Cascading Adversarial Bias from Injection to Distillation in Language Models

Name: Cascading Adversarial Bias from Injection to Distillation in Language Models
Uploaded: 2026-01-27T00:00:00.000Z
Duration: 32 min 31 s
Description: Phantom is a trigger-based poisoning attack on retrieval-augmented generation (RAG) systems. The attacker places one crafted document in the knowledge base, and it gets retrieved only when the user's query contains a...

Google TechTalks · 32:31

Phantom is a trigger-based poisoning attack on retrieval-augmented generation (RAG) systems. The attacker places one crafted document in the knowledge base, and it gets retrieved only when the user's query contains a...

Read the full summary on tuber