Research map / Label-efficient and robust learning / Recognition and long-tail
With Argus Eyes: Assessing Retrieval Gaps via Uncertainty Scoring to Detect and Remedy Retrieval Blind Spots
Zeinab Sadat Taghavi, Ali Modarressi, Hinrich Schuetze, Andreas Marfurt
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 ARGUS Document Expansion | dense retrievers (avg over 8) · BRIGHT full avg | nDCG@5/10 average: +3.44 | Improved |
| add ARGUS Document Expansion | dense retrievers (avg over 8) · ImpliRet full avg | nDCG@5/10 average: +6.76 | Improved |
| add ARGUS Document Expansion | dense retrievers (avg over 8) · RAR-b full avg | nDCG@5/10 average: +1.68 | Improved |
| add ARGUS LLM Synthesis | dense retrievers (avg over 8) · BRIGHT full avg | nDCG@5/10 average: +2.44 | Improved |
| add ARGUS LLM Synthesis | dense retrievers (avg over 8) · ImpliRet full avg | nDCG@5/10 average: +2.21 | Improved |
| add ARGUS LLM Synthesis | dense retrievers (avg over 8) · RAR-b full avg | nDCG@5/10 average: +1.81 | Improved |
| use LLM Synthesis instead of Document Expansion | Jina-v3 · ImpliRet | nDCG@10: 17.9 -> 16.4 | Hurt |
| add low-RPS targeting instead of random KB augmentation (ablation) | BGE-M3 · ImpliRet (2/2) | nDCG: 21.74 -> 36.15 | Improved |
| add RPS-based selection over augmenting all entities (ablation) | BGE-M3 · ImpliRet (2/2) | nDCG: 31.42 -> 36.15 | Improved |
| add RPS-based selection over augmenting all entities (ablation) | ReasonIR · ImpliRet (2/2) | nDCG: 27.41 -> 30.60 | Improved |
| raise threshold tau from 0.3 to 1.0 (augment all) (ablation) | BGE-M3 · ImpliRet (2/2) | nDCG: 36.15 -> 33.45 | Hurt |
| lower threshold tau from 0.3 to 0.2 (ablation) | ReasonIR · BRIGHT (3/10) | nDCG: 21.13 -> 18.91 | Hurt |
Closest papers by meaning
- RAGGED: Towards Informed Design of Scalable and Stable RAG Systems (ICML 2025 · similarity 0.88)
- Worse than Zero-shot? A Fact-Checking Dataset for Evaluating the Robustness of RAG Against Misleading Retrievals (NeurIPS 2025 · similarity 0.87)
- SeCon-RAG: A Two-Stage Semantic Filtering and Conflict-Free Framework for Trustworthy RAG (NeurIPS 2025 · similarity 0.87)
- Completing Missing Annotation: Multi-Agent Debate for Accurate and Scalable Relevant Assessment for IR Benchmarks (ICLR 2026 · similarity 0.86)
- Follow My Instruction and Spill the Beans: Scalable Data Extraction from Retrieval-Augmented Generation Systems (ICLR 2025 · similarity 0.86)
- Speculative RAG: Enhancing Retrieval Augmented Generation through Drafting (ICLR 2025 · similarity 0.86)
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.
