Research

An inference-time decoding strategy that boosts LLM QA robustness under noisy and adversarial prompt conditions.

A Taiwan-centric tri-modal benchmark for Traditional Chinese QA that measures both model accuracy and latency trade-offs.