LOCA identifies an average of six minimal interpretable changes in intermediate representations that causally induce refusal on otherwise successful jailbreaks for Gemma and Llama models.
Title resolution pending
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
citation-role summary
dataset 1
citation-polarity summary
years
2026 5verdicts
UNVERDICTED 5roles
dataset 1polarities
use dataset 1representative citing papers
Adaptive Instruction Composition uses a neural contextual bandit with RL to adaptively combine crowdsourced texts, generating more effective and diverse LLM jailbreaks than random or prior adaptive methods on Harmbench.
DR-Smoothing introduces a disrupt-then-rectify prompt processing scheme into smoothing defenses, delivering tight theoretical bounds on success probability against both token- and prompt-level jailbreaks.
Suppressing one refusal neuron or amplifying one concept neuron bypasses safety alignment in LLMs from 1.7B to 70B parameters without training or prompt engineering.
Stylistic rewrites of harmful prompts raise attack success rates from 3.84% to 36.8-65% across 31 frontier models, indicating weak generalization in safety refusals.
citing papers explorer
-
Minimal, Local, Causal Explanations for Jailbreak Success in Large Language Models
LOCA identifies an average of six minimal interpretable changes in intermediate representations that causally induce refusal on otherwise successful jailbreaks for Gemma and Llama models.
-
Adaptive Instruction Composition for Automated LLM Red-Teaming
Adaptive Instruction Composition uses a neural contextual bandit with RL to adaptively combine crowdsourced texts, generating more effective and diverse LLM jailbreaks than random or prior adaptive methods on Harmbench.
-
Guaranteed Jailbreaking Defense via Disrupt-and-Rectify Smoothing
DR-Smoothing introduces a disrupt-then-rectify prompt processing scheme into smoothing defenses, delivering tight theoretical bounds on success probability against both token- and prompt-level jailbreaks.
-
A Single Neuron Is Sufficient to Bypass Safety Alignment in Large Language Models
Suppressing one refusal neuron or amplifying one concept neuron bypasses safety alignment in LLMs from 1.7B to 70B parameters without training or prompt engineering.
-
Adversarial Humanities Benchmark: Results on Stylistic Robustness in Frontier Model Safety
Stylistic rewrites of harmful prompts raise attack success rates from 3.84% to 36.8-65% across 31 frontier models, indicating weak generalization in safety refusals.