Yutaka Matsuo

Does “Do Differentiable Simulators Give Better Policy Gradients?” Give Better Policy Gradients?

Quantization-Aware Diffusion Models For Maximum Likelihood Training

C-Voting: Confidence-Based Test-Time Voting without Explicit Energy Functions

SELF-HARMONY: LEARNING TO HARMONIZE SELF-SUPERVISION AND SELF-PLAY IN TEST-TIME REINFORCEMENT LEARNING

RL Squeezes, SFT Expands: A Comparative Study of Reasoning LLMs

Emergence of Exploration in Policy Gradient Reinforcement Learning via Retrying

Geometry of Nash Mirror Dynamics: Adaptive $\beta$-Control for Stable and Bias-Robust Self-Improving LLM Agents

Unlocking Noise-Resistant Vision: Key Architectural Secrets for Robust Models

WorldPack: Compressed Memory Improves Spatial Consistency in Video World Modeling

Leave No Observation Behind: Real-time Correction for VLA Action Chunks

Recurrent model for Sequential reasoning

Improving Dynamic Object Interactions in Text-to-Video Generation with AI Feedback

Emergent Misalignment from Superposition

Vertical Attention: Automatic Exploration of Inter-Layer Connections in Transformer-based Language Models

Provably Efficient RL under Episode-Wise Safety in Constrained MDPs with Linear Function Approximation

Inference-Time Text-to-Video Alignment with Diffusion Latent Beam Search

Topology of Reasoning: Understanding Large Reasoning Models through Reasoning Graph Properties

Beyond Induction Heads: In-Context Meta Learning Induces Multi-Phase Circuit Emergence

Near-Optimal Policy Identification in Robust Constrained Markov Decision Processes via Epigraph Form

Rethinking Evaluation of Sparse Autoencoders through the Representation of Polysemous Words

CityNav: Language-Goal Aerial Navigation Dataset Using Geographic Information

Bridging Lottery Ticket and Grokking: Understanding Grokking from Inner Structure of Networks

FullDiffusion: Diffusion Models Without Time Truncation

ToM-agent: Large Language Models as Theory of Mind Aware Generative Agents with Counterfactual Reflection

The Geometry of Phase Transitions in Diffusion Models: Tubular Neighbourhoods and Singularities

MMA: Benchmarking Multi-Modal Large Language Model in Ambiguity Contexts

RAGDP: Retrieve-Augmented Generative Diffusion Policy

Maximum Likelihood Estimation for Flow Matching by Direct Second-order Trace Objective

Curse of Instructions: Large Language Models Cannot Follow Multiple Instructions at Once