AINewsnow

Retrieval Is Not a Safety Boundary: Poisoned-Corpus and Unsupported-Question Evaluation of a RAG Pipeline

Abstract. I built a retrieval-augmented generation service with hybrid retrieval (dense + BM25), a relevance threshold that is supposed to make it refuse unanswerable questions, and an extractive generator that quotes retrieved text with citations. I tested the two failure surfaces the standard eva…

Read the full story at DEV Community — Machine Learning ↗

Timeline · 1 report

  1. 2026-11-03 04:37 · DEV Community — Machine Learning
    Retrieval Is Not a Safety Boundary: Poisoned-Corpus and Unsupported-Question Evaluation of a RAG Pipeline

More stories

  1. Anthropic reports $17B quarterly revenue, $900B spending commitments ahead of IPO — Reuters AI
  2. Anthropic Reveals Plan to Spend More Than $895 Billion Ahead of Public Debut — Bloomberg AI
  3. OpenAI appoints former US Army Secretary Dan Driscoll as Senior Counselor for National Security Modernization — OpenAI News
  4. Google expands Dreambeans AI app to India in first international launch — TechCrunch AI
  5. Marvell to acquire Lumentum in $115 billion all-stock AI data-centre deal — DigiTimes Asia
  6. OpenAI cuts funding target to about $20 billion as tech stocks sell off — Economic Times Tech
  7. Anthropic SEC filing warns of imminent threshold for AI-driven research acceleration — TechCrunch AI
  8. OpenAI to merge ChatGPT's Chat and Work tabs into a single desktop experience — TechCrunch AI

Get the daily brief of stories like this at 6:30 every morning →