Poisoning Attacks on LLMs Require a Near-constant Number of Poison Samples
Tests pretraining-data poisoning across model and dataset scales to examine the amount of malicious data needed for a targeted effect.
Verify byline in the original ↗PUBLICATION BYLINE
Indexed publications with this author name. Affiliation and author identity require publication-specific evidence.
Tests pretraining-data poisoning across model and dataset scales to examine the amount of malicious data needed for a targeted effect.
Verify byline in the original ↗