mlsanityData to 14 Oct 2026

AttnDiff: Attention-based Differential Fingerprinting for Large Language Models

arXiv 2604.05502 ↗First posted 7 Apr 2026cs.CR cs.LGcited by 4

On the map: LLM alignment / RLHF / preference optimisation → Safety alignment and jailbreak attacks

Abstract (arXiv)

Protecting the intellectual property of open-weight large language models (LLMs) requires verifying whether a suspect model is derived from a victim model despite common laundering operations such as fine-tuning (including PPO/DPO), pruning/compression, and model merging. We propose textscAttnDiff, a data-efficient white-box framework that extracts fingerprints from models via intrinsic information-routing behavior. textscAttnDiff probes minimally edited prompt pairs that induce controlled semantic conflicts, captures differential attention patterns, summarizes them with compact spectral descriptors, and compares models using CKA. Across Llama-2/3 and Qwen2.5 (3B--14B) and additional open-source families, it yields high similarity for related derivatives while separating unrelated model families (e.g., >0.98 vs. <0.22 with M=60 probes). With 5--60 multi-domain probes, it supports practical provenance verification and accountability.

Key facts. Cited by 4 papers in the map (data through 14 Oct 2026). Placed in Safety alignment and jailbreak attacks, one of 10 directions in LLM alignment / RLHF / preference optimisation. Builds on 29 of its 49 references that are in the map. First posted on arXiv 7 Apr 2026.

Builds on

29 of 49 references.

References · 29

  1. 2022Training language models to follow instructions with human feedbackcited by 15,971
  2. 2024The Llama 3 Herd of Modelscited by 14,715
  3. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
  4. 2022Constitutional AI: Harmlessness from AI Feedbackcited by 3,050
  5. 2021TruthfulQA: Measuring How Models Mimic Human Falsehoodscited by 2,975
  6. 2022GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformerscited by 1,919
  7. 2019Similarity of Neural Network Representations Revisitedcited by 1,755
  8. 2023Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunchcited by 694
  9. 2023WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instructcited by 625
  10. 2021Datasets: A Community Library for Natural Language Processingcited by 530
  11. 2023Llemma: An Open Language Model For Mathematicscited by 388
  12. 2023Efficient and Effective Text Encoding for Chinese LLaMA and Alpacacited by 305
17 more
  1. 2024Model Merging in LLMs, MLLMs, and Beyond: Methods, Theories, Applications and Opportunitiescited by 254
  2. 2024Evolutionary Optimization of Model Merging Recipescited by 208
  3. 2023Model Breadcrumbs: Scaling Multi-Task Model Merging with Sparse Maskscited by 133
  4. 2024DELLA-Merging: Reducing Interference in Model Merging through Magnitude-Based Samplingcited by 81
  5. 2024FuseChat: Knowledge Fusion of Chat Modelscited by 55
  6. 2024Hey, That's My Model! Introducing Chain & Hash, An LLM Fingerprinting Techniquecited by 37
  7. 2024TRAP: Targeted Random Adversarial Prompt Honeypot for Black-Box Identificationcited by 36
  8. 2024Double-I Watermark: Protecting Model Copyright for LLM Fine-tuningcited by 21
  9. 2024UTF:Undertrained Tokens as Fingerprints A Novel Approach to LLM Identificationcited by 18
  10. 2025Intrinsic Fingerprint of LLMs: Continue Training is NOT All You Need to Steal A Model!cited by 18
  11. 2023Functional Invariants to Watermark Large Transformerscited by 15
  12. 2025LLM DNA: Tracing Model Evolution via Functional Representationscited by 15
  13. 2025EverTracer: Hunting Stolen Large Language Models via Stealthy and Robust Probabilistic Fingerprintcited by 12
  14. 2025CTCC: A Robust and Stealthy Fingerprinting Framework for Large Language Models via Cross-Turn Contextual Correlation Backdoorcited by 12
  15. 2025SeedPrints: Fingerprints Can Even Tell Which Seed Your Large Language Model Was Trained Fromcited by 10
  16. 2025Are Robust LLM Fingerprints Adversarially Robust?cited by 8
  17. 2025Towards the Resistance of Neural Network Watermarking to Fine-tuningcited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 4.

4 papers.

  1. 2026MemMark: State-Evolution Attribution Watermarking for Agent Long-Term Memory Systemscited by 16
  2. 2026KBF: Knowledge Boundary as Fingerprint for Language Model and Black-Box API Auditingcited by 3
  3. 2026Witness Overlap: Directional Provenance Inside Open-Weight Model Familiescited by 0
  4. 2026Emergent Structure in the Marginal Attention Space of Language Modelscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Fingerprinting Multimodal Large Language Models
  2. 2026A Behavioral Fingerprint for Large Language Models: Provenance Tracking via Refusal Vectors
  3. 2025RAFP: Identifying LLM Lineages via Rare-Region Fingerprints
  4. 2024LLMmap: Fingerprinting For Large Language Models
  5. 2025AWM: Accurate Weight-Matrix Fingerprint for Large Language Models
  6. 2025Fingerprinting LLMs via Prompt Injection
  7. 2025Gradient-Based Model Fingerprinting for LLM Similarity Detection and Family Classification
  8. 2025CoTSRF: Utilize Chain of Thought as Stealthy and Robust Fingerprint of Large Language Models