Research map / Large language models / Retrieval and knowledge
Scientific Document Retrieval using Multi-level Aspect-based Queries
Jianyou Wang, Kaicheng Wang, Xiaoyue Wang, Prudhviraj Naidu, Leon Bergen, Ramamohan Paturi
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 |
|---|---|---|---|
| fine-tune with supervised contrastive learning on DORIS-MAE training queries | E5-L-v2 · DORIS-MAE 60 test queries | MAP: 40.62 -> 49.15 | Improved |
| fine-tune with SCL | E5-L-v2 · DORIS-MAE 60 test queries | MRR@10: 14.53 -> 23.16 | Improved |
| fine-tune with SCL | E5-L-v2 · DORIS-MAE 60 test queries | R@20: 42.38 -> 52.45 | Improved |
| give retrievers concatenated aspects instead of the full query | SentBERT · DORIS-MAE | R@5: 17.73 | Improved |
| use 2-aspect sub-queries instead of full queries | ada-002 · DORIS-MAE | RP: 35.81 -> 47.35 | Improved |
| use 2-aspect sub-queries instead of full queries | ada-002 · DORIS-MAE | MAP: 40.37 -> 51.67 | Improved |
| give ada-002 aspect access | ada-002 · DORIS-MAE | MAP: 40.37 -> 37.62 | Hurt |
| use aspect-aware ASPIRE models instead of general models | TSAspire/OTAspire · DORIS-MAE | R@5: 14.26 vs 16.51 | No clear effect |
| apply to DORIS-MAE rather than MS MARCO | ColBERT-v2 · MRR@10 | MRR@10: 39.7 -> 6.69 | Hurt |
| apply to DORIS-MAE rather than NQ | RocketQA-v2 · R@20 | R@20: 83.7 -> 45.41 | Hurt |
| ChatGPT instead of a human annotator | Anno-GPT annotation · Annotation agreement | accuracy: 64.13 vs 62.07 | No clear effect |
| ChatGPT instead of a human annotator | Anno-GPT annotation · Annotation agreement | Spearman rho: 54.31 vs 46.61 | No clear effect |
Closest papers by meaning
- SPIQA: A Dataset for Multimodal Question Answering on Scientific Papers (NeurIPS 2024 · similarity 0.85)
- BRIGHT: A Realistic and Challenging Benchmark for Reasoning-Intensive Retrieval (ICLR 2025 · similarity 0.85)
- FreshStack: Building Realistic Benchmarks for Evaluating Retrieval on Technical Documents (NeurIPS 2025 · similarity 0.84)
- UDA: A Benchmark Suite for Retrieval Augmented Generation in Real-World Document Analysis (NeurIPS 2024 · similarity 0.84)
- Docopilot: Improving Multimodal Models for Document-Level Understanding (CVPR 2025 · similarity 0.84)
- Holistic Reasoning with Long-Context LMs: A Benchmark for Database Operations on Massive Textual Data (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.
