SMRTR AIOct 9, 2025Hacker News

A small number of samples can poison LLMs of any size

SMRTR summary

Researchers found that attackers can create backdoor vulnerabilities in large language models by injecting just 250 malicious documents into training data, making poisoning attacks surprisingly practical.

SMRTR provides this summary for quick context. The original article belongs to Hacker News.

Read the original article
SMRTR AI

Get the next batch of curated summaries in your inbox.

This archive is built from SMRTR newsletter summaries. Subscribe for hand-picked stories without the extra noise.