REPAIR: Resolving Long-Tail Confusion in Scientific Retrievers via Fact-Verified Iterative Refinement
cs.AI
Submitted: 2026-09-16
Updated: 2026-09-16
Comments: Accepted to EMNLP 2026 (Main Conference). 30 pages, 5 figures, 20 tables. Code: https://github.com/yerimoh/REPAIR
Code: https://github.com/yerimoh/REPAIR
License: http://creativecommons.org/licenses/by/4.0/
The gist: Precise retrieval of scientific information is fundamentally constrained by long-tailed concepts and high fact-sensitivity of scientific corpora.
Terminology
Abstract
Precise retrieval of scientific information is fundamentally constrained by long-tailed concepts and high fact-sensitivity of scientific corpora. These challenges often limit the effectiveness of dense retrievers and hallucination-prone LLM augmentation. To address this, we present REPAIR, a self-evolving data augmentation framework for scientific dense retrievers. REPAIR iteratively synthesizes training data to address knowledge gaps by cycling through diagnosis of long-tail concepts, API-guided evidence expansion, and differentiation via hard negative mining. This process effectively grounds retrieval in factual reality to resolve fine-grained distinctions. Extensive experiments demonstrate that REPAIR significantly outperforms 19 strong baselines on nine materials science and biomedical benchmarks. Our work highlights that diagnosing and factually augmenting data to long-tail deficits is essential for robust scientific retrieval.
Sources
- MS MARCO: A Human Generated MAchine Reading COmprehension Dataset
- LLM2Vec: Large Language Models Are Secretly Powerful Text Encoders
- M3-Embedding: Multi-Linguality, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation
- MedDialog: Two Large-scale Medical Dialogue Datasets
- Unsupervised Dense Information Retrieval with Contrastive Learning
- MELT: Materials-aware Continued Pre-training for Language Model Adaptation to Materials Science
- SGPT: GPT Sentence Embeddings for Semantic Search
- Text and Code Embeddings by Contrastive Pre-Training
- Gemma: Open Models Based on Gemini Research and Technology
- Llama 2: Open Foundation and Fine-Tuned Chat Models
- Text Embeddings by Weakly-Supervised Contrastive Pre-training
- Benchmarking Retrieval-Augmented Generation for Medicine
- Qwen2 Technical Report
- COCO-DR: Combating Distribution Shifts in Zero-Shot Dense Retrieval with Contrastive and Distributionally Robust Learning
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems
- VSAL: A Vision Solver with Adaptive Layouts for Graph Property Detection