Language Models are Human Parity Zero-Shot Text to Speech Synthesizers

Q-Sparse: All Large Language Models can be Fully Sparsely-Activated

MoRA: High-Rank Updating for Parameter-Efficient Fine-Tuning

Textual Aesthetics in Large Language Models

One Language, Many Gaps: Evaluating Dialect Fairness and Robustness of Large Language Models in Reasoning Tasks

Next Block Prediction: Video Generation via Semi-Auto-Regressive Modeling