Model-Centric Diagnostics: A Framework for Internal State Readouts
Fuente:
arXiv
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911412849213440 |
|---|---|
| author | Wu, Fangzheng Summa, Brian |
| author_facet | Wu, Fangzheng Summa, Brian |
| contents | We present a model-centric diagnostic framework that treats training state as a latent variable and unifies a family of internal readouts -- head-gradient norms, confidence, entropy, margin, and related signals -- as anchor-relative projections of that state. A preliminary version of this work introduced a head-gradient probe for checkpoint selection. In this version, we focus on the unifying perspective and structural diagnostics; full algorithmic details, theoretical analysis, and experimental validation will appear in a forthcoming paper.
We outline the conceptual scaffold: any prediction head induces a local loss landscape whose geometry (gradient magnitude, curvature, sharpness) reflects how well the upstream features are aligned with the task. Different readout choices -- gradient norms, softmax entropy, predictive margin -- correspond to different projections of this geometry, each with complementary strengths. The framework suggests that checkpoint selection, early stopping, and lightweight architecture pre-screening can all be viewed as querying the same underlying state through different lenses.
Illustrative experiments on ImageNet classification and COCO detection/segmentation hint at the practical potential; rigorous benchmarks and ablations are deferred to the full paper. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_16874 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Model-Centric Diagnostics: A Framework for Internal State Readouts Wu, Fangzheng Summa, Brian Computer Vision and Pattern Recognition Artificial Intelligence We present a model-centric diagnostic framework that treats training state as a latent variable and unifies a family of internal readouts -- head-gradient norms, confidence, entropy, margin, and related signals -- as anchor-relative projections of that state. A preliminary version of this work introduced a head-gradient probe for checkpoint selection. In this version, we focus on the unifying perspective and structural diagnostics; full algorithmic details, theoretical analysis, and experimental validation will appear in a forthcoming paper. We outline the conceptual scaffold: any prediction head induces a local loss landscape whose geometry (gradient magnitude, curvature, sharpness) reflects how well the upstream features are aligned with the task. Different readout choices -- gradient norms, softmax entropy, predictive margin -- correspond to different projections of this geometry, each with complementary strengths. The framework suggests that checkpoint selection, early stopping, and lightweight architecture pre-screening can all be viewed as querying the same underlying state through different lenses. Illustrative experiments on ImageNet classification and COCO detection/segmentation hint at the practical potential; rigorous benchmarks and ablations are deferred to the full paper. |
| title | Model-Centric Diagnostics: A Framework for Internal State Readouts |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2601.16874 |