@mailwoman/neural-weights-en-us 4.2.0 → 4.3.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (3) hide show
  1. package/model-card.json +47 -38
  2. package/model.onnx +0 -0
  3. package/package.json +1 -1
package/model-card.json CHANGED
@@ -1,17 +1,17 @@
1
1
  {
2
2
  "name": "neural-weights-en-us",
3
- "version": "4.2.0",
4
- "model_lineage": "v1.0.2-consolidation-runB / step 20000 — consolidation of the parity campaign (unit + affix + country gazetteer-anchor + multi-locale balance) init_from consolidation v1.0.0 step-040000 (fresh optimizer — NOT resume; recorded honestly, see docs/articles/evals/2026-06-10-consolidation-session.md) @ affix 17x on corpus v0.4.12-consolidation — shipped as the unified 4.2.0 release version; tokenizer 0.6.0-a0",
5
- "phase": "Stage 3 — v1.0 consolidation: parity flag-plant (spine + country anchor + affix existence)",
3
+ "version": "4.3.0",
4
+ "model_lineage": "v1.1.0-relabel-consolidation / step 40000 — the v4.2.0 consolidation recipe FROM SCRATCH on a label-consistent mix (#511 affix-split relabel pass: the base corpus's ~467M monolithic street labels rewritten to agree with the affix shard, ending the 1,039:1 contradiction the #492 probe ladder isolated), with both anti-contradiction compensations reverted (synth-affix 17.0→2.0, affix tag weights →1.5). Ships with the locale head exported (locale_logits) driving the address-system conventions mask (#478 slice 1) — declared inference config addressSystemConventions:'auto'. Gate: docs/articles/evals/2026-06-11-v4.3.0-ship-gate.md; tokenizer 0.6.0-a0",
5
+ "phase": "Stage 3 — v1.1 relabel: affix parity closed (data-consistency lever) + conventions layer slice 1",
6
6
  "license": "AGPL-3.0-only",
7
7
  "locale": "en-us",
8
8
  "training": {
9
- "corpus_version": "0.4.12-consolidation",
9
+ "corpus_version": "0.4.12-consolidation (+ loader affix-relabel pass, lexicon affix-relabel-v1)",
10
10
  "tokenizer_version": "0.6.0-a0",
11
- "steps": 20000,
12
- "best_step": 20000,
11
+ "steps": 40000,
12
+ "best_step": 40000,
13
13
  "hardware": "NVIDIA A100-SXM4-40GB (Modal cloud)",
14
- "recipe": "Run B of the consolidation campaign: init_from the clean v1.0.0 consolidation step-040000 (every proven lever: unit shard, affix-ml shard, country balanced shard + gazetteer soft anchor + channel choreography, both-order German), synth-affix 17x, 20k steps, CE-only, lr=1.5e-4, seed 42. Selected over v1.0.0/A/C at the fork: strongest stable variant (US postcode 97.3, country 89.8, FR hn 94.6). STATED RE-BASELINES vs canonical bars: affix 64.9/48.8 (vs 78/67), US street 76.2 (vs 80.4), unit 90.6 (vs 92) — measured 29M stability ceiling, see issue #492. GATE NUMBERS ARE REPAIRS-OFF (#486). Eval procedure REQUIRES --gazetteer-lexicon + --suppress-gaz-near-postcode (zero-filled clues degrade country recall and fake an affix crash)."
14
+ "recipe": "v1.1.0-relabel-consolidation: from-scratch 40k on v0.4.12 with the #511 affix-split relabel pass at load time (trailing USPS suffix→street_suffix, leading directional→street_prefix, exact shard-builder parity, after augmentation), synth-affix 2.0 (natural weight — the contradiction it fought is gone), affix tag class-weights 1.5, gazetteer anchor + train-time choreography, CE-only, lr=1.5e-4, seed 42."
