Get Started
Home
Topics
Search
Library
Research questionCan phase-transition counts during fine-tuning predict final test accuracy across architectures and distribution shifts?During fine-tuning, class separability may change in discrete jumps, but the relationship between the number of jumps and eventual test accuracy may depend on architecture and whether evaluation data are i.i.d. or corrupted.
Computer Vision
Evaluation & Benchmarks
Machine Learning
Statistical Machine Learning
Latest papersRecent research connected to this question, newest first.Phase Transition Frequency as a Training Time Predictor of Test Accuracy in ResNetsEvidence comes from 75 ResNet-18, ResNet-50, and ResNet-101 fine-tuning experiments across CIFAR-10, CIFAR-100, TinyImageNet, and CIFAR-10-C, using five to ten seeds per configuration. The association is strongest on the i.i.d. benchmarks and weaker under distributional stress; architecture-depth adjustment is demonstrated for CIFAR-100, while stricter categorical conditioning and comparisons with effective-rank, sharpness, Fisher-information, margin, and neural-collapse measures remain unestablished.research paper · Sep 4, 2026
Related questions
How can validation loss track language-model capabilities across changing training distributions without benchmark evaluation?How can small-scale pretraining mixture experiments stay reliable when scarce high-quality data is repeated at target scale?Can post-training ternarization make language models smaller without unacceptable capability loss or slower inference?How can interval-based time-series classifiers speed up frequent inference without materially reducing accuracy?