Research map / Probabilistic and scientific modeling / Molecules and proteins
Exploiting a Zoo of Checkpoints for Unseen Tasks
Jiaji Huang, Qiang Qiu, Kenneth Church
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 |
|---|---|---|---|
| Select five checkpoints using MMI instead of the bert-large-cased best layer | Checkpoint feature selection · CoNLL2000 Chunking | score: 93.64 -> 93.88 | Improved |
| Select five checkpoints using MMI instead of the bert-large-cased best layer | Checkpoint feature selection · CoNLL2003 NER | score: 84.44 -> 88.64 | Improved |
| Select five checkpoints using MMI instead of the bert-large-cased best layer | Checkpoint feature selection · PTB POS | score: 96.73 -> 97.57 | Improved |
| Select five checkpoints using MMI instead of the bert-large-cased best layer | Checkpoint feature selection · Semantic Tagging | score: 93.83 -> 94.35 | Improved |
| Select five checkpoints using MMI instead of the bert-large-cased best layer | Checkpoint feature selection · ItHappened-v2 Event Factuality | Pearson correlation: 76.25 -> 78.35 | Improved |
| Select five checkpoints using MMI instead of the bert-large-cased best layer | Checkpoint feature selection · PTB Parent tagging | score: 96.10 -> 96.64 | Improved |
| Select five checkpoints using MMI instead of the bert-large-cased best layer | Checkpoint feature selection · PTB Gparent tagging | score: 82.46 -> 85.85 | Improved |
| Select five checkpoints using MMI instead of the bert-large-cased best layer | Checkpoint feature selection · PTB GGparent tagging | score: 67.90 -> 73.32 | Improved |
| Select checkpoints with MMI rather than randomly | Checkpoint feature selection · CIFAR100 unseen 30-way tasks | average validation accuracy | Improved |
| Choose checkpoints based on the first unseen task rather than MMI selection | Checkpoint feature selection · CIFAR100 unseen 30-way tasks | average validation accuracy | Mixed |
Closest papers by meaning
- A Bayesian Model Selection Criterion for Selecting Pretraining Checkpoints (ICML 2025 · similarity 0.84)
- Taskonomy: Disentangling Task Transfer Learning (CVPR 2018 · similarity 0.84)
- Localizing Task Information for Improved Model Merging and Compression (ICML 2024 · similarity 0.84)
- SAFE: Multitask Failure Detection for Vision-Language-Action Models (NeurIPS 2025 · similarity 0.82)
- Identifying Selections for Unsupervised Subtask Discovery (NeurIPS 2024 · similarity 0.82)
- Treasure Hunt: Real-time Targeting of the Long Tail using Training-Time Markers (NeurIPS 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.
