AI

SetPrint: Black-Box LLM Fingerprinting via Setwise Correspondence of Response Changes

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 11/11/2026, 05:03:24 UTC.

Briefings

SetPrint is a training-free black-box LLM fingerprinting method that detects model origins by comparing response changes from two models under identical perturbations. It achieves superior performance on LeaFBench compared to existing training-free baselines, driven by its ability to capture both internal response change patterns and cross-model correspondence.

Key points

• SetPrint uses a training-free approach to detect whether a target LLM was derived from a source model by comparing response changes under identical perturbations.
• The method consistently outperforms existing training-free black-box fingerprinting methods on the LeaFBench benchmark.

Context

Black-box LLM fingerprinting aims to determine if a target model was derived from a source model using only output responses to input queries. This research provides a new method to reveal this relationship through Setwise Correspondence of Response Changes.

Limitations

The study is based on a single preprint (arXiv:2611.12950v1) from November 2026, with no additional experimental data, validation results, or source-supported background provided.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 11/11/2026, 07:10:50 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeic5xzd3chu55jgglfyh26mxiqxeazfuhgslwk7ah2isifrewxnori