Xueqi Cheng

Fine-tuning Done Right in Model Editing

Beyond Text-Only: Towards Multimodal Table Retrieval in Open-World

Data, Not Model: Explaining Bias toward LLM Texts in Neural Retrievers

Annotation-Efficient Honesty Alignment via Confidence Elicitation and Calibration

Parameters vs. Context: Fine-Grained Control of Knowledge Reliance in Language Models

DIVA-GRPO: Enhancing Multimodal Reasoning through Difficulty-Adaptive Variant Advantage

RefineX: Learning to Refine Pre-training Data at Scale from Expert-Guided Programs

How Knowledge Popularity Influences and Enhances LLM Knowledge Boundary Perception

Does Generative Retrieval Break through the Limitations of Dense Retrieval?

Latent Reasoning in LLMs as a Vocabulary-Space Superposition

Focusing by Contrastive Attention: Enhancing VLMs' Visual Reasoning

a1: Steep Test-time Scaling Law via Environment Augmented Generation

Estimating Commonsense Plausibility through Semantic Shifts

Innate Reasoning is Not Enough: In-Context Learning Enhances Reasoning Large Language Models with Less Overthinking

Lost in Time: Systematic Temporal Bias in Large Audio Language Models

PilotRAG: Teaching LLMs Multi-Turn Hybrid RAG via Reinforcement Learning

SCD: Soft $top-k$ Contrastive Decoding for Universal LLM Detoxification

GLOSE: Understanding and Mitigating Toxicity in LLMs via Global Toxic Subspace

Inference-time Alignment in Continuous Space

Is Factuality Enhancement a Free Lunch For LLMs? Better Factuality Can Lead to Worse Context-Faithfulness

Cross-Modal Safety Mechanism Transfer in Large Vision-Language Models

A Theory for Token-Level Harmonization in Retrieval-Augmented Generation

CLIPure: Purification in Latent Space via CLIP for Adversarially Robust Zero-Shot Classification

Everything is Editable: Extend Knowledge Editing to Unstructured Data in Large Language Models

HiddenGuard: Fine-Grained Safe Generation with Specialized Representation Router

Decoding by Contrasting Knowledge: Enhancing LLMs' Confidence on Edited Facts

Visual Transformation Telling

Latent Variable Identifiability in Nonlinear Causal Models with Single-domain Data under Minimality Condition