New MDAR Benchmark Challenges AI Models on Dynamic Audio Reasoning

The new MDAR benchmark offers 3,000 audio Q&A pairs for dynamic reasoning; Qwen2.5‑Omni hit 76.67% accuracy on single‑choice while GPT‑4o Audio reached 68.47%. Read more: getnews.me/new-mdar-benchmark-chall... #mdar #audioreasoning #ai