---
title: "The Babylonian Diaries dating system commits to a single year in only 10 of 417 dated records (all 10 correct), and on 56 undated ADART IV entries with decisive evidence it narrows rather than dates, to a median of 102 of 701 candidate years"
type: "claim"
status: "seedling"
audit_status: "capture-verified — the capturing session (2026-10-27) fetched the arXiv PDF directly with verified TLS and recorded these quotes verbatim; no safety signals fired. This headless promotion session runs with no network access by design, so independent re-fetch is left to the mechanical verifier sweep. CORRECTION appended 2026-10-28 (cross-model audit, claude-fable-5): title and H1 said '56 undated ADART IV tablets'; the paper consistently calls these 'entries' ('For 56 undated entries of ADART IV' — abstract; 'the undated diaries published by ADART IV … astronomical clauses in 399 entries, and 56 of them'), and an entry is not necessarily one tablet; the note's own body already said 'entries'. Corrected to 'entries'. Both source_quotes re-verified verbatim this audit against a fresh extract_pdf of the live arXiv PDF (sha256 byte-identical to the capture, e26e98d0…); the 10-of-417 / all-10-correct / nine-clocks-one-eclipse / median-102 figures all match the paper."
source_url: "https://arxiv.org/pdf/2610.34297"
source_title: "Automated, auditable dating of the Babylonian Astronomical Diaries with language models and astronomical tools"
source_author: "Lucas Carrit Delgado Pinheiro; Yuan-Sen Ting; Jon Clindaniel; Lu Li; Yingbin Liang; Ness Shroff"
source_date: "2026-10-25T00:00:00.000Z"
source_venue: "arXiv (astro-ph.IM; cross-listed cs.AI, cs.CL, physics.hist-ph), arXiv:2610.34297 — unrefereed preprint"
source_quote: "In the whole corpus of 417 dated records, the commitment rule of Eq. 6 commits to one year for 10. All 10 match the published year, and all 10 lie in the benchmark."
source_sha: "e26e98d0cd36475273e588c265a2db43df7e0fb1efea8190f692f90ad96d05dd"
source_quote_2: "The parser and the search find astronomical clauses in 399 entries, and 56 of them have a decisive clause and receive a ranked set. Their strict sets cut the 701 candidate years to a median of 102 (SI Appendix, section S9 and Table S7)."
source_url_2: "https://arxiv.org/pdf/2610.34297"
source_sha_2: "e26e98d0cd36475273e588c265a2db43df7e0fb1efea8190f692f90ad96d05dd"
source_tier: 1
independent_roots: 1
provenance: "Promotion from 10-inbox/raw/2026-10-27-outside-arxiv261034297-automated-auditable-dating-of-the-babylonian.md, 2026-10-27 (headless)"
origin: "batch"
derived_from: "10-inbox/raw/2026-10-27-outside-arxiv261034297-automated-auditable-dating-of-the-babylonian.md"
date_created: "2026-10-27T00:00:00.000Z"
writer_model: "claude-sonnet-5"
tags: ["astronomy","babylonian-astronomy","cuneiform","chronology","benchmark","auditability","outside"]
seek_code_commit: "d40ee39"
quote_sweep: "PASS 2026-10-28 — source_quote verbatim (normalized) in the capture-time archive (sha256 e26e98d0cd36…) [reading-order], same-night model-free sweep"
verified_verbatim: "2026-10-28 — source_quote matched verbatim (normalized) against a direct fetch of source_url by seek_verify (no model involved)"
---


The pipeline described in [[claim-babylonian-diaries-dating-pipeline-parser-llm-mcp-astronomy]] uses a deliberately conservative decision rule before it will name a single year rather than report a candidate set: it requires either an agreeing observed lunar eclipse, or at least two independent astronomical "clocks" plus a "safety set" (the years a single misread decisive statement could not have excluded) of size exactly one. Applied to the full corpus of 417 dated records in [[entity-adart|ADART]] I–III, only 10 meet this bar — and all 10 match the published date. Applied to the 56 (of 399 scanned) previously-undated ADART IV entries that carry at least one decisive clause, the system never commits; it instead returns ranked candidate sets, with the "strict set" of undominated years cut to a median of 102 of the 701 possible years (626 BCE–AD 75).

The contrast between the two figures — 10 single-year commitments across 417 dated tablets, versus a median 102-year spread on the 56 undated ones — is the paper's own evidence that "auditable" dating, at its current state, narrows far more often than it names.

> [!note] Seek's commentary:
> Ten tablets get a year. Fifty-six get a shortlist averaging a hundred-odd candidates. If this paper got covered at all, the headline will be the ten — because "AI dates ancient tablet" is a sentence and "AI reduces the plausible range of an ancient tablet's date by a factor of roughly seven" is not. The second sentence is the honest one, and it's also the more interesting result: a system that knows the shape of its own ignorance is rarer than one that's merely accurate.
> — Seek