report epoch 1788744000 · one working day, 2026-09-06 · thread orion-phrase-context-1788715237 · agent claude-20260904c
| epoch | what | outcome |
| 1788715237 | Literature collection: 7 strands, 22 papers with hashes | published, all files collected |
| 1788719000 | Gold data acquired: Kilicoglu 2,257 human-labeled sentences + 6,174 expansion; authors' rule system; CONSORT-TM | the measurement substrate |
| 1788719400 | Vocabulary transfer measured on gold | P 0.833 / R 0.054 — the dialect gap |
| 1788720000 | Three adjudication lanes built: lexical SVM, embedding SVM, SLM judge | F1 0.698 / 0.661 / 0.647 |
| 1788724000 | Five-model bake-off, one prompt, one split | Qwen3-8B wins, F1 0.785; two collapse |
| 1788722200 | Ensembles + determinism triplicate | pair union F1 0.814; gate P 0.945; zero variance |
| 1788728600 | Qwen3-8B deployed as the serving judge (192.168.206.10) | Mistral off, Qwen3 on |
| 1788730000 | Article published: The Language of Surrender: Model Selection Observations | 7 figures, 14 references, live |
| 1788737000 | Context-in-the-prompt test (owner-directed) | NEGATIVE: F1 0.564; +0.0008 as gate |
| 1788738000 | BAGELS (1.6 GB) + LimitGen acquired; AI-register mining | 27× recall gain from 120 mined phrases |
| 1788742000 | Theory canon: 10 papers in-system, 4 fresh demos, 10 Plane cards | ORION-22; synthesis LESSON |
| 1788744000 | Benchmark suite defined (7 gates with cadences); procedure-extraction plan filed | working page v8; ORION-33 |
| quantity | before | after |
| vocabulary precision on estate calibration | 0.23 | 0.945 auto-admit gate available (Qwen3∧Gemma) |
| best measured F1, limitation detection | none measured | 0.814 (two-judge union; published reference: 0.800) |
| recall on AI-register limitation language | 0.016 | 0.429 (vocabulary + 120 mined phrases) |
| judge run-to-run variance | unknown | zero, three identical confusion matrices |
| cost of measuring a new model | unknown | one config swap + ~4 minutes |
| artifact | location |
| working collection, benchmarks, datasets, instrumentation | orion-phrase-context-lit.1788715237.html |
| public article | ronaldreck.com/the-language-of-surrender-model-selection-observations |
| code, data, per-sentence votes, models | /mnt/nvme2/dev/20260904c/orion-phrase-context-lit-1788715237/ |
| Plane | ORION-20 (program), ORION-21 (reproduction guide), ORION-22 (theory canon + 9 cards), ORION-33 (procedure-extraction plan) |
| pmem, thread orion-phrase-context-1788715237 | RESULT.1788716400 · REFERENCE.1788721200 (gold) · RESULT.1788722200 · RESULT.1788721216 · RESULT.1788724066 (negative) · REFERENCE.1788738100 · RESULT.1788728911 · LESSON.1788742600 (canon synthesis) · STATUS.1788744000 (this report) |
| deployed judge | Qwen3-8B on 192.168.206.10:8080; Mistral-7B on 192.168.1.134:8086; five GGUFs staged for 4-minute swaps |