提出EAR方法，可精准擦除图像模型中的特定概念而不损失生成质量。
EAR: Erasing Concepts from Unified Autoregressive Models
- 采用窗口梯度累积与阈值损失掩码，精准对齐删除目标与生成内容。
- 在Janus-Pro模型上，擦除效果显著提升，且整体生成能力保持良好。
- 构建新基准ECGVF，支持更严格的概念擦除评估，适合模型安全研究者。
自回归（AR）模型在视觉理解与图像生成任务中表现出统一而强大的性能。然而，在保持整体生成质量的前提下，从AR模型中移除不期望的概念仍是一个开放挑战。本文提出擦除自回归模型（EAR），一种用于有效且保留实用性的概念擦除微调方法。具体地，我们引入窗口梯度累积（WGA）策略，使局部解码与擦除目标对齐；并采用阈值损失掩码（TLM）策略，在微调过程中保护与目标概念无关的内容。此外，我们提出一个新型基准——概念擦除生成器与视觉过滤器（ECGVF），旨在为AR模型中的概念擦除提供更严格、全面的评估基础。首先，利用多样化大语言模型结构化模板预生成大规模目标替换概念提示对；随后，基于这些提示生成图像，并通过视觉分类器进行严格筛选以确保概念保真度与一致性。在基于AR模型Janus-Pro的ECGVF基准上的大量实验表明，EAR在擦除有效性与模型实用性保持方面均有显著提升。代码已开源：https://github.com/immc-lab/ear/
原文摘要 · Abstract (English)
Autoregressive (AR) models have achieved unified and strong performance across both visual understanding and image generation tasks. However, removing undesired concepts from AR models while maintaining overall generation quality remains an open challenge. In this paper, we propose Erasure Autoregressive Model (EAR), a fine-tuning method for effective and utility-preserving concept erasure in AR models. Specifically, we introduce Windowed Gradient Accumulation (WGA) strategy to align patch-level decoding with erasure objectives, and Thresholded Loss Masking (TLM) strategy to protect content unrelated to the target concept during fine-tuning. Furthermore, we propose a novel benchmark, Erase Concept Generator and Visual Filter (ECGVF), aim at provide a more rigorous and comprehensive foundation for evaluating concept erasure in AR models. Specifically, we first employ structured templates across diverse large language models (LLMs) to pre-generate a large-scale corpus of target-replacement concept prompt pairs. Subsequently, we generate images from these prompts and subject them to rigorous filtering via a visual classifier to ensure concept fidelity and alignment. Extensive experimental results conducted on the ECGVF benchmark with the AR model Janus-Pro demonstrate that EAR achieves marked improvements in both erasure effectiveness and model utility preservation. Code is available at: https://github.com/immc-lab/ear/
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。