2026 ACII DaiKon Workshop Sets Benchmark for Dyadic Affect Modeling

The 2026 ACII Dyadic Conversations (ACII-DaiKon) Workshop & Challenge sets a new standard for understanding interpersonal affect and social dynamics in two-person dialogues. Although there has been significant progress in modeling conversational affect, existing benchmarks often focus on individual speakers and fail to capture the evolving interactions between partners, such as directional influence, coordination of conversational timing, and the development of rapport. To fill this void, ACII-DaiKon introduces three interconnected sub-challenges that utilize a common dataset: predicting directional interpersonal influence, forecasting turn-taking (next speaker and time until next speech), and analyzing rapport trajectories throughout interactions. This challenge is based on the Hume-DaiKon dataset, which includes 945 dyadic conversations (totaling 743.4 hours of audiovisual data) gathered in natural settings across five languages, supporting multimodal modeling.