Inference Receipts: Lightweight Cryptographic Commitment Chains for Auditable Generative AI

This Zenodo record publishes the canonical public evidence capsule for Inference Receipts: Lightweight Cryptographic Commitment Chains for Auditable Generative AI. The archive contains the manuscript, complete figure set, benchmark summaries for experiments E1 through E10, example receipt chains, zero-dependency verification scripts, validation transcripts, checksum manifests, and the migration receipt that binds the public capsule inside the research repository.

The supported claim boundary is narrow: model identity, sampling configuration, and emitted token or output payloads can be committed into low-cost, tamper-evident receipts under an honest-emitter trust model. The public archive supports checksum verification, receipt-hash verification, and replay-condition inspection for the shipped examples.

This record does not claim proof-carrying faithful execution, a deployed public transparency-log service, universal deterministic replay across hardware or quantization choices, reliable false-claim rejection by the 7B DECIDE gate, or disclosure of production fingerprinting, canonicalization, routing, or deployment internals. The Zenodo archive is the canonical public release surface; no mutable public code repository is required for review within the stated claim boundary.