analiza aiqbench
REPAIR introduces a self-evolving data augmentation framework specifically designed to mitigate long-tail confusion and hallucinations in scientific dense retrievers. By iteratively synthesizing training data through fact-verified evidence expansion, the method significantly improves retrieval accuracy on materials science and biomedical benchmarks without relying on generic LLM augmentation. This research advances the robustness of AI systems handling high-fact-sensitivity domains by grounding retrieval processes in factual reality.
REPAIR: Resolving Long-Tail Confusion in Scientific Retrievers via Fact-Verified Iterative Refinement
arXiv:2609.18262v1 Announce Type: new 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.
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.
Accepted to EMNLP 2026 (Main Conference). 30 pages, 5 figures, 20 tables. Code: this https URL
Artificial Intelligence (cs.AI)
arXiv:2609.18262 [cs.AI]
(or arXiv:2609.18262v1 [cs.AI] for this version)
https://doi.org/10.48550/arXiv.2609.18262 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 227 cuvinte
Subiecte conexe
What's Continual Learning, and Why Might We Expect To See It In Advanced LLM Agents?
LessWrong · 12 iun. 2026
Debating RSI, the US-China Gap, and Jaggedness with JS Denain of Epoch AI
Interconnects · 22 sept. 2026
I Did This Without Your Knowledge
LessWrong · 17 oct. 2026
The Entangled Dimensions of Decision Theory
LessWrong · 30 iul. 2026