LibriSpeech-Real-QA
:
A Benchmark for Streaming Question Answering
Qwen2.5-Omni + SFT
· 160 ms chunk streaming demo
test-clean
chunk
0.16 s
live model output vs. static transcript
real model output