Research map / Architectures and efficiency / RNNs and normalization
Autoregressive Entity Retrieval
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 |
|---|---|---|---|
| Generate meaningful entity names rather than numerical IDs (ablation) | GENRE · WNED-KILT validation | accuracy: 68.5 -> 88.8 | Helped |
| Use constrained decoding with the candidate set rather than unconstrained generation (ablation) | GENRE · Entity disambiguation average | micro F1 (InKB): 79.6 -> 88.8 | Helped |
| Use the provided candidate set instead of the full KB (ablation) | GENRE · Entity disambiguation average | micro F1 (InKB): 85.1 -> 88.8 | Helped |
| Train on BLINK data in addition to AIDA data (ablation) | GENRE · Entity disambiguation average | micro F1 (InKB): 80.0 -> 88.8 | Helped |
| Train on AIDA data in addition to BLINK data (ablation) | GENRE · Entity disambiguation average | micro F1 (InKB): 88.1 -> 88.8 | Helped |
| Train jointly on the KILT tasks rather than only BLINK data (ablation) | GENRE · KILT average | R-Precision: 43.8 -> 69.7 | Helped |
| Train jointly on KILT tasks rather than only DPR data (ablation) | GENRE · KILT average | R-Precision: 42.1 -> 69.7 | Helped |
| Generate entity names instead of numerical IDs (ablation) | GENRE · KILT average | R-Precision: 19.0 -> 69.7 | Helped |
| Remove constrained decoding (ablation) | GENRE · KILT average | R-Precision: 63.0 -> 69.7 | Hurt |
| Use GENRE rather than BLINK + flair | page retrieval · KILT average | R-Precision: +13.7 R-precision points | Helped |
| Use GENRE rather than BLINK | entity retrieval · model/index memory footprint | memory: 14 times less memory | Helped |
| Use entity names rather than atomic numerical IDs (ablation) | GENRE · WNED-KILT validation, partial and no-match cases | accuracy: partial: 63.8 -> 86.9; no match: 55.0 -> 59.9; exact match: 76.0 -> 96.6 | Mixed |
Closest papers by meaning
- DocKS-RAG: Optimizing Document-Level Relation Extraction through LLM-Enhanced Hybrid Prompt Tuning (ICML 2025 · similarity 0.84)
- An Autoregressive Text-to-Graph Framework for Joint Entity and Relation Extraction (AAAI 2024 · similarity 0.84)
- UniversalNER: Targeted Distillation from Large Language Models for Open Named Entity Recognition (ICLR 2024 · similarity 0.84)
- Span Graph Transformer for Document-Level Named Entity Recognition (AAAI 2024 · similarity 0.84)
- Select and Augment: Enhanced Dense Retrieval Knowledge Graph Augmentation (Abstract Reprint) (AAAI 2024 · similarity 0.84)
- Web-Scale Visual Entity Recognition: An LLM-Driven Data Approach (NeurIPS 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.