Research map / Probabilistic and scientific modeling / Molecules and proteins
Search-Guided, Lightly-Supervised Training of Structured Prediction Energy Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 reward-guided truncated randomized search to generate local-improvement constraints | SPEN training · Cora citation-field extraction | token-level accuracy: 48.3\% -> 57.1\% | Helped |
| Use reward-guided truncated randomized search to generate local-improvement constraints | SPEN training · Bibtex | F1: 40.1 -> 44.0 | Helped |
| Use reward-guided truncated randomized search to generate local-improvement constraints | SPEN training · Bookmarks | F1: 30.6 -> 38.4 | Helped |
| Use search-guided training instead of neural shape parser policy gradients | Shape parsing · Shape parsing test | IOU: 32.4\% -> 56.3\% | Helped |
| Use a reward margin of 0.001 rather than the selected larger margin (ablation) | SG-SPEN · Bibtex training | F1 | Hurt |
| Use a reward margin of 0.002 rather than 0.01 (ablation) | SG-SPEN · Bibtex training | F1 | Hurt |
| Increase the reward margin and search budget (ablation) | SG-SPEN · structured prediction tasks | accuracy | Helped |
| Use a smaller reward margin and search budget (ablation) | SG-SPEN · structured prediction tasks | training speed | Helped |
| Use unlabeled training examples alongside limited labeled data (ablation) | SG-SPEN · Cora citation extraction, 5 labeled examples | score: 53.0 -> 65.5 | Helped |
| Use unlabeled training examples alongside limited labeled data (ablation) | SG-SPEN · Cora citation extraction, 10 labeled examples | score: 62.4 -> 71.7 | Helped |
| Use unlabeled training examples alongside limited labeled data (ablation) | SG-SPEN · Cora citation extraction, 50 labeled examples | score: 81.6 -> 82.9 | Helped |
| Use exponentiated gradient descent instead of other inference updates | SPEN inference · structured prediction | inference performance | Helped |
Closest papers by meaning
- Structured Prediction Energy Networks (ICML 2016 · similarity 0.91)
- Only Strict Saddles in the Energy Landscape of Predictive Coding Networks? (NeurIPS 2024 · similarity 0.84)
- Benchmarking Predictive Coding Networks -- Made Simple (ICLR 2025 · similarity 0.84)
- Bandit and Delayed Feedback in Online Structured Prediction (NeurIPS 2025 · similarity 0.83)
- SOInter: A Novel Deep Energy-Based Interpretation Method for Explaining Structured Output Models (ICLR 2024 · similarity 0.83)
- Encodings for Prediction-based Neural Architecture Search (ICML 2024 · similarity 0.83)
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.