MMAR

The MMAR dataset is used to evaluate the performance of models in general audio understanding tasks.

Papers using MMAR (5)

DIFFA-2: A Practical Diffusion Large Language Model for General Audio Understanding2026EntangleCodec: A Unified Discrete Audio Tokenizer via Semantic-Acoustic Entanglement2026MiMo-Audio: Audio Language Models are Few-Shot Learners2025AQA-TTRL: Self-Adaptation in Audio Question Answering with Test-Time Reinforcement Learning2025Omni-R1: Do You Really Need Audio to Fine-Tune Your Audio LLM?2025