An Interpretable N-gram Perplexity Threat Model for Large Language Model Jailbreaks

Valentyn Boreiko, Alexander Panfilov, Václav Voráček, Matthias Hein, Jonas Geiping
2/14/2026

No official code found. Best implementation: valentyn1boreiko/llm-threat-model (12 stars, MIT, last commit 2025-06-11).

Semantic Scholar

Code Implementations(1)

valentyn1boreiko/llm-threat-modelCommunity62%
MIT

An Interpretable N-gram Perplexity Threat Model for Large Language Model Jailbreaks [ICML 2025]

121Shell, PythonOct 14, 20241 years agoMIT

Similar Papers

QuantSR: Accurate Low-bit Quantization for Efficient Image Super-Resolution

70% match1 repo532026

ERVD: An Efficient and Robust ViT-Based Distillation Framework for Remote Sensing Image Retrieval

70% match1 repo22024

Deep Kernel Relative Test for Machine-generated Text Detection

69% match2026

MVDoppler: Unleashing the Power of Multi-View Doppler for MicroMotion-based Gait Classification

67% match2026

GraphCL: Graph-based Clustering for Semi-Supervised Medical Image Segmentation

66% match2026

S3GCL: Spectral, Swift, Spatial Graph Contrastive Learning

64% match1 repo92026

Cite this paper

@article{boreiko2026interpretable,
  title  = {An Interpretable N-gram Perplexity Threat Model for Large Language Model Jailbreaks},
  author = {Valentyn Boreiko and Alexander Panfilov and Václav Voráček and Matthias Hein and Jonas Geiping},
  year   = {2026},
  url    = {https://api.semanticscholar.org/CorpusID:384edfc8f9364abbbf3644c8ac5d1d9ce6e9a938},
  journal = {ICML 2025 2025}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/384edfc8f9364abbbf3644c8ac5d1d9ce6e9a938.svg)](https://codeofpaper.com/paper/384edfc8f9364abbbf3644c8ac5d1d9ce6e9a938)

Discussion