Research map / Large language models / Language, safety and interpretability
Learning Contextual Representations for Semantic Parsing with Generation-Augmented Pre-Training
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 |
|---|---|---|---|
| Replace the BART encoder with GAP Model | RAT-SQL · Spider test | exact set match accuracy: 0.651 -> 0.697 | Helped |
| Replace the BART encoder with GAP Model | YXJ · Criteria-to-SQL test | SQL accuracy: 0.307 -> 0.327 | Helped |
| Replace the BART encoder with GAP Model | YXJ · Criteria-to-SQL test | execution accuracy: 0.558 -> 0.594 | Helped |
| Add GenSQL pre-training (ablation) | IRNet with GAP pre-training · Spider dev | accuracy: 0.680 -> 0.699 | Helped |
| Add MLM pre-training without GenSQL (ablation) | IRNet + BART Encoder · Spider dev | accuracy: 0.680 -> 0.697 | Helped |
| Add CPred pre-training without GenSQL (ablation) | IRNet + BART Encoder · Spider dev | accuracy: 0.680 -> 0.699 | Helped |
| Add CRec pre-training without GenSQL (ablation) | IRNet + BART Encoder · Spider dev | accuracy: 0.680 -> 0.705 | Helped |
| Add CRec to MLM while using GenSQL (ablation) | IRNet + GAP MLM + GenSQL · Spider dev | accuracy: 0.717 -> 0.728 | Helped |
| Add CPred to MLM while using GenSQL (ablation) | IRNet + GAP MLM + GenSQL · Spider dev | accuracy: 0.717 -> 0.716 | No clear effect |
| Remove utterance input during MLM pre-training (ablation) | GAP MLM · Spider dev | accuracy: 0.697 -> 0.678 | Hurt |
| Remove schema input during MLM pre-training (ablation) | GAP MLM · Spider dev | accuracy: 0.697 -> 0.679 | Hurt |
| Replace generated utterances with table surrounding text (ablation) | GAP MLM · Spider dev | accuracy: 0.697 -> 0.679 | Hurt |
Closest papers by meaning
- Narrowing the Gap between Supervised and Unsupervised Sentence Representation Learning with Large Language Model (AAAI 2024 · similarity 0.86)
- Provence: efficient and robust context pruning for retrieval-augmented generation (ICLR 2025 · similarity 0.86)
- In-Context Pretraining: Language Modeling Beyond Document Boundaries (ICLR 2024 · similarity 0.86)
- Filling Memory Gaps: Enhancing Continual Semantic Parsing via SQL Syntax Variance-Guided LLMs Without Real Data Replay (AAAI 2025 · similarity 0.86)
- Enhancing Tabular Learners with Context-Aware Semantic Embeddings (NeurIPS 2026 · similarity 0.85)
- In-Context Language Learning: Architectures and Algorithms (ICML 2024 · similarity 0.85)
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.