Search papers, labs, and topics across Lattice.
This paper details the BioASQ Task 14B 2026 system, focusing on a dual approach to retrieval and answer combination that enhances performance while minimizing costs. A hybrid retrieval strategy, integrating dense BGE, BM25, and RRF, achieved a remarkable R@200 score of 99.3%, while a cost-pragmatic re-retrieval policy significantly improved list F1 and precision by 12% at a lower cost compared to traditional methods. Additionally, the study reveals that a synonym-union resolver outperformed on list recall metrics, leading to a top placement on multiple leaderboards, showcasing the effectiveness of their multi-model combiners in complex question-answering scenarios.
A cost-pragmatic retrieval strategy not only boosts performance metrics but also reduces operational costs, challenging conventional wisdom in multi-model systems.
We describe our BioASQ Task 14B 2026 system. The work centers on two design decisions: how aggressively to re-retrieve when first-stage retrieval is weak, and how to combine multiple language-model answers. Retrieval unions two parallel pipelines - a hybrid first stage (dense BGE + BM25 + RRF, reaching R@200 = 99.3% on the BioASQ-13b historical archive) and an agent-driven pipeline that decomposes the question over PubMed, Europe PMC, and iCite - with a BGE cross-encoder quality gate flagging weakly-supported questions for selective re-retrieval. On Task 12B 2024 validation, a cost-pragmatic re-retrieval policy beats a skill-strict baseline significantly on list F1 and list precision, at 12% lower re-retrieval cost. Holding prompt and model fixed across val and test 13B (different question sets), list F1 rises by +0.132 absolute on the BioASQ-released gold-input pool, consistent with substantial retrieval-side headroom. For Phase B answering we decompose multi-model ensemble lift into a selection component bounded by the per-question oracle and a fusion component that aggregators can exceed. The decomposition predicts before any experiment that LLM-as-judge wins on selection-dominated metrics (yes/no, multi-reference ROUGE) but is structurally insufficient on the recall component of fusion-friendly metrics (factoid rank-1, list recall). On Task 13B 2025 our synonym-union resolver wins list recall on every head, while GPT-5.5 solo retains the list-F1 lead because the resolver's wider item set costs precision. On the Task 14B 2026 preliminary leaderboard our team places first on the combined-exact aggregate on three of the eight (phase x batch) leaderboards, wins four individual question-type cells, and takes #1 on Phase B b3 ideal.
