# Not All Forgetting Is Equal: Retention Dynamics in Fine-Tuned Image Classifiers

arXiv: https://arxiv.org/abs/2604.11508 · HTML: https://mlsanity.com/paper/2604.11508/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 17 of its 34 references that are in the map.
- First posted on arXiv 13 Apr 2026.

## Builds on (most cited first, 17 of 17)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [Overcoming catastrophic forgetting in neural networks](https://mlsanity.com/paper/1612.00796/) (2016)
- [Training data-efficient image transformers & distillation through attention](https://mlsanity.com/paper/2012.12877/) (2020)
- [Do Vision Transformers See Like Convolutional Neural Networks?](https://mlsanity.com/paper/2108.08810/) (2021)
- [Deep Learning on a Data Diet: Finding Important Examples Early in Training](https://mlsanity.com/paper/2107.07075/) (2021)
- [Beyond neural scaling laws: beating power law scaling via data pruning](https://mlsanity.com/paper/2206.14486/) (2022)
- [Does Learning Require Memorization? A Short Tale about a Long Tail](https://mlsanity.com/paper/1906.05271/) (2019)
- [Dataset Cartography: Mapping and Diagnosing Datasets with Training Dynamics](https://mlsanity.com/paper/2009.10795/) (2020)
- [Identifying Mislabeled Data using the Area Under the Margin Ranking](https://mlsanity.com/paper/2001.10528/) (2020)
- [Common pitfalls and recommendations for using machine learning to detect and prognosticate for COVID-19 using chest radiographs and CT scans](https://mlsanity.com/paper/2008.06388/) (2020)
- [Let's Agree to Agree: Neural Networks Share Classification Order on Real Datasets](https://mlsanity.com/paper/1905.10854/) (2019)
- [Towards Robust and Reproducible Active Learning Using Neural Networks](https://mlsanity.com/paper/2002.09564/) (2020)
- [Parting with Illusions about Deep Active Learning](https://mlsanity.com/paper/1912.05361/) (2019)
- [How I failed machine learning in medical imaging -- shortcomings and recommendations](https://mlsanity.com/paper/2103.10292/) (2021)
- [Characterizing Datapoints via Second-Split Forgetting](https://mlsanity.com/paper/2210.15031/) (2022)
- [OCTDL: Optical Coherence Tomography Dataset for Image-Based Deep Learning Methods](https://mlsanity.com/paper/2312.08255/) (2023)
- [Human-like Forgetting Curves in Deep Neural Networks](https://mlsanity.com/paper/2506.12034/) (2025)

## Abstract (arXiv)

Fine-tuning a pretrained classifier leaves some samples reliably learned and others cycling between correct and incorrect. Curriculum learning, data pruning and dataset cartography assume that pattern is a property of the sample, untested. We record per-sample correctness at every epoch while fine-tuning ResNet-18 and DeiT-Small on an imbalanced retinal OCT dataset and CUB-200-2011, matching samples by image identity and holding the split fixed across seeds. Per-sample retention is reproducible: cross-run Spearman correlation of the fitted decay constant is 0.37 to 0.59 over ten seeds. It is architecture-specific: two runs of one backbone agree more than two backbones on identical data (0.45 and 0.59 within against 0.30 between on OCTDL). Loss after five frozen-backbone epochs predicts a different run's decay constant at 0.29 to 0.43. The Ebbinghaus exponential does not survive: monotone decay, the one shape it can represent, is 0.1% to 0.8% of samples, and on traces that do forget mean R-squared is negative in all four configurations. A power law and a free-asymptote variant fail on the same traces: the defect is monotonicity. Across five sampling arms with matched exposure, prioritisation ratios of 2.7x to 28x, and an online variant, none of 48 comparisons against uniform sampling survives Benjamini-Hochberg correction, though three seeds detect only about four accuracy points. A stable, cheap difficulty score does not buy generalisation through sampling. Patient-grouped splitting, the remedy for a leak reaching 76% to 78% of OCT test images, moves that dataset's headline metrics by less than their run-to-run spread.