The authors present a submission to the QANTA 2026 shared challenge at the ICML 2026 Workshop on Efficient Multimodal Question Answering (EMM-QA), achieving the highest overall leaderboard score of 0.402.
- The system employs a task-specific two-agent architecture designed for pyramid-style questions with incrementally revealed text and images.
- The Tossup agent uses a GPT-4o-mini-class model with confidence-calibrated answering and a numeric reasoning policy to reduce overconfident predictions.
- The Bonus agent utilizes a GPT-4o-class model with leadin-aware reasoning, structured relational reasoning, and multimodal evidence integration.
- The approach avoids retrieval pipelines or model ensembles, relying instead on efficient reasoning policies within a hosted-only environment.
The results demonstrate that lightweight, task-specific reasoning strategies can provide strong performance on resource-constrained multimodal question answering benchmarks.