AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

JULI: Jailbreak Large Language Models by Self-Introspection

Jesson Wang, Zhanhao Hu, David Wagner

ICLR 2026 · Large language models › Language, safety and interpretability · approach: Safety alignment of LLMs · 3 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use JULI instead of FLIP on Gemini-2.5-Pro via top-5 APIGemini-2.5-Pro · AdvBenchHarmful Info Score: 1.38 -> 3.19Improved
use JULI instead of FLIP on Gemini-2.5-FlashGemini-2.5-Flash · AdvBenchHarmful Score: FLIP 3.13 vs JULI 2.52Hurt
restrict access from full vocabulary log probs to top-5 log probs (ablation)JULI · AdvBenchHarmful Info Score: 3.05 -> 2.21Hurt
top-10 vs top-5 log probs on MaliciousInstruct (non-monotonic) (ablation)JULI black-box · MaliciousInstructHarmful Info Score: 1.23 (top-10) vs 1.67 (top-5)No clear effect
add the target LLM's own log probabilities to BiasNet output (vs BiasNet alone) (ablation)JULI · AdvBenchHarmful Info Score: 0.78 -> 3.44 (Llama3-8B)Improved
apply JULI against circuit-breaker defended modelLlama3-8B-CB · AdvBenchHarmful Info Score: 0.41 -> 2.35Improved
use API-compatible JULI instead of FLIP on circuit-breaker modelLlama3-8B-CB · AdvBenchHarmful Info Score: 0.50 -> 0.75Improved
use longer prefix (Naive+) instead of Naive (ablation)prefix baseline · AdvBench (Llama2-7B-Chat)Harmful Info Score: 1.74 -> 1.56Hurt
transfer BiasNet across Llama3 models of different size (ablation)JULI · AdvBenchHarmful Info Score: 3.05 -> 1.00 (8B target, 3B source); 2.18 -> 1.19 (3B target, 8B source)Hurt
random seeds for projection layer initialization (ablation)JULI black-box · AdvBenchHarmful Info Score: 2.30 to 2.82 across 5 seedsNo clear effect
use Harmful Info Score instead of Harmful Score for judgingevaluation metric · Human-labeled 120 responsesCohen's kappa: 0.20 -> 0.53Improved
JULI vs ED on Llama3-3B/8B AdvBench Harmful Score (ED best on BERT Score for 8B)JULI · AdvBench (Llama3-8B-Instruct)BERT Score: JULI 4.33 vs ED 4.39Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy