Research map / Large language models / Language, safety and interpretability
ICL-D3IE: In-Context Learning with Diverse Demonstrations Updating for Document Information Extraction
Jiabang He, Lei Wang, Yi Hu, Ning Liu, Hui Liu, Xing Xu, Heng Tao Shen
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 |
|---|---|---|---|
| add hard demonstrations (ablation) | ICL-D3IE (Davinci-003) · FUNSD (ID) | entity F1: 78.20 -> 90.32 | Improved |
| add layout-aware demonstrations (ablation) | ICL-D3IE (Davinci-003) · CORD (ID) | entity F1: 84.13 -> 94.12 | Improved |
| add layout-aware demonstrations (ablation) | ICL-D3IE (Davinci-003) · SROIE (ID) | entity F1: 96.83 -> 97.88 | No clear effect |
| add label mapping (ablation) | ICL-D3IE (Davinci-003) · CORD (ID) | entity F1: 87.94 -> 94.12 | Improved |
| add formatting demonstrations (ablation) | ICL-D3IE (Davinci-003) · SROIE (ID) | entity F1: 90.58 -> 97.88 | Improved |
| add formatting demonstrations (ablation) | ICL-D3IE (Davinci-003) · CORD (ID) | entity F1: 93.07 -> 94.12 | Improved |
| use ICL-D3IE instead of standard ICL | Davinci-003 · FUNSD (ID) | entity F1: 71.52 -> 90.32 | Improved |
| ICL-D3IE vs fully fine-tuned LayoutLMv3 on OOD text perturbations | ICL-D3IE (Davinci-003) · CORD (OOD) | entity F1: 75.23 -> 91.23 | Improved |
| ICL-D3IE vs fully fine-tuned LayoutLMv3 on ID CORD | ICL-D3IE (Davinci-003) · CORD (ID) | entity F1: 96.56 -> 94.12 | Hurt |
| use ChatGPT instead of Davinci-003 as backbone | ICL-D3IE · CORD (ID) | entity F1: 94.12 -> 87.13 | Hurt |
| order demonstrations M-H-L-F instead of M-L-H-F (ablation) | ICL-D3IE (Davinci-003) · FUNSD, CORD, SROIE | entity F1 | Improved |
| update hard demonstrations iteratively (up to 20 iterations) (ablation) | ICL-D3IE (Davinci-003) · FUNSD, CORD, SROIE | entity F1 | Improved |
Later papers that compared with this method
1 later accepted paper reports ICL-D3IE as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- SAIL: Sample-Centric In-Context Learning for Document Information Extraction (AAAI 2025 · better on 3 of 3 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- LIVE: Learnable In-Context Vector for Visual Question Answering (NeurIPS 2024 · similarity 0.88)
- Making Text Embedders Few-Shot Learners (ICLR 2025 · similarity 0.88)
- In-Context Pretraining: Language Modeling Beyond Document Boundaries (ICLR 2024 · similarity 0.87)
- From Unstructured Data to In-Context Learning: Exploring What Tasks Can Be Learned and When (NeurIPS 2024 · similarity 0.87)
- Mixture of Demonstrations for In-Context Learning (NeurIPS 2024 · similarity 0.87)
- In-Context Language Learning: Architectures and Algorithms (ICML 2024 · similarity 0.87)
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.
