Songlin Hu

EntropyLong: Effective Long-Context Training via Predictive Uncertainty

Aligning at the Source: Steering Corrective to the Origins of Harmfulness in LLMs

LightRetriever: A LLM-based Text Retrieval Architecture with Extremely Faster Query Inference

Libra-Emo: A Large Dataset for Multimodal Fine-grained Negative Emotion Detection

AgentAlign: Navigating Safety Alignment in the Shift from Informative to Agentic Large Language Models

Mitigating Object Hallucinations in Large Vision-Language Models via multi-scale visual integration

Resolving the Security-Auditability Dilemma with Auditable Latent Chain-of-Thought

Modulating LLM Behavior via Context-Specific Activation Steering

NExtLong: Toward Effective Long-Context Training without Long Documents

LongMagpie: A Self-synthesis Method for Generating Large-scale Long-context Instructions

Quest: Query-centric Data Synthesis Approach for Long-context Scaling of Large Language Model

CodePMP: Scalable Preference Model Pretraining for Large Language Model Reasoning

Beyond Imitation: Learning Key Reasoning Steps from Dual Chain-of-Thoughts in Reasoning Distillation