Introduces Neighbor Leakage Rate showing high trigger leakage in VLAS backdoors at 3% poisoning, caused by broad activation regions in fine-tuning that hard-negative samples can narrow.
Test-time backdoor attacks on multimodal large language models
8 Pith papers cite this work. Polarity classification is still indexing.
8
Pith papers citing it
citation-role summary
background 2
citation-polarity summary
verdicts
UNVERDICTED 8roles
background 2polarities
background 2representative citing papers
EntropyScan detects backdoored LVLMs by quantifying structural anomalies in visual attention distributions on benign samples via Tsallis entropy and reference-anchored Z-score normalization.
ProjLens shows that backdoor parameters in MLLMs are encoded in low-rank subspaces of the projector and that embeddings shift toward the target direction with magnitude linear in input norm, activating only on poisoned samples.
ReShift is a reasoning-level backdoor framework for VLMs that uses poisoned data construction and joint optimization to shift CoT trajectories on trigger while preserving surface coherence.
BYORn defends autoregressive vision-language models against backdoor attacks in supervised fine-tuning by dynamically replacing semantically implausible poisoned responses with model-generated alternatives, improving robustness while preserving clean performance.
Phantasia is a new backdoor attack on VLMs that dynamically aligns malicious outputs with input context to achieve higher stealth and state-of-the-art success rates compared to static-pattern attacks.
GLA backdoor attack on DriveVLM uses naturalistic graffiti and cross-lingual triggers to reach 90% ASR at 10% poisoning ratio while improving some clean-task metrics like BLEU-1.
A comprehensive survey that taxonomizes safety threats to large models and agents, reviews defenses and benchmarks, and outlines open challenges.
citing papers explorer
-
Beyond Attack Success Rate: Examining Trigger Leakage in Vision-Language Agentic Systems
Introduces Neighbor Leakage Rate showing high trigger leakage in VLAS backdoors at 3% poisoning, caused by broad activation regions in fine-tuning that hard-negative samples can narrow.
-
EntropyScan: Towards Model-level Backdoor Detection in LVLMs via Visual Attention Entropy
EntropyScan detects backdoored LVLMs by quantifying structural anomalies in visual attention distributions on benign samples via Tsallis entropy and reference-anchored Z-score normalization.
-
ProjLens: Unveiling the Role of Projectors in Multimodal Model Safety
ProjLens shows that backdoor parameters in MLLMs are encoded in low-rank subspaces of the projector and that embeddings shift toward the target direction with magnitude linear in input norm, activating only on poisoned samples.
-
ReShift: Aha-Moment-Driven Reasoning-Level Backdoor Attacks on Vision-Language Models
ReShift is a reasoning-level backdoor framework for VLMs that uses poisoned data construction and joint optimization to shift CoT trajectories on trigger while preserving surface coherence.
-
BYORn: Bootstrap Your Own Responses to Defend Large Vision-Language Models Against Backdoor Attacks
BYORn defends autoregressive vision-language models against backdoor attacks in supervised fine-tuning by dynamically replacing semantically implausible poisoned responses with model-generated alternatives, improving robustness while preserving clean performance.
-
Phantasia: Context-Adaptive Backdoors in Vision Language Models
Phantasia is a new backdoor attack on VLMs that dynamically aligns malicious outputs with input context to achieve higher stealth and state-of-the-art success rates compared to static-pattern attacks.
-
Multimodal Backdoor Attack on VLMs for Autonomous Driving via Graffiti and Cross-Lingual Triggers
GLA backdoor attack on DriveVLM uses naturalistic graffiti and cross-lingual triggers to reach 90% ASR at 10% poisoning ratio while improving some clean-task metrics like BLEU-1.
-
Safety at Scale: A Comprehensive Survey of Large Model and Agent Safety
A comprehensive survey that taxonomizes safety threats to large models and agents, reviews defenses and benchmarks, and outlines open challenges.