共计 52 篇文章

---

2025

12-05

五种LORA

12-04

Soft Instruction De-Escalation Defense

12-03

Jigsaw-Agile Community Rules Classification第一名方案

12-02

DoRA： Weight-Decomposed Low-Rank Adaptation

11-29

IBD：通过图像有偏解码减轻大型视觉-语言模型中的幻觉

11-28

Implicit Bias Injection Attacks against Text-to-Image Diffusion Models

11-27

VASparse：通过视觉感知的 token 稀疏化实现高效视觉幻觉缓解

11-26

Be My Eyes：通过多智能体协作将大型语言模型扩展到新模态

11-24

把MoE整合进LLaVA

11-18

UpSafe℃： Upcycling for Controllable Safety in Large Language Models