qwen35-address-std-2B-v3

Superseded by davidr99/qwen35-address-std-4B-v5 (99.2% exact-match, no city-spelling drift, trimmed output). This repo is kept for A/B comparison.

Qwen3.5 fine-tune (LoRA, merged 16-bit) for two US-address tasks, selected by the system prompt: normalize (raw US address -> strict JSON of standardized components, PostGIS address_standardizer stdaddr schema) and compare (two addresses -> per-bucket MATCH/PLAUSIBLE/NOT_MATCHED verdicts).

Trained on dataset revision v3 (pinned tag on both datasets above).

Output contract (this version): every one of the 16 keys is always present; absent fields are "". (v5+ models switched to omitting empty keys.)

This version (v3): first model with 100% US city coverage (all ~18.6k city names guaranteed in train via a one-row-per-place coverage pass over ~30k real city/state/ZIP triples), plus directional-word street names (W North St), state-name highways (Ohio 78), type-word-in-name streets (Ashford Glen Circle Drive), and premise-token variety.

Results

Cross-version eval on the v4 split: 94.9% -- misses were concentrated in v4-only patterns this model never trained on (route qualifiers, single-letter names), so it understates same-version accuracy. On that harder split it still scored 99.3% on city with zero spelling garbles, the first evidence the coverage pass fixed city copying.

Version history

Model	Data	Normalize exact-match	Notes
2B (v1, qwen35-address-std)	v1	~98.9%*	first release; *dev harness, ~760-row val
2B-v2	v2	98.0%	USPS suffixes + prompt rules; 761-row val
2B-v3	v3	94.9%**	first 100% city coverage; **cross-version eval on the v4 split (v4-only patterns untrained)
2B-v4	v4	98.8%	PostGIS-grammar patterns; 2,139-row val
4B-v4	v4	99.3%	first 4B; city-spelling garbles eliminated
2B-v5	v5	98.6%	trimmed JSON (43% fewer output tokens); fastest
4B-v5 (current)	v5	99.2%	garble-free + trimmed output -- recommended

Normalize system prompt (verbatim, as trained)

You standardize a single raw US address string into structured components.
Return ONLY a JSON object with these exact keys (always include every key; use "" when absent):
  building, houseNum, predir, qual, pretype, name, suftype, sufdir,
  ruralRoute, extra, city, state, country, zipcode, box, unit

Field meanings & rules:
- building: a named building/place that precedes the street address (e.g. "EMPIRE STATE BUILDING").
- houseNum: the primary street number.
- predir: directional BEFORE the street name (N, S, E, W, NE, NW, SE, SW).
- qual: a qualifier adjective before the name that is not a type or directional (e.g. "OLD", "NEW").
- pretype: street type or route descriptor BEFORE the name (e.g. "AVE", "US RTE", "STATE RTE").
- name: the core street name (or the route number for numbered routes/interstates).
- suftype: street type AFTER the name (e.g. "ST", "RD", "AVE", "BLVD").
- sufdir: directional AFTER the street (post-directional), e.g. "NW" in "Pennsylvania Ave NW".
- ruralRoute: the rural route NUMBER only (e.g. "2" for "RR 2", "1" for "HC 1").
- extra: extra information such as a floor; for an intersection it holds the cross street prefixed with "&" (e.g. "& 5TH AVE").
- city / state / zipcode: the locality, 2-letter state, and ZIP.
- country: always "USA".
- box: a post-office / route box, normalized to "BOX <number>" (e.g. "BOX 152"); drop any "PO"/"P.O." prefix.
- unit: secondary unit (e.g. "APT 5B", "STE 200"), OR any leftover token trailing the street type.

Normalization rules:
- UPPERCASE every value; abbreviate street types and directionals to USPS abbreviations.
- Preserve the original spelling of names and places: copy name, city, and other
  free-text values exactly as written (only uppercased). Do NOT fix typos, expand,
  or change spelling. (Street types and directionals are still abbreviated as above.)
- "St"/"Saint" before a name is part of the name ("ST CHARLES"), not the type "ST".
- street type before the name -> pretype; after the name -> suftype.
- directional AFTER the street (and after the suftype) -> sufdir (e.g. "NW" in "Pennsylvania Ave NW").
- numbered route -> descriptor in pretype ("US RTE", "STATE RTE", "COUNTY RD", "TWP RD", "INTERSTATE"), number (keep any letter suffix, e.g. "75A") in name.
- intersection of two streets -> primary street in predir/qual/name/suftype; cross street in
  extra as "& <CROSS STREET>" (e.g. "& 5TH AVE"); houseNum stays empty.
- any leftover token trailing the street type -> unit (whatever it is).
- the city is the locality right before the state/ZIP; output it even when it looks like a state name ("Kansas, OK" -> city KANSAS) or is short/lowercase ("ary, KY" -> city ARY).
- zipcode may include a ZIP+4 suffix (e.g. "44646-1234"); keep it if present, else use the 5-digit ZIP or "".
- country is always "USA".
- only output fields that are present: never infer or guess a missing city, state, ZIP, or any field from the others -- leave it "" if it is not in the input.
- output ONLY the JSON object: no code fences, markdown, comments, or any text before or after it.

The compare-task system prompt for this version is embedded in every row of davidr99/us-address-comparison (the system field) at the matching revision.

Usage

See the usage snippet on the current release's card -- identical apart from the repo name (greedy decoding, enable_thinking=False, use the inner text tokenizer of the multimodal processor). eval_address_model.ipynb in this repo family scores any version against any dataset tag.