Yang Liu

SteeringSafety: A Systematic Safety Evaluation Framework of Representation Steering in LLMs

SceneTransporter: Optimal Transport-Guided Compositional Latent Diffusion for Single-Image Structured 3D Scene Generation

DRAGON: Guard LLM Unlearning in Context via Negative Detection and Reasoning

Skywork-Reward-V2: Scaling Preference Data Curation via Human-AI Synergy

RepIt: Steering Language Models with Concept-Specific Refusal Vectors

Compressing Large MoE Models via Efficient Pruning and Data-Aware Calibration

Projection-Domain Adaptation for 3D Transformers: From Perspective to Panoramic Scene Reconstruction

FEWL: Measuring and Mitigating LLM Hallucination Without Gold-Standard Answers

Label Smoothing Improves Machine Unlearning

Unlearning Isn't Deletion: Investigating Reversibility of Machine Unlearning in LLMs

AgentOrchestra: Orchestrating Hierarchical Multi-Agent Intelligence with the Tool-Environment-Agent (TEA) Protocol

Geo-Refine: Geometry–Appearance Synergy for Robust Single-Image 3D Scene Generation

Adversarial Test Case Generation via Reinforcement Learning Extends Scaling Laws

Supervised Fine-Tuning on Ambiguous Preference Pairs Boosts LLM Alignment

DRAGON: Guard LLM Unlearning in Context via Negative Detection and Reasoning

Learning Counterfactual Outcomes Under Rank Preservation

Evaluating LLM-contaminated Crowdsourcing Data Without Ground Truth

Regretful Decisions under Label Noise

LLM Unlearning via Loss Adjustment with Only Forget Data

Understanding Chain-of-Thought in LLMs Through Information Theory

Incentivizing LLMs to Self-Verify Their Answers

DAPO : Improving Multi-Step Reasoning Abilities of Large Language Models with Direct Advantage-Based Policy Optimization

Adversarial Machine Unlearning

Noisy Test-Time Adaptation in Vision-Language Models

Automatic Dataset Construction (ADC): Sample Collection, Data Curation, and Beyond

ACC-Collab: An Actor-Critic Approach to Multi-Agent LLM Collaboration

Improving Data Efficiency via Curating LLM-Driven Rating Systems

Revisiting inverse Hessian vector products for calculating influence functions

Understanding Chain-of-Thought in LLMs through Information Theory

Robust Multi-bit Text Watermark with LLM-based Paraphrasers

Token Cleaning: Fine-Grained Data Selection for LLM Supervised Fine-Tuning

Prompting Fairness: Integrating Causality to Debias Large Language Models

Learning to Watermark LLM-generated Text via Reinforcement Learning

DiffTell: A Comprehensive Dataset for Image Difference Captioning

Robust RLHF with Noisy Rewards

Progressive LLM Alignments Using Two-Player Games

Counterfactual Learning under Rank Preservation

Human-Instruction-Free LLM Self-Alignment with Limited Samples