← Chen, Ho Yiing — Research Records

Internal Specialization Scales with Model Size: Quantitative Evidence from qwen2.5 3B vs 7B

Chen, Ho Yiing · 2026-05-21 · Zenodo

doi:10.5281/zenodo.20325676 · PDF

Abstract

© Copyright 2026 Chen, Ho Yiing (ORCID: 0009-0006-6816-9891), Independent researcher, charenix.com. All rights reserved by the author. This work is licensed under the Creative Commons Attribution 4.0. Folk wisdom in the LLM community holds that bigger transformers 'specialize more.' I give that claim a cell-level test. Running identical ten-prompt observation passes through qwen2.5:3B and qwen2.5:7B on a Mercury sensor grid produces directly comparable signature counts at each size. The 7B model has 2.08x as many cross-lingual physics-topic cells , 1.66x as many Chinese-only sensors, and 1.56x

Chen, Ho Yiing (norika) · Independent Researcher, Taiwan · ORCID 0009-0006-6816-9891