Research map / Large language models / Language, safety and interpretability
Gradient-Based Inference for Networks with Output Constraints
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 |
|---|---|---|---|
| Apply gradient-based inference with parse-span constraints | SRL-100 · OntoNotes v5.0 test set | F1: 84.40 -> 85.63 (+1.23) | Helped |
| Use GBI instead of constrained A* decoding | SRL-X · OntoNotes v5.0, SRL-X failure sets | F1: 59.70 vs. 48.83 | Helped |
| Apply GBI with valid-tree constraints | Net2 · PTB section 23 | F1: 86.54 -> 87.57 | Helped |
| Apply GBI with input-dependent output-count constraint | 32-unit seq2seq LSTM · Synthetic transduction failure set | accuracy: 75.2% -> 82.4% | Helped |
| Apply GBI on out-of-domain examples | SRL-NW · Out-of-domain SRL genres | F1: 39.72 -> 52.4 (+12.68) | Helped |
| Use GBI with noisy constraints instead of constrained A* | SRL-100 · Noisy-constraint SRL | F1: 84.40 -> 84.87 (+0.47); 84.40 -> 77.48 (-6.92) | Mixed |
| Increase GBI maximum epochs from 10 to 30 | SRL-100 · OntoNotes v5.0 | F1: +0.34 overall F1; +3.4 F1 on failure set | Helped |
Closest papers by meaning
- Learning Approximate Inference Networks for Structured Prediction (ICLR 2018 · similarity 0.87)
- Neural Language Modeling by Jointly Learning Syntax and Lexicon (ICLR 2018 · similarity 0.87)
- Deep networks learn to parse uniform-depth context-free languages from local statistics (ICML 2026 · similarity 0.87)
- Sequence-to-Sequence Learning with Latent Neural Grammars (NeurIPS 2021 · similarity 0.86)
- A Primal Dual Formulation For Deep Learning With Constraints (NeurIPS 2019 · similarity 0.86)
- Training Neural Networks as Recognizers of Formal Languages (ICLR 2025 · similarity 0.86)
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.