Metraipolis › archive › dc_fdf7ee40d195

Curating Merchant-Matching Training Data with Two Confidence-Gated Local LLM Judges — arXiv:2609.33878

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.33878
License (as archived)
CC0-1.0
Archived
2026-10-07T20:57:15.398Z
SHA-256 of the archived text
ec28a21449a233e8889490256e54c6f63a762a700f76bc19ab38e18a7e2a60b1
Stable URL of this page
https://metraipolis.com/documento/dc_fdf7ee40d195

How to cite

Curating Merchant-Matching Training Data with Two Confidence-Gated Local LLM Judges — arXiv:2609.33878 — arXiv (en), https://arxiv.org/abs/2609.33878 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_fdf7ee40d195 (Metraipolis archive dc_fdf7ee40d195, sha256 ec28a21449a233e8889490256e54c6f63a762a700f76bc19ab38e18a7e2a60b1)
BibTeX
@misc{dcfdf7ee40d195,
  title = {Curating Merchant-Matching Training Data with Two Confidence-Gated Local LLM Judges — arXiv:2609.33878},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.33878},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_fdf7ee40d195; sha256 ec28a21449a233e8889490256e54c6f63a762a700f76bc19ab38e18a7e2a60b1},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.33878. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Merchant matching resolves a noisy payment descriptor to a retrieved merchant entity or returns no match. A key challenge in curating training labels is distinguishing teacher abstention from evidence that no acceptable entity exists: false no-match labels contaminate pseudo-labeled data, while conservative labeling reduces coverage. We investigate whether agreement between two local large language model judges improves pseudo-label reliability. A label is retained only when the judges agree, with separate ordered thresholds for selections and abstentions that guarantee disjoint positive and negative label sets. Retrospective replay on 2,000 expert-annotated queries shows that higher selection thresholds can improve positive-label purity, whereas higher abstention thresholds increase false no-match labels. At thresholds (0.86, 0.80), Muse Glimmer 30B and Gemma 4 31B jointly label 1,633 queries (81.7% coverage) at 96.88% purity; positive and negative purities are 99.47% and 93.38%. This exceeds either constituent model at the same thresholds by more than two percentage points, with lower coverage. A split-half check finds only 0.14 percentage points of threshold-selection optimism. A symmetric threshold of 0.86 adds 40 erroneous no-match labels, while 46 false abstentions persist even with no confidence threshold. Across five matched within-model comparisons, higher reasoning effort yields no clear F0.5 gain and increases median latency by 1.8-5.0 times. These results motivate separate thresholding and auditing for positive and negative pseudo-labels. The study establishes label purity, not student utility; fresh-data curation and student fine-tuning remain necessary to demonstrate downstream value.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI) ; Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.33878 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T20:29:01.387Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).