analiza aiqbench
ReplayAudit introduces a telemetry-guided framework for verifying outsourced LLM fine-tuning by sealing training records and selectively replaying computation intervals. By leveraging loss changes and parameter-update norms to stratify sampling, the system significantly reduces audit time while effectively detecting false updates or skipped steps in real LoRA training scenarios. This work addresses critical trust and verification challenges inherent in distributed machine learning workflows where providers may be untrusted.
ReplayAudit: Auditing Outsourced LLM Fine-Tuning via Sealed Records and Selective Replay
De la arXiv
Text integralarXiv:2611.09607v1 Announce Type: cross Abstract: Auditing outsourced LLM fine-tuning requires checking an untrusted provider's training while controlling verification cost. Uniform sampling ignores training-time signals, while checking a suspicious record can require replaying several preceding steps. We present ReplayAudit, a telemetry-guided auditing system that seals training records before challenges and independently replays sampled intervals from sealed checkpoints. Loss changes and parameter-update norms guide stratified random sampling; replay determines whether the selected computation agrees with its records.
We relate sampling probabilities to replay work and evaluate the system on real LoRA training. A 7B BF16 prototype rejects a false update-norm report during a random audit; targeted checks also reject false loss reports and skipped updates. We calibrate sampling budgets offline to at least a 90% probability of directly selecting an injected step. For two learning-rate deviation settings, executing these budgets on the same honest 1.5B trajectory reduces median complete-audit wall time by 49% and 67% relative to uniform sampling.
When skipped updates evade the high-anomaly stratum, the corresponding audit instead costs 38% more. These results show how deviation capture and actual replay work jointly determine the cost of telemetry-guided auditing.
Abstract:Auditing outsourced LLM fine-tuning requires checking an untrusted provider's training while controlling verification cost. Uniform sampling ignores training-time signals, while checking a suspicious record can require replaying several preceding steps. We present ReplayAudit, a telemetry-guided auditing system that seals training records before challenges and independently replays sampled intervals from sealed checkpoints. Loss changes and parameter-update norms guide stratified random sampling; replay determines whether the selected computation agrees with its records.
We relate sampling probabilities to replay work and evaluate the system on real LoRA training. A 7B BF16 prototype rejects a false update-norm report during a random audit; targeted checks also reject false loss reports and skipped updates. We calibrate sampling budgets offline to at least a 90% probability of directly selecting an injected step. For two learning-rate deviation settings, executing these budgets on the same honest 1.5B trajectory reduces median complete-audit wall time by 49% and 67% relative to uniform sampling.
When skipped updates evade the high-anomaly stratum, the corresponding audit instead costs 38% more. These results show how deviation capture and actual replay work jointly determine the cost of telemetry-guided auditing.
Cryptography and Security (cs.CR); Machine Learning (cs.LG)
arXiv:2611.09607 [cs.CR]
(or arXiv:2611.09607v1 [cs.CR] for this version)
https://doi.org/10.48550/arXiv.2611.09607 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 281 cuvinte