Di Wang

Controlling Repetition in Protein Language Models

Dual-Kernel Adapter: Expanding Spatial Horizons for Data-Constrained Medical Image Analysis

Predicting LLM Output Length via Entropy-Guided Representations

Understanding and Improving Continuous LLM Adversarial Training via In-context Learning Theory

The Price of Amortized inference in Sparse Autoencoders

When Modalities Conflict: How Unimodal Reasoning Uncertainty Governs Preference Dynamics in MLLMs

Dissecting Representation Misalignment in Contrastive Learning via Influence Function

Evaluating Data Influence in Meta Learning

Mechanistic Analysis of Demonstration Conflicts in In-Context Learning

Understanding Private Learning From Feature Perspective

Benign Overfitting in Adversarial Training for Vision Transformers

Untargeted Jailbreak Attack

MSRS: Adaptive Multi-Subspace Representation Steering for Attribute Alignment in Large Language Models

MONICA: Real-Time Monitoring and Calibration of Chain-of-Thought Sycophancy in Large Reasoning Models

Robust Learning of Diffusion Models with Extremely Noisy Conditions

D-LEAF: Localizing and Correcting Hallucinations in Multimodal LLMs via Layer-to-head Attention Diagnostics

Efficient and Stable Grouped RL Training for Large Language Models

Dynamic Target Attack

Curriculum-RLAIF: Curriculum Alignment with Reinforcement Learning from AI Feedback

Flattery in Motion: Benchmarking and Analyzing Sycophancy in Video-LLMs

CACE-Net: Cascade Coupling Effect for Link Prediction in Multi-layer Networks

Investigating CoT Monitorability in Large Reasoning Models

Attributing Data for Sharpness-Aware Minimization

Pre-trained Encoder Inference: Revealing Upstream Encoders In Downstream Machine Learning Services

TAD-Net: Reinforced Anomaly Generation and Wavelet-enhanced Prediction for Temporal Anomaly Detection

Concept-Based Dictionary Learning for Inference-Time Safety in Vision–Language–Action Models

Flexible Feature Distillation for Large Language Models

CoLa: A Choice Leakage Attack Framework To Expose Privacy Risks In Subset Training

Goal-oriented Backdoor Attack against Vision-Language-Action Models via Physical Objects

Model-agnostic Adversarial Attack and Defense for Vision-Language-Action Models

Backdooring CLIP through Concept Confusion

Compositional Architecture of Regret in Large Language Models

Locate-then-edit for Multi-hop Factual Recall under Knowledge Editing

Second-Order Convergence in Private Stochastic Non-Convex Optimization

Editable Concept Bottleneck Models

Short-length Adversarial Training Helps LLMs Defend Long-length Jailbreak Attacks: Theoretical and Empirical Evidence

EAP-GP: Mitigating Saturation Effect in Gradient-based Automated Circuit Identification

Private Training Large-scale Models with Efficient DP-SGD

Private Stochastic Optimization for Achieving Second-Order Stationary Points

Scalable Zeroth-Order Fine-Tuning for Extremely Large Language Models with Limited GPU Memory

Locate-then-edit for Multi-hop Factual Recall under Knowledge Editing

Towards User-level Private Reinforcement Learning with Human Feedback

Editable Concept Bottleneck Models

Dissecting Misalignment of Multimodal Large Language Models via Influence Function

FlashDP: Memory-Efficient and High-Throughput DP-SGD Training for Large Language Models

Private Stochastic Convex Optimization with Tysbakov Noise Condition and Large Lipschitz Constant

Low-cost Enhancer for Text Attributed Graph Learning via Graph Alignment

Representation Confusion: Towards Representation Backdoor on CLIP via Concept Activation

XTraffic: A Dataset Where Traffic Meets Incidents with Explainability and More

ZO-Offloading: Fine-Tuning LLMs with 100 Billion Parameters on a Single GPU

Understanding Reasoning in Chain-of-Thought from the Hopfieldian View

What Makes Your Model a Low-empathy or Warmth Person: Exploring the Origins of Personality in LLMs