List your company

Paralinguistic-aware dialogue on ParaS2SBench Synthetic (Human Evaluation)

4.922Average Paralinguistic Awareness ScoreGPT-TTS
2.610083.210293.81054.41071Nov 11, 2025

Evaluation results

MethodLinks
GPT-TTSCategory=Topline2025.114.922————
GPT-TTSCategory=Topline2025.114.8444.7524.8894.9234.813
Kimi-Audio GRPOCategory=Ours2025.114.4074.3164.514.3814.422
Kimi-Audio GRPOCategory=Ours2025.114.303————
Kimi-Audio SFTCategory=Ours2025.114.0894.1924.2233.8124.131
Kimi-Audio SFTCategory=Ours2025.113.996————
GPT-4o voice modeCategory=Closed Source2025.113.564————
Qwen2.5 OmniCategory=Open Source2025.113.538————
Qwen2.5 OmniCategory=Open Source2025.113.4833.3523.9533.4963.131
GPT-4o voice modeCategory=Closed Source2025.113.4573.3753.8333.5423.078
Whisper-GPT-TTSCategory=Baseline2025.113.23————
GLM 4Category=Open Source2025.113.23————
GLM 4Category=Open Source2025.113.1343.0123.5143.2282.781
Whisper-GPT-TTSCategory=Baseline2025.113.1023.2123.0413.0423.112
Kimi-AudioCategory=Open Source2025.112.9243.2782.3823.1212.912
Kimi-AudioCategory=Open Source2025.112.699————