Research map / Probabilistic and scientific modeling / Molecules and proteins
LasUIE: Unifying Information Extraction with Latent Adaptive Structure-aware Generative Language Model
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 |
|---|---|---|---|
| add structural broadcaster (SB) (ablation) | LasUIE · 12 IE datasets avg | F1: 74.6 -> 75.9 | Helped |
| add structure diversifying regularization (ablation) | LasUIE · 12 IE datasets avg | F1: 75.6 -> 75.9 | Helped |
| add task-oriented structure fine-tuning (ablation) | LasUIE · 12 IE datasets avg | F1: 74.9 -> 75.9 | Helped |
| induce latent structure via post-training instead of injecting external parse trees | UIE · 12 IE datasets avg | F1: 74.6 -> 75.9 | Helped |
| inject dependency and constituency syntax from external parser | UIE · 12 IE datasets avg | F1: 73.1 -> 74.6 | Helped |
| inject both dependency and constituency syntax instead of one type (ablation) | GEN-T5 · 12 IE datasets avg | F1: 73.2 -> 73.9 (DepSyn only vs both) | Helped |
| unified IE modeling instead of separate | GEN-T5 vs UIE · 12 IE datasets avg | F1: 72.0 -> 73.1 | Helped |
| use LasUIE in 1-shot low-resource setting | UIE +Dep&ConSyn · 1-shot avg | F1: 16.4 -> 34.3 | Helped |
| use LasUIE with 10% data | UIE +Dep&ConSyn · 10% data avg | F1: 64.4 -> 67.2 | Helped |
| use LasUIE on CoNLL03 1% data vs UIE with IE pretraining | UIE with supervised IE pretraining · CoNLL03 1% data | F1: 82.8 vs 82.1 | Hurt |
Closest papers by meaning
- UniGen: A Unified Generative Framework for Retrieval and Question Answering with Large Language Models (AAAI 2024 · similarity 0.84)
- Dynamically Structured Diffusion Language Model Decoding via Bayesian Inference (NeurIPS 2026 · similarity 0.84)
- Precise Information Control in Long-Form Text Generation (NeurIPS 2025 · similarity 0.84)
- DISP-LLM: Dimension-Independent Structural Pruning for Large Language Models (NeurIPS 2024 · similarity 0.84)
- STAR: Boosting Low-Resource Information Extraction by Structure-to-Text Data Generation with Large Language Models (AAAI 2024 · similarity 0.84)
- Generative Adapter: Contextualizing Language Models in Parameters with A Single Forward Pass (ICLR 2025 · similarity 0.84)
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.