AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

Hey, That's My Model! Introducing Chain & Hash, An LLM Fingerprinting Technique

Mark Russinovich, Yanan Cai, Ahmed Salem

ICLR 2026 · Large language models › Instruction and fine-tuning · approach: Safety alignment after fine-tuning · 49 citations (OpenAlex)

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 changedWhereMeasuredEffect
add meta-prompt diversification to fingerprint training (ablation)Chain & Hash (random questions) · Instruct models under adversarial meta-promptsfingerprint strength: drops to 0 without -> >99% withImproved
add meta-prompt diversification to fingerprint training (ablation)Chain & Hash (natural questions) · Instruct models under adversarial meta-promptsfingerprint strength: mean 82.8%Improved
use random questions instead of natural questions (ablation)Chain & Hash · Llama-3-8B-Instruct ChatDoc fine-tuning, no meta-promptrequired trials: 1 -> 25Hurt
use random questions instead of natural questions (ablation)Chain & Hash · persistence under meta-prompts (Phi-3-Mini, Llama-3-8B-Instruct)required trials: natural up to 21 -> random 1Improved
add random padding augmentation to fingerprint samplesChain & Hash · post fine-tuning robustnessfingerprint robustnessImproved
expand fingerprint question set from 10 to 100 (ablation)Chain & Hash · Llama-3-8B-Instruct meta-prompt persistencefingerprint strength / convergenceImproved
reduce number of training meta-prompts to six (ablation)Chain & Hash (random questions) · Llama-3-8B-Instructfingerprint strengthNo clear effect
fingerprint model with Chain & HashLlama-3-8B · IFEvalIFEval change %: +46.7Improved
fingerprint model with Chain & HashLlama-3-8B and Phi-3-Mini-Instruct · GSM8KGSM8K change %: -6.2 (Llama-3-8B natural), -3.21 (Phi-3 natural)Hurt
apply INT8 quantization to fingerprinted modelChain & Hash · fingerprinted modelsfingerprint strength: <0.5% drop in most, <2.5% maxNo clear effect
paraphrase output with GPT-4o before verificationChain & Hash · stronger adversaryfingerprint detection: 99% -> 20% (2/10 fingerprints)Hurt
paraphrase input with GPT-4o before queryingChain & Hash · stronger adversaryfingerprint strength: 99% -> 79%Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy