Source · Tier B · PreprintSource · Verifying LLM Inference to Detect Model Weight Exfiltration
Verifying LLM Inference to Detect Model Weight Exfiltration
R. Rinberg, A. Karvonen, A. Hoover, D. Reuter, K. Warr. 2025. arXiv.
| Original | https://arxiv.org/abs/2511.02620 |
|---|---|
| DOI | 10.48550/arXiv.2511.02620 |
| arXiv | 2511.02620 |
| Version | arXiv v1 4 Nov 2025, v2 10 Dec 2025, v3 12 Mar 2026; v3 read. No journal or conference version found as of 2026-09-25. |
| Accessed | 2026-10-20 |
| Imported from | hodgkins-ai-verification-papers@c71e59ff0e8e |
| Note | Listed under "Inference verification" in the Hodgkins bibliography (CC BY 4.0). |
Cited by
- Bounding unexplained information in outputsMechanism
- Deterministic and bit-exact inferenceMechanism
- Hardware-attested weight bindingMechanism
- Network taps and certifiersMechanism
- Sampled inference recomputationMechanism
- AI 2040 inference-only verification stackImplementation
- DiFR (Divergence From Reference)Implementation
- Model weights have not left the facilityClaim
- RecomputationConcept
- Weight exfiltrationConcept