15
15
  },
16
16
  "architecture": {
17
17
  "hidden_size": 384,
@@ -79,14 +79,14 @@
79
79
  "intersection_a",
80
80
  "intersection_b"
81
81
  ],
82
- "notes": "v4.1.0 — secondary-unit coverage. Same Stage-3 33-BIO-label schema as 4.0.0 (no schema change). Adds a format-diverse synth-unit shard (USPS Pub-28 C2 designators: APT/STE/FL/… across unit-after, unit-first, bare, and venue-prefixed layouts) on top of the v0.9.3 multi-locale base. `unit` recognition 0%→92.3% on a held-out real-designator eval; by 'negative space' it also raised US `street` +3.3pp and lifted `country` (US +6pp, FR +15pp) — covering the missing tag sharpened its neighbors. No regression vs 4.0.0 on any US/FR golden tag; DE native-order locality held (90.6%).",
82
+ "notes": "v4.3.0 — the affix lever closed by data consistency, not weights. street_prefix 64.9→93.6, street_suffix 48.8→96.6 (int8, real-affix eval); on the new 193-row NAD-native eval the shipped v4.2.0 scored 18.2/8.9 F1 where this model scores 92.2/90.3 at P=100. First shipped slice of the conventions layer: the locale head (now exported as locale_logits) detects the address system and the decoder OBEYS it — fr forbids the USPS affix decomposition and pins the 5-digit postcode shape (fixes the FR leading-postcode digit-split this run's first gate caught). Consumers: pass addressSystemConventions:'auto' (the declared ship config) or omit for byte-stable pre-#511 behavior. Known follow-ups: po_box (deferred lever, both Montréal gate rows), FR region tail (27.6→16.2, unfloored), de/gb conventions rows as evidence arrives.",
83
83
  "format": {
84
84
  "model": "ONNX int8 dynamic (quantized from fp32)",
85
85
  "tokenizer": "SentencePiece unigram, byte_fallback=true, vocab_size=48000",
86
86
  "max_sequence_length": 128,
87
87
  "opset": 17,
88
- "fp32_size_mb": 112.9,
89
- "int8_size_mb": 28.6
88
+ "fp32_size_mb": 118.4,
89
+ "int8_size_mb": 29.8
90
90
  },
91
91
  "files": {
92
92
  "model": "model.onnx",
@@ -103,51 +103,60 @@
103
103
  },
104
104
  "base_relpath": "/data/output-v097-unit-v3-s42/checkpoints/step-020000",
105
105
  "eval": {
106
- "ship_gate_2026_06_10": {
106
+ "ship_gate_2026_06_11": {
107
+ "promotion_gate": "PASS 12/12 (gates/v4.3.0-relabel.json, int8-graded, max fp32-int8 delta 0.2pp)",
107
108
  "honest_eval_vt": {
108
109
  "n": 1428,
109
- "region_match_pct": 99.9,
110
+ "region_match_pct": 99.6,
110
111
  "coord_p50_km": 3.4,
111
112
  "coord_p90_km": 7.4,
112
- "pip_coverage_adj_pct": 47.1,
113
- "baseline_v410_region_pct": 100.0,
113
+ "pip_coverage_adj_pct": 46.9,
114
+ "baseline_v420_region_pct": 99.9,
114
115
  "verdict": "PASS"
115
116
  },
116
- "demo_presets": "PASS — 5/6 identical to v4.1.0; 6th is the intended affix split",
117
- "int8_vs_fp32": "PASS — all gate tags within 0.1pp; quant deterministic",
118
- "de_native_order_int8_pct": 90.9
117
+ "demo_presets": "PASS — affix split fires live (1060 W Addison St → W + Addison + St)",
118
+ "int8_vs_fp32": "PASS — all gate tags within 0.2pp",
119
+ "affix_stability_20k_to_40k": "v2 prefix 88.3→91.6, suffix 89.4→90.3 (rose, no decay)"
119
120
  },
120
- "per_component_int8_gazfed": {
121
+ "per_component_int8_gazfed_conventions": {
121
122
  "us": {
122
- "postcode": 97.3,
123
- "country_homograph": 89.8,
124
- "micro": 84.8,
125
- "locality": 72.9,
123
+ "postcode": 97.8,
124
+ "country_homograph": 85.1,
125
+ "micro": 85.1,
126
+ "locality": 74.4,
126
127
  "region": 89.1,
127
- "street": 76.2,
128
- "street_prefix": 64.9,
129
- "street_suffix": 48.8,
130
- "unit": 90.6,
131
- "house_number": 96.9
128
+ "street": 75.5,
129
+ "street_prefix": 93.6,
130
+ "street_suffix": 96.6,
131
+ "unit": 92.1
132
132
  },
133
133
  "fr": {
134
- "postcode": 99.6,
135
- "house_number": 94.6,
136
- "region": 27.6
134
+ "postcode": 99.7,
135
+ "house_number": 97.7,
136
+ "region": 16.2
137
137
  },
138
138
  "de": {
139
- "native_locality_anchor_on": 90.9
139
+ "native_locality_anchor_on": 90.1
140
+ },
141
+ "affix_nad_native_v2": {
142
+ "street_prefix_f1": 92.2,
143
+ "street_suffix_f1": 90.3,
144
+ "precision": 100.0,
145
+ "baseline_v420": {
146
+ "street_prefix_f1": 18.2,
147
+ "street_suffix_f1": 8.9
148
+ }
140
149
  }
141
150
  },
142
- "known_regressions_vs_4_1_0": {
143
- "us_street": -2.3,
144
- "unit": -1.7,
145
- "us_postcode": -1.0,
146
- "mitigations": "arbitration layer #478; architecture escalation #492"
151
+ "known_regressions_vs_4_2_0": {
152
+ "fr_region": -11.4,
153
+ "us_street_exact": -0.7,
154
+ "country_homograph": "-4.7 (NOT comparable: the v4.2.0 figure was measured under the gaz-starved country leg fixed this gate; both clear the 83.3 floor)",
155
+ "mitigations": "FR region tail rides the next corpus pass; arbitration layer #478 continues"
147
156
  }
148
157
  },
149
158
  "files_md5": {
150
- "model.onnx": "9eb4a99f6db06cccff57939f657c09f9",
159
+ "model.onnx": "9ab47793a4a454c8432c5de05567ad0f",
151
160
  "tokenizer.model": "b6137e8c52914c9715374268ecaa4bc6"
152
161
  }
153
- }
162
+ }
package/model.onnx CHANGED
Binary file
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@mailwoman/neural-weights-en-us",
3
- "version": "4.2.0",
3
+ "version": "4.3.0",
4
4
  "description": "Mailwoman neural-classifier weights for locale 'en-us'. Data-only package — loaded by @mailwoman/neural at runtime.",
5
5
  "license": "AGPL-3.0-only",
6
6
  "repository": {