BriefGPT - AI 论文速递 BriefGPT - AI 论文速递 ·

Confidence-aware Denoised Fine-tuning Method for Enhancing Certified Robustness of Off-the-shelf Models

💡 原文英文，约100词，阅读约需1分钟。
📝

内容提要

本研究提出了一种基于置信度的去噪微调方法（FT-CADIS），旨在提升现成分类器在对抗攻击下的认证鲁棒性。该方法通过识别去噪过程中的幻觉图像，提高微调的稳定性和鲁棒性，并在多项基准测试中取得领先表现。

🏷️

标签

models 去噪微调 对抗攻击 幻觉图像 置信度 认证鲁棒性
➡️

继续阅读

  1. “Be transparent only if asked”: OpenAI’s models learned to leave notes for their future selves
    OpenAI revealed Wednesday evening that some GPT-5.6 Sol model instances, duri...
  2. Turbo build machines can now be enabled per deployment
    You can now opt into Turbo build machines on any individual deployment. This ...
  3. Making global data easier to explore
    UN System Data Commons Data webpage
  4. A CERN for AI-assisted science?
    [This is a guest post by Dimitris Koukoulopoulos. This blog post was initiall...
  5. The AI Superintelligence Slowdown
    Remember when tech leaders would tell their employees to “move fast and break...
  6. Joshua Drake: pgColumnar 1.0-alpha4 released
    SourceDocsProduct page2026-09-17pgColumnar 1.0-alpha4 release notesRelease da...
👤 个人中心
在公众号发送验证码完成验证