# ReplayAudit: Auditing Outsourced LLM Fine-Tuning via Sealed Records and Selective Replay

“ReplayAudit: Auditing Outsourced LLM Fine-Tuning via Sealed Records and Selective Replay” (2026), by Yihan Xia et al. Computer Science research paper. Original abstract below, unchanged. A free PDF is available.

- Authors: Yihan Xia, Wenxin Xu, Taotao Wang
- Published: November 6, 2026
- Field: Computer Science
- DOI: https://doi.org/10.48550/arxiv.2611.09607
- Free PDF: https://arxiv.org/pdf/2611.09607

## 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.

Canonical page: https://nerdi.pages.dev/paper/W7225622086
Read three real papers a day in Nerdi (iPhone): https://apps.apple.com/app/apple-store/id6813826387?pt=128757794&ct=web-paper&mt=8