Search papers, labs, and topics across Lattice.
This paper tackles the challenges of mixed-format medical visual question answering (VQA) by introducing a candidate-expanding routing mechanism that integrates a permutation-stabilized vision-language expert with an answer-text memory. The method addresses the instability in multiple-choice predictions and the evaluation difficulties of open answers, achieving a significant accuracy improvement from 88.95% to 91.73% in a retrospective analysis of 1,403 cases. Additionally, the approach enhances oracle coverage from 90.31% to 96.15%, demonstrating the effectiveness of candidate expansion in improving routing performance and output validity in medical contexts.
Expanding candidate options in medical VQA boosts accuracy by nearly 3 percentage points, revealing the critical role of controlled routing in complex question-answering tasks.
Mixed-format medical visual question answering (VQA) requires stable option selection and machine-readable free-text output. The two formats fail differently: multiple-choice predictions can change with option symbols or positions, while clinically plausible open answers can fail automated evaluation when serialization is malformed. We address both challenges with an answer-text memory, a permutation-stabilized vision--language expert, and a sparse candidate- expanding router. The cyclic schedule follows prior work; our contribution is to make expert top-2 a routable candidate alongside memory and expert top-1. On a 1,403-case retrospective internal analysis, this expansion improves a matched binary router from 88.95% to 91.73% (+2.78 percentage points; 95% CI 1.57--3.99), with 56 rescued errors and 17 regressions. Oracle coverage rises from 90.31% to 96.15%, and the final submitted configuration reaches 92.23% on the same retrospective split. For open questions, strict generation and deterministic guards produce 475/475 schema- valid participant-facing outputs without repair, retry, or hard-gate failure. Visual ablations reveal substantial textual dependence. Candidate expansion supplies the principal controlled routing gain; open-path evidence establishes output-contract validity rather than clinical correctness in medical use or deployment.
