Pick Your Poison: Learning to Select Poison Sets for Stronger LLM Backdoor Attacks

A new research paper proposes learning to select poison sets rather than sampling them at random, aiming to strengthen backdoor poisoning attacks on large language models. Backdoor poisoning attacks add poisoned examples to otherwise-clean finetuning data, pairing a trigger with a target behavior the model learns to produce when the trigger appears, and existing evaluations typically fix the number of poisoned examples and sample them at random from a candidate pool.