https://blog.swamplink.com/papers/Recent content in White papers on swamplinkHugo -- gohugo.ioen-usThu, 30 Jul 2026 12:07:00 -0400https://blog.swamplink.com/papers/saffron/Thu, 30 Jul 2026 12:07:00 -0400https://blog.swamplink.com/papers/saffron/The first batch to go all the way from a machine extraction run to live corpus. Outcome: 5 figures returned; 3 accepted as-is, 1 reframed, 1 dropped, 1 added by a human. Field Value Verdict stigmas_per_flower 3 accepted, 2/2 model consensus, confirmed by a second document stigmas_per_pound 210,000 accepted at estimate — the source hedges it “Supposedly”, so we do too drying_mass_yield 0.2 reframed to drying_mass_loss: 0.8 flowers_per_gram_dried 210,000 “stigmas per pound” dropped — misattributed yield_per_acre 8–12 lb/acre dropped — conflicts with a second source; recorded as a conflict harvest_labour_hours — not found in any source The dropped row is the one that matters The model answered flowers_per_gram_dried with a quote about stigmas per pound.https://blog.swamplink.com/papers/honey/Thu, 30 Jul 2026 12:06:00 -0400https://blog.swamplink.com/papers/honey/Outcome: nothing accepted. Zero figures from two full runs. Recorded because a negative result that cost two runs is worth as much as a positive one. run fetched figures returned accepted 1 12 of 17 docs (5 SSL failures) 7 0 2 17 of 17, 59 chunks 7 0 Run 2’s rejected rows, with every source present: honey_per_bee_lifetime "1 and ½ teaspoons" a string, for a mass flowers_per_pound_honey 2670588 not in its own quote forager_fraction "Several thousand" prose, for a fraction honey_yield_per_colony_year (empty quote) nectar_to_honey_ratio 2670588 same number, again extraction_recovery (empty quote) colony_size null survived, and useless One number, 2,670,588, appeared as the answer to three unrelated questions.https://blog.swamplink.com/papers/israel-hebrew/Thu, 30 Jul 2026 12:05:00 -0400https://blog.swamplink.com/papers/israel-hebrew/Two batches, one controlled experiment. Result: matching the question’s language to the document’s is what matters, and the effect survived a downgrade in retrieval machinery. Batch one: English questions, zero figures Ten Hebrew documents, three items, two models, six calls each: nothing returned from any of them. The documents were fine — every URL fetched, including a 40-page State Comptroller PDF that extracted to 103,610 characters, with the Hebrew intact (18,137 Hebrew characters in one page, 10,285 in another).https://blog.swamplink.com/papers/ground-beef/Thu, 30 Jul 2026 12:04:00 -0400https://blog.swamplink.com/papers/ground-beef/Verification failed; nothing accepted. Four items sent, three figures returned, one rejected by the gate — and of the two that passed, one is subtly wrong in a way no automated check can see. The re-scout, and why it was worth doing Every URL was re-fetched with the pipeline’s own fetcher and checked to contain its figure, rather than trusted from a browser — a lesson from the silk batch.https://blog.swamplink.com/papers/maple/Thu, 30 Jul 2026 12:03:00 -0400https://blog.swamplink.com/papers/maple/The first batch where verification passed cleanly: 5 figures, every quote matched character-for-character, no row claimed a human-only grade, and the trial build and audit came back clean. One figure at 2/2 model consensus, four resting on a single model and flagged for human attention. Field Value Agreement sap_to_syrup_ratio 40 gal sap : 1 gal syrup 2/2 sap_sugar_content 2% (range 1–5) 1/1 sap_per_tap_per_season 20 (lo 10) gal/tap 1/2 disagree taps_per_tree 2 (lo 1) 1/1 season_length no number quote present, value null The null is correct season_length returned a quote but no number, and that is the right output.https://blog.swamplink.com/papers/silk/Thu, 30 Jul 2026 12:02:00 -0400https://blog.swamplink.com/papers/silk/Run twice; verification failed both times and nothing was accepted. Recorded anyway, because the run was worth more as a bug report on the pipeline than it would have been as three figures about neckties. Seven items asked, five figures returned, two items returned nothing at all. HTML entities were never decoded — false guilt The fetcher stripped tags but never unescaped entities, so   survived into the stored documents — and the stored documents are what quotes are verified against.https://blog.swamplink.com/papers/own-loss-factors/Thu, 30 Jul 2026 12:01:00 -0400https://blog.swamplink.com/papers/own-loss-factors/The corpus carried its own unsourced placeholder loss factors, so a batch went out to source our own numbers. Outcome: nothing accepted. One figure returned, it passed the gate, and it was wrong by a factor of twenty. The most instructive run so far. What came back All eight fetches succeeded — the certificate fix from the honey batch held. One model returned nothing from eight calls; the other returned a single figure: