Newest papers

Auditor-Assisted Summary-Channel Verification for Hosted LLM Identity Substitution

Ziyi Zhang, Ziyao Wang, Guoheng Sun, Ang Li, Jian Li · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.25982 · cited by 0 works (OpenAlex) · open-access PDF

Portuguese

Os serviços de LLM hospedados estão cada vez mais expondo a identidade do modelo como uma reivindicação de produto, mas usuários externos e auditores muitas vezes não conseguem inspeccionar qual configuração do modelo serviu a uma solicitação. Estudamos um regime assistido por auditores em que um auditor confiável registra identidades reivindicadas e prepara material de verificação antes da implantação, enquanto a verificação do tempo de auditoria usa apenas a identidade reivindicada e a saída liberada, sem acesso em tempo real a internos de modelos, logs de inferência, decisões de roteamento ou metadados do provedor. Proponemos SUMMARK, um quadro de watermarking de canal de resumo que usa resumos de raciocínio visíveis ao usuário como uma superfície de auditoria chave. Durante a configuração, a SUMMARK

Machine translation (Meta M2M100); the English original follows.

Abstract

Hosted LLM services increasingly expose model identity as a product claim, but external users and auditors often cannot inspect which model configuration served a request. We study an auditor-assisted regime in which a trusted auditor enrolls claimed identities and prepares verification material before deployment, while audit-time verification uses only the claimed identity and released output, without runtime access to model internals, inference logs, routing decisions, or provider metadata. We propose SUMMARK, a summary-channel watermarking framework that uses user-visible reasoning summaries as a keyed audit surface. During setup, SUMMARK constructs identity-specific vocabulary carriers, trains lightweight summary adapters, and calibrates per-identity thresholds. At deployment time, a keyed statistical test assesses whether the released summary is consistent with the claimed identity and, when needed, attributes the most likely enrolled identity. Across model families, scales, and task domains, SUMMARK reliably detects cross-family and cross-scale substitutions at low false-positive rates while preserving task-facing answer quality. We further study FP16->INT8/INT4 precision-consistency diagnostics and robustness under bounded post-processing, rewriting stress tests, and text-only spoofing, revealing both strong performance and clear failure boundaries. Our code is publicly available at https://github.com/trinitr0toluene/Summark.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Computer science, Identity (music), Rewriting, Computer security, Robustness (evolution), Audit

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange