Light Alignment Improves LLM Safety via Model Self-Reflection with a Single Neuron (arxiv 2026) 2026-02-12 #深度学习 #大模型
FraudShield: Knowledge Graph Empowered Defense for LLMs against Fraud Attacks (WWW 2026) 2026-02-11 #深度学习 #大模型
PoisonedEye:Knowledge Poisoning Attack on Retrieval-Augmented Generation based Large Vision-Language Models (ICML 2025) 2026-02-09 #深度学习 #RAG
TraceDet: Hallucination Detection from the Decoding Trace of Diffusion Large Language Models (ICLR 2026) 2026-02-09 #深度学习 #大模型
REACT:SYNERGIZING REASONING AND ACTING IN LANGUAGE MODELS (ICLR 2023) 姚顺雨腾讯实习之作。 2026-02-08 #深度学习 #大模型 #agent
RAGLens: Toward Faithful Retrieval-Augmented Generation with Sparse Autoencoders (ICLR 2026) 2026-02-08 #深度学习 #大模型 #RAG
Why Steering Works:Toward a Unified View of Language Model Parameter Dynamics (arxiv 2026) 将全量微调、LoRA、激活干预统一。 很好的文章! 2026-02-06 #深度学习 #大模型
LLM-VA: Resolving the Jailbreak-Overrefusal Trade-off via Vector Alignment (arxiv 2026) Large Language Model Vector Alignment (LLM-VA) 2026-02-06 #深度学习 #大模型
AlphaSteer: Learning Refusal Steering with Principled Null-Space Constraint (ICLR 2026) 之前看alphaedit和nullu就有过将二者融合起来。没想到已经有人做了。 2026-02-06 #深度学习 #大模型