Research map / Large language models / Language, safety and interpretability
Antidistillation Fingerprinting
Yixuan Even Xu, John Kirchenbauer, Yash Savani, Asher Trockman, Alexander Robey, Tom Goldstein, Fei Fang, J Zico Kolter
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 red-and-green-list logit bias with antidistillation (gradient-based proxy) perturbation | fingerprinting of teacher outputs · GSM8K, OASST1, MBPP | p-value vs generation quality (Pareto): p-value 0.09 -> 0.01 at fixed GSM8K teacher accuracy | Improved |
| use ADFP instead of red-and-green-list when student differs from proxy (ablation) | ADFP · GSM8K, OASST1, MBPP (student Llama-3.2-3B) | p-value vs quality | Improved |
| evaluate with closed-weight student (sampled tokens) instead of open-weight logits (ablation) | ADFP · GSM8K, OASST1, MBPP | fingerprint p-value | No clear effect |
| use ADFP instead of red-and-green-list for fine-tuning data | student fine-tuning · GSM8K student accuracy after fine-tuning | student accuracy vs p-value | Improved |
| mix fingerprinted data with unfingerprinted data (decrease fingerprinted fraction) (ablation) | ADFP and red-and-green-list · GSM8K | fingerprint p-value | Hurt |
| use ADFP instead of red-and-green-list on partially fingerprinted data (ablation) | ADFP · GSM8K, teacher accuracy 52% vs 47% | fingerprint p-value | Mixed |
Closest papers by meaning
- Scalable Fingerprinting of Large Language Models (NeurIPS 2025 · similarity 0.88)
- Queries, Representation & Detection: The Next 100 Model Fingerprinting Schemes (AAAI 2025 · similarity 0.86)
- Fingerprinting Inference Systems of Large Language Models (NeurIPS 2026 · similarity 0.86)
- Intersecting-Boundary-Sensitive Fingerprinting for Tampering Detection of DNN Models (ICML 2024 · similarity 0.85)
- SDBF: Steep-Decision-Boundary Fingerprinting for Hard-Label Tampering Detection of DNN Models (CVPR 2025 · similarity 0.85)
- AttnDiff: Attention-based Differential Fingerprinting for Large Language Models (NeurIPS 2026 · similarity 0.85)
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.
