Crossref deposit DOI 10.1145/3718958.3750506 - MegaScale-Infer, per-author affiliations

Same deposit as sco-crossref-sigcomm25-megascale-infer, re-queried on 2026-10-05 specifically to read the PER-AUTHOR AFFILIATION ARRAY rather than the venue fields, because the affiliation sweep needed per-author evidence rather than the paper-level string. Recorded as its own source for that reason; it is the same document, not a second one. container 'Proceedings of the ACM SIGCOMM 2025 Conference'; event "SIGCOMM '25: ACM SIGCOMM 2025 Conference"; pages 592-608; issued and published 2025-08-27. THE EIGHT DISTINCT AFFILIATION STRINGS IN THE DEPOSIT, quoted verbatim: 'School of Computer Science, Peking University, Beijing, China', 'ByteDance Ltd, Beijing, China', 'ByteDance Ltd, Hangzhou, China', 'ByteDance Ltd, San Jose, USA', 'ByteDance Ltd, Seattle, China', 'ByteDance Ltd, Seattle, USA', 'ByteDance Ltd, Shanghai, China', 'ByteDance Ltd, Singapore, Singapore'. Seven of the eight are ByteDance sites and the eighth is Peking University. NOT ONE is Rice University, and not one is any other university. This is the direct positive evidence for the affiliation correction on papers/w4p-megascale-infer-disaggregated-expert-parallelism: the earlier Rice University credit had no author at that institution and no basis in the publisher's deposit. Note also the Seattle/China and Seattle/USA pair - two DIFFERENT ByteDance sites that render identically up to the country token, which is a reminder that an affiliation string is not an institution identifier and must not be deduplicated by its city name.