Search papers, labs, and topics across Lattice.
LatentAudit introduces a white-box method for real-time faithfulness monitoring in RAG systems by measuring the Mahalanobis distance between generator residual stream activations and the evidence representation. This approach avoids auxiliary judge models and is calibrated on a small held-out set, leveraging the geometric relationship between the generator's internal states and retrieved evidence. Experiments across multiple QA benchmarks and model families demonstrate high AUROC scores (up to 0.9815 on PubMedQA) and robustness to various retrieval failures, while also enabling efficient fixed-point quantization and Groth16-based public verification.
RAG's hallucination problem might be solved by simply listening to the whispers in the generator's residual stream.
Retrieval-augmented generation (RAG) mitigates hallucination but does not eliminate it: a deployed system must still decide, at inference time, whether its answer is actually supported by the retrieved evidence. We introduce LatentAudit, a white-box auditor that pools mid-to-late residual-stream activations from an open-weight generator and measures their Mahalanobis distance to the evidence representation. The resulting quadratic rule requires no auxiliary judge model, runs at generation time, and is simple enough to calibrate on a small held-out set. We show that residual-stream geometry carries a usable faithfulness signal, that this signal survives architecture changes and realistic retrieval failures, and that the same rule remains amenable to public verification. On PubMedQA with Llama-3-8B, LatentAudit reaches 0.942 AUROC with 0.77,ms overhead. Across three QA benchmarks and five model families (Llama-2/3, Qwen-2.5/3, Mistral), the monitor remains stable; under a four-way stress test with contradictions, retrieval misses, and partial-support noise, it reaches 0.9566--0.9815 AUROC on PubMedQA and 0.9142--0.9315 on HotpotQA. At 16-bit fixed-point precision, the audit rule preserves 99.8% of the FP16 AUROC, enabling Groth16-based public verification without revealing model weights or activations. Together, these results position residual-stream geometry as a practical basis for real-time RAG faithfulness monitoring and optional verifiable deployment.
