Masashi Sugiyama

Decoupling the Class Label and the Target Concept in Machine Unlearning

Towards Scalable Oversight via Partitioned Human Supervision

Accessible, Realistic, and Fair Evaluation of Positive-Unlabeled Learning Algorithms

Rethinking Consistent Multi-Label Classification Under Inexact Supervision

Weak-to-Strong Diffusion with Reflection

Practical estimation of the optimal classification error with soft labels and calibration

Towards Understanding Valuable Preference Data for Large Language Model Alignment

LLM Routing with Dueling Feedback

Unlocking the Power of Co-Occurrence in CLIP: A DualPrompt-Driven Method for Training-Free Zero-Shot Multi-Label Classification

Are Large Vision-Language Models Good Annotators for Image Tagging?

Accelerating Discrete Diffusion Models with Parallel Sampling

Learning from Label Proportions via Proportional Value Classification

Bilateral Information-aware Test-time Adaptation for Vision-Language Models

Unsupervised Prompt Learning with Few-shot Examples for Answering Objective Questions

Sequence Length Matters in Data Scheduling for Accelerating Language Model Pretraining

Towards Scalable Oversight with Collaborative Multi-Agent Debate in Error Detection

What Is Preference Optimization Doing, How and Why?

From Coefficients to Directions: Rethinking Model Merging with Directional Alignment

CARPRT: Class-Aware Zero-Shot Prompt Reweighting for Vision-Language Model

Learning Robust Diffusion Models from Imprecise Supervision

Rethinking Toxicity Evaluation in Large Language Models: A Multi-Label Perspective

Color Blindness Test Images as Seen by Large Vision-Language Models

Reinforcement Learning with Verifiable yet Noisy Rewards under Imperfect Verifiers

Parallel Simulation for Log-concave Sampling and Score-based Diffusion Models

Generalized Linear Bandits: Almost Optimal Regret with One-Pass Update

Adaptive Localization of Knowledge Negation for Continual LLM Unlearning

Realistic Evaluation of Deep Partial-Label Learning Algorithms

The Adaptive Complexity of Minimizing Relative Fisher Information

The adaptive complexity of parallelized log-concave sampling

Off-Policy Corrected Reward Modeling for Reinforcement Learning from Human Feedback

Sharpness-Aware Black-Box Optimization

Towards Out-of-Modal Generalization without Instance-level Modal Correspondence

Decoupling the Class Label and the Target Concept in Machine Unlearning

CARPRT: Class-Aware Prompt Reweighting for Pre-Trained Vision-Language Models

Parallel simulation for sampling under isoperimetry and score-based diffusion models

On the Cost-Effectiveness of Partially-Annotating Methods for Multi-Label Learning

Towards Effective Evaluations and Comparisons for LLM Unlearning Methods

Learning View-invariant World Models for Visual Robotic Manipulation

Causal Graph Learning via Distributional Invariance of Cause-Effect Relationship

Delving into Weakly Supervised Learning with Pre-Trained Models

Non-stationary Online Learning for Curved Losses: Improved Dynamic Regret via Mixability

Beyond Simple Sum of Delayed Rewards: Non-Markovian Reward Modeling for Reinforcement Learning

On Unsupervised Prompt Learning for Classification with Black-box Language Models