arXiv Artificial Intelligence

Token by Token, Compromised: Backdoor Vulnerabilities in Unified Autoregressive Models

1 hour ago 09.10.2026 07:00 👁 1 views
Token by Token, Compromised: Backdoor Vulnerabilities in Unified Autoregressive Models

Quick summary

arXiv:2605.19227v2 Announce Type: replace-cross Abstract: Unified autoregressive models (UAMs) are transformer models that generate text as well as image tokens within a single autoregressive pass. Shared parameters and a multimodal vocabulary simplify the training pipeline and facilitate flexible multimodal generation, yet might introduce new vulnerabilities. In particular, we are the first to show that this unified architecture enables multimodal backdoor attacks, where a trigger can propagate malicious effects across multiple output modalities. Specifically, we present the Token by Token Ba

Key takeaways

  • arXiv:2605.19227v2 Announce Type: replace-cross Abstract: Unified autoregressive models (UAMs) are transformer models that generate text as well as image tokens within a single autoregressive pass.
  • Shared parameters and a multimodal vocabulary simplify the training pipeline and facilitate flexible multimodal generation, yet might introduce new vulnerabilities.
  • In particular, we are the first to show that this unified architecture enables multimodal backdoor attacks, where a trigger can propagate malicious effects across multiple output modalities.

Why it matters

“Token by Token, Compromised: Backdoor Vulnerabilities in Unified Autoregressive Models” should be evaluated beyond branding and benchmark scores. Its practical importance will emerge in task accuracy, latency, unit cost, safety and integration with real workflows.

Kaynak sitede devamını oku: arXiv Artificial Intelligence ↗