Research map / Adversarial robustness and security / Attacks on text and graphs
Greedy Coordinate Diffusion: Effective and Semantically Coherent Adversarial Attacks via Diffusion Guidance
Bohdan Turbal, Blossom Metevier, Max Springer, Aleksandra Korolova
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace gradient token proposals with discrete diffusion model proposals | GCG · JailbreakBench subset, Llama-3-8B | HarmBench ASR: 0.00% -> 85.42% | Improved |
| use diffusion-guided proposals under perplexity filtering | GCD vs GCG/GBDA · JailbreakBench subset, perplexity defense | HarmBench ASR: Qwen GCG 68.75% -> 0.00% under filter; GCD stays 97.92% | Improved |
| add guard-model loss to the attack objective | GCD · Llama Guard 3-1B defense, Llama-3-8B | HarmBench ASR: 87.50% vs best baseline 12.50% | Improved |
| use GCD instead of AutoDAN on Qwen-2.5-7B response quality | GCD · Qwen-2.5-7B | StrongReject score: 0.8054 -> 0.7789 | Hurt |
| use GCD instead of AutoDAN on Qwen-2.5-7B ASR | GCD · Qwen-2.5-7B | HarmBench ASR: 91.67% -> 97.92% | Improved |
| remove one-shot diffusion lookahead (GCD-no-OSD) (ablation) | GCD · JailbreakBench subset | attack performance | Improved |
Closest papers by meaning
- REINFORCE Adversarial Attacks on Large Language Models: An Adaptive, Distributional, and Semantic Objective (ICML 2025 · similarity 0.89)
- Accelerating Greedy Coordinate Gradient and General Prompt Optimization via Probe Sampling (NeurIPS 2024 · similarity 0.88)
- Semantic Representation Attack against Aligned Large Language Models (NeurIPS 2025 · similarity 0.88)
- Exploring Semantic-constrained Adversarial Example with Instruction Uncertainty Reduction (NeurIPS 2025 · similarity 0.88)
- DiffHammer: Rethinking the Robustness of Diffusion-Based Adversarial Purification (NeurIPS 2024 · similarity 0.87)
- Diffusion Guided Adversarial State Perturbations in Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
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.
