Nicholas Carlini

Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces

The Attacker Moves Second: Stronger Adaptive Attacks Bypass Defenses Against LLM Jailbreaks and Prompt Injections

ImpossibleBench: Measuring LLMs' Propensity of Exploiting Test Cases

Auditing Agents for Adversarial Fine-tuning Detection

Poisoning Attacks on LLMs Require a Near-constant Number of Poison Samples

AutoAdvExBench: Benchmarking Autonomous Exploitation of Adversarial Example Defenses

Adversarial Perturbations Cannot Reliably Protect Artists From Generative AI

Measuring Non-Adversarial Reproduction of Training Data in Large Language Models

Scalable Extraction of Training Data from Aligned, Production Language Models

On Evaluating the Durability of Safeguards for Open-Weight LLMs

Exploring and Mitigating Adversarial Manipulation of Voting-Based Leaderboards

AutoAdvExBench: Benchmarking Autonomous Exploitation of Adversarial Example Defenses

IF-Guide: Influence Function-Guided Detoxification of LLMs

Evaluating Privacy Risks of Parameter-Efficient Fine-Tuning

Persistent Pre-training Poisoning of LLMs

Certified Robustness to Clean-label Poisoning Using Diffusion Denoising

Stealing User Prompts from Mixture-of-Experts Models