Selecting Open-Weight Language Models for Zero-Shot Intent Classification: A Systematic Evaluation of 41 Models
Source
arxiv.org
Author
Parishruthi Ganesh, Gerry Dozier, Cheryl Seals
Date
Why it matters
Concrete evidence that parameter count is the wrong selection axis for classification-style routing: tuned 3B models beat larger base models, and the benchmarks most people cite no longer separate anything.
Terms in this piece · Glossary
open weights — A model whose trained parameters are published for anyone to download and run — unlike API-only models you can access but never possess.
calibration — How well a model's confidence matches reality — a calibrated model saying "90% sure" is right about 90% of the time.
articleOptimismBench: Forecasting Bias and the Alignment Effect in Language Model JudgmentSeonglae Cho, Adriano Koshiyama
articleThe Open Weight Models That Matter June 2026OpenRouter
articleMeasuring the Microtask Eligibility Gap: When Is an Off-the-Shelf SLM Enough for an Agent Harness?Jundong Hu, Shekar Ramachandran