SeedPrints: Fingerprints Can Even Tell Which Seed Your Large Language Model Was Trained From

• URL: http://arxiv.org/abs/2509.26404v1
• Date: Tue, 30 Sep 2025 15:34:08 GMT
• Title: SeedPrints: Fingerprints Can Even Tell Which Seed Your Large Language Model Was Trained From
• Authors: Yao Tong, Haonan Wang, Siquan Li, Kenji Kawaguchi, Tianyang Hu,
• Abstract summary: We propose a stronger and more intrinsic notion of LLM fingerprinting: SeedPrints.<n>We show that untrained models exhibit reproducible token selection biases conditioned solely on their parameters.<n> Experiments on LLaMA-style and Qwen-style models show that SeedPrints achieves seed-level distinguishability and can provide birth-to-lifecycle identity verification akin to a biometric fingerprint.
• Score: 65.75182441010327
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Fingerprinting Large Language Models (LLMs) is essential for provenance verification and model attribution. Existing methods typically extract post-hoc signatures based on training dynamics, data exposure, or hyperparameters -- properties that only emerge after training begins. In contrast, we propose a stronger and more intrinsic notion of LLM fingerprinting: SeedPrints, a method that leverages random initialization biases as persistent, seed-dependent identifiers present even before training. We show that untrained models exhibit reproducible token selection biases conditioned solely on their parameters at initialization. These biases are stable and measurable throughout training, enabling our statistical detection method to recover a model's lineage with high confidence. Unlike prior techniques, unreliable before convergence and vulnerable to distribution shifts, SeedPrints remains effective across all training stages and robust under domain shifts or parameter modifications. Experiments on LLaMA-style and Qwen-style models show that SeedPrints achieves seed-level distinguishability and can provide birth-to-lifecycle identity verification akin to a biometric fingerprint. Evaluations on large-scale pretrained models and fingerprinting benchmarks further confirm its effectiveness under practical deployment scenarios. These results suggest that initialization itself imprints a unique and persistent identity on neural language models, forming a true ''Galtonian'' fingerprint.

Related papers

• A Behavioral Fingerprint for Large Language Models: Provenance Tracking via Refusal Vectors [43.11304710234668]
  We introduce a novel fingerprinting framework that leverages the behavioral patterns induced by safety alignment.<n>In a large-scale identification task across 76 offspring models, our method achieves 100% accuracy in identifying the correct base model family.<n>We propose a theoretical framework to transform this private fingerprint into a publicly verifiable, privacy-preserving artifact.
  arXiv  Detail & Related papers  (2026-02-10T05:57:35Z)
• Natural Fingerprints of Large Language Models [19.87526607747389]
  We show that even when large language models are trained on exactly the same dataset, their outputs remain distinguishable.<n>We refer to these unintended, distinctive characteristics as natural fingerprints.<n>These results suggest that training dynamics can systematically shape model behavior, independent of data or architecture.
  arXiv  Detail & Related papers  (2025-04-21T05:48:52Z)
• UTF:Undertrained Tokens as Fingerprints A Novel Approach to LLM Identification [9.780530666330007]
  Fingerprinting large language models (LLMs) is essential for verifying model ownership, ensuring authenticity, and preventing misuse.<n>In this paper, we introduce a novel and efficient approach to fingerprinting LLMs by leveraging under-trained tokens.<n>Our method has minimal overhead and impact on model's performance, and does not require white-box access to target model's ownership identification.
  arXiv  Detail & Related papers  (2024-10-16T07:36:57Z)
• Adaptive Pre-training Data Detection for Large Language Models via Surprising Tokens [1.2549198550400134]
  Large language models (LLMs) are extensively used, but there are concerns regarding privacy, security, and copyright due to their opaque training data. Current solutions to this problem leverage techniques explored in machine learning privacy such as Membership Inference Attacks (MIAs) We propose an adaptive pre-training data detection method which alleviates this reliance and effectively amplify the identification.
  arXiv  Detail & Related papers  (2024-07-30T23:43:59Z)
• HuRef: HUman-REadable Fingerprint for Large Language Models [44.9820558213721]
  HuRef is a human-readable fingerprint for large language models.<n>It uniquely identifies the base model without interfering with training or exposing model parameters to the public.
  arXiv  Detail & Related papers  (2023-12-08T05:01:47Z)
• Self-Distillation for Further Pre-training of Transformers [83.84227016847096]
  We propose self-distillation as a regularization for a further pre-training stage. We empirically validate the efficacy of self-distillation on a variety of benchmark datasets for image and text classification tasks.
  arXiv  Detail & Related papers  (2022-09-30T02:25:12Z)
• Masked Autoencoders As The Unified Learners For Pre-Trained Sentence Representation [77.47617360812023]
  We extend the recently proposed MAE style pre-training strategy, RetroMAE, to support a wide variety of sentence representation tasks. The first stage performs RetroMAE over generic corpora, like Wikipedia, BookCorpus, etc., from which the base model is learned. The second stage takes place on domain-specific data, e.g., MS MARCO and NLI, where the base model is continuingly trained based on RetroMAE and contrastive learning.
  arXiv  Detail & Related papers  (2022-07-30T14:34:55Z)
• On the Transferability of Pre-trained Language Models: A Study from Artificial Datasets [74.11825654535895]
  Pre-training language models (LMs) on large-scale unlabeled text data makes the model much easier to achieve exceptional downstream performance. We study what specific traits in the pre-training data, other than the semantics, make a pre-trained LM superior to their counterparts trained from scratch on downstream tasks.
  arXiv  Detail & Related papers  (2021-09-08T10:39:57Z)
• Is BERT a Cross-Disciplinary Knowledge Learner? A Surprising Finding of Pre-trained Models' Transferability [74.11825654535895]
  We investigate whether the power of the models pre-trained on text data, such as BERT, can be transferred to general token sequence classification applications. We find that even on non-text data, the models pre-trained on text converge faster than the randomly models.
  arXiv  Detail & Related papers  (2021-03-12T09:19:14Z)
• EaZy Learning: An Adaptive Variant of Ensemble Learning for Fingerprint Liveness Detection [14.99677459192122]
  Fingerprint liveness detection mechanisms perform well under the within-dataset environment but fail miserably under cross-sensor and cross-dataset settings. To enhance the generalization abilities, robustness and the interoperability of the fingerprint spoof detectors, the learning models need to be adaptive towards the data. We propose a generic model, EaZy learning which can be considered as an adaptive midway between eager and lazy learning.
  arXiv  Detail & Related papers  (2021-03-03T06:40:19Z)
• Document Ranking with a Pretrained Sequence-to-Sequence Model [56.44269917346376]
  We show how a sequence-to-sequence model can be trained to generate relevance labels as "target words" Our approach significantly outperforms an encoder-only model in a data-poor regime.
  arXiv  Detail & Related papers  (2020-03-14T22:29:50Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.