Defu Lian

Is Softmax Loss all you need? A Principled Analysis of Softmax Loss and its Variants

MR$^2$-Bench: Going Beyond Matching to Reasoning in Multimodal Retrieval

GTM: A General Time-series Model for Enhanced Representation Learning of Time-Series data

Scaling Reasoning Hop Exposes Weaknesses: Demystifying and Improving Hop Generalization in Large Language Models

Evolving Graph Structured Programs for Circuit Generation with Large Language Models

ChemEval: A Multi-level and Fine-grained Chemical Capability Evaluation for Large Language Models

EditScore: Unlocking Online RL for Image Editing via High-Fidelity Reward Modeling

CoSteer: Collaborative Decoding-Time Personalization via Local Delta Steering

Preference Optimization via Key-step Error Exploration for Multi-step Reasoning in LLMs

Revisiting Overestimation Bias of Q-learning: Breaking Bias Propagation Chains Does Well

Retro*: Optimizing LLMs for Reasoning-Intensive Document Retrieval

Beyond Raw Detection Scores: Markov-Informed Calibration for Boosting Machine-Generated Text Detection

Advancing and Benchmarking Personalized Tool Invocation for LLMs

Can Recommender Systems Teach Themselves? A Recursive Self-Improving Framework with Fidelity Control

Machine-Generated Text Detection Requires Fewer Machine-Human Mixed Texts

MAR: Medical Asymmetric Retriever for Efficient Chinese Medical Dense Retrieval

Unify in Isolation: Unified Data Synthesis for Divergent Multi-Stage Systems

Antagonistic Evolution for LLM Tool Use

Unlocking the Power of Function Vectors for Characterizing and Mitigating Catastrophic Forgetting in Continual Instruction Tuning

LogicTree: Improving Complex Reasoning of LLMs via Instantiated Multi-step Synthetic Logical Data

Advancing Machine-Generated Text Detection from an Easy to Hard Supervision Perspective

Making Text Embedders Few-Shot Learners

P-Law: Predicting Quantitative Scaling Law with Entropy Guidance in Large Recommendation Models

Accurate KV Cache Eviction via Anchor Direction Projection for Efficient LLM Inference

ToolACE: Winning the Points of LLM Function Calling

RecFlow: An Industrial Full Flow Recommendation Dataset

HyperTree Planning: Enhancing LLM Reasoning via Hierarchical Thinking

Making Transformer Decoders Better Differentiable Indexers

TDDBench: A Benchmark for Training data detection

Multi-play Multi-armed Bandit Model with Scarce Sharable Arm Capacities

From Feature Interaction to Feature Generation: A Generative Paradigm of CTR Prediction Models

Harnessing Query Heterogeneity for Cost-Effective Proactive Caching in LLM Inference

Multiple-play Stochastic Bandits with Prioritized Resource Sharing

Strengthening Federated Learning: Surrogate Data-Guided Aggregation for Robust Backdoor Defense