All papers

Cancerous Backdoor Attack changes LLM backdoors from one-time activation to multi-stage reasoning corruption that accumulates progressively with triggers.

DIVA 是针对带参考图的视频生成模型的免训练 black-box 越狱：用一致性机制本身堵住语义漂移这条安全逃逸路径。。

SinoGlyphBench是一个配对诊断基准，用来看内容审核在可还原的中文字形混淆下是否还保持原决定。

DSR 是一种 black-box 跨模态越狱：输入侧不放显式有害内容，而让商业 MLLM 在推理中把无害图文重组成有害图像。

四类视觉越狱用来检验前沿 VLM 的文本安全训练能否挡住由图像传达的有害意图。。

StructAttack is a black-box, single-query jailbreak against LVLMs: it decomposes a harmful instruction into separate seemingly harmless slots, draws them into a structural diagram, and then makes the model reassemble the slots into harmful content.

COMET is a black-box jailbreak for VLMs with multimodal reasoning, used to red-team cross-modal safety alignment. The gap the authors address is: distributing malicious cues into fixed combinations of image and text cannot detect vulnerabilities that only appear in multi-step reasoning. The attack does not access gradients; it only constructs image and text inputs.

作者把这篇工作定位为立场论文：再摄入存在时，按无状态设计的 LLM 仍可能经自身输出携带状态。要处理的问题是，这种不依赖外部记忆模块的通道会让风险跨过单次请求，而单触发测试看不到完整条件。

BVS 用图文对诱导 MLLM 生成有害图像。作者认为，相对文本越狱和单模态图像诱导，视觉生成的安全审查仍不充分。