LivePI benchmark reports indirect prompt injection success rates of 10.7-29.6% across five models on seven input surfaces and shows a two-layer defense blocking all malicious completions while preserving utility.
Agentdojo: A dynamic environment to evaluate prompt injection attacks and defenses for llm agents.Advances in Neural Information Processing Systems, 37:82895–82920
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CR 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
LivePI: More Realistic Benchmarking of Agents Against Indirect Prompt Injection
LivePI benchmark reports indirect prompt injection success rates of 10.7-29.6% across five models on seven input surfaces and shows a two-layer defense blocking all malicious completions while preserving utility.