Research map / Large language models / Language, safety and interpretability
Model Spider: Learning to Rank Pre-Trained Models Efficiently
Yi-Kai Zhang, Ting-Ji Huang, Yao-Xiang Ding, De-Chuan Zhan, Han-Jia Ye
Read the paperCheck 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 |
|---|---|---|---|
| use RankAgg aggregated ranking as training supervision instead of H-Score (ablation) | Model Spider · 8 datasets mean | weighted tau_w: 0.642 -> 0.765 | Improved |
| use RankAgg instead of LogME as supervision (ablation) | Model Spider · 8 datasets mean | weighted tau_w: 0.689 -> 0.765 | Improved |
| use RankAgg instead of H-Score supervision (ablation) | Model Spider · CIFAR10 | weighted tau_w: 0.386 -> 0.845 | Improved |
| use RankAgg instead of LogME supervision (ablation) | Model Spider · CIFAR10 | weighted tau_w: 0.695 -> 0.845 | Improved |
| add PTM-specific features (k greater than 0) to general tokens (ablation) | Model Spider · Aircraft/Caltech101 ablation | weighted tau_w | Improved |
| weaker fixed feature extractor psi (tiny Swin from EsViT, half the parameters) (ablation) | Model Spider · single-source zoo | weighted tau_w | No clear effect |
| halve the number of training tasks (ablation) | Model Spider · single-source zoo | weighted tau_w | Mixed |
| Model Spider vs best baseline LogME (standard) | LogME · 8 datasets mean | weighted tau_w: 0.676 -> 0.800 | Improved |
| Model Spider vs best baseline LogME (few-shot) | LogME · 8 datasets mean few-shot | weighted tau_w: 0.574 -> 0.750 | Improved |
| Model Spider vs LogME on Aircraft (standard) | LogME · Aircraft | weighted tau_w: 0.540 -> 0.506 | Hurt |
| Model Spider vs LogME on dSprites regression | LogME · dSprites | weighted tau_w: 0.612 -> 0.679 | Improved |
| Model Spider vs H-Score on UTKFace regression | H-Score · UTKFace | weighted tau_w: 0.075 -> 0.364 | Improved |
Closest papers by meaning
- Model Swarms: Collaborative Search to Adapt LLM Experts via Swarm Intelligence (ICML 2025 · similarity 0.83)
- ModelLens: Finding the Best for Your Task from Myriads of Models (NeurIPS 2026 · similarity 0.82)
- Weak-to-Strong Generalization under Distribution Shifts (NeurIPS 2025 · similarity 0.82)
- How to Train Your LLM Web Agent: A Statistical Diagnosis (NeurIPS 2025 · similarity 0.82)
- Optimal Representation Size: High-Dimensional Analysis of Pretraining and Linear Probing (NeurIPS 2026 · similarity 0.82)
- Specialized Foundation Models Struggle to Beat Supervised Baselines (ICLR 2025 · similarity 0.82)
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.
