Research map / Label-efficient and robust learning / Recognition and long-tail
Ranking Neural Checkpoints
Yandong Li, Xuhui Jia, Ruoxin Sang, Yukun Zhu, Bradley Green, Liqiang Wang, Boqing Gong
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace classification-head softmax probabilities with GMM cluster posteriors in LEEP | Checkpoint ranking with LEEP · NeuCRaB Group II | Kendall: 55.49 -> 67.49 | Improved |
| Replace classification-head softmax probabilities with GMM cluster posteriors in LEEP | Checkpoint ranking with LEEP · NeuCRaB Group III | Kendall: 45.06 -> 51.14 | Improved |
| Apply PCA dimensionality reduction before mutual-information estimation (ablation) | MI (α=0.01) · NeuCRaB Group I | Kendall: -0.31 -> 20.67 | Improved |
| Apply PCA dimensionality reduction before mutual-information estimation (ablation) | MI (α=0.01) · NeuCRaB Group II | Kendall: -17.81 -> 51.34 | Improved |
| Use α=0.01 instead of α=0.50 (ablation) | MI w/ PCA · NeuCRaB Group III | Kendall: 21.26 -> 38.83 | Improved |
| Use converged rather than 1-epoch linear-classifier training (ablation) | Linear checkpoint ranking · NeuCRaB Group I | Kendall: 18.44 -> 53.43 | Improved |
| Use converged rather than 1-epoch fine-tuning (ablation) | Fine-tuning checkpoint ranking · NeuCRaB Group I | Kendall: 22.15 -> 36.78 | Mixed |
| Use 5-epoch rather than 1-epoch fine-tuning (ablation) | Fine-tuning checkpoint ranking · NeuCRaB Group III | Kendall: 11.59 -> 15.72 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- LEEP — LEEP: A New Measure to Evaluate Transferability of Learned Representations (ICML 2020) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Boost Neural Networks by Checkpoints (NeurIPS 2021 · similarity 0.86)
- Optimal Gradient Checkpoint Search for Arbitrary Computation Graphs (CVPR 2021 · similarity 0.84)
- A Bayesian Model Selection Criterion for Selecting Pretraining Checkpoints (ICML 2025 · similarity 0.81)
- Maximizing Intermediate Checkpoint Value in LLM Pretraining with Bayesian Optimization (ICML 2025 · similarity 0.81)
- Enhancing Adversarial Transferability with Checkpoints of a Single Model's Training (CVPR 2025 · similarity 0.81)
- When to Learn and When to Stop: Quitting at the Optimal Time (Student Abstract) (AAAI 2025 · similarity 0.80)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
