List your company
ResearchFormerly WizWand

Speech-to-Speech Question Answering on Llama Questions

79.33AccuracyFreezeEmpath
57.843663.42186974.5782Mar 10, 2026Mar 16, 2026Mar 23, 2026Mar 30, 2026Apr 6, 2026Apr 13, 2026Apr 20, 2026
Updated 6mo ago

Evaluation results

MethodLinks
FreezeEmpathModality=S2T2026.0479.33
Kimi-AudioModality=S2T2026.0479
VocalNetModel Parameters=8B2026.0378.6
Fun-Audio-Chat-8BModality=S2T2026.0477.76
Qwen2.5-Omni-7BModel Parameters=7B2026.0377.33
AudioKVModel=Qwen2.5-Omni-7B, KV cache budget=40%, max_gen_tokens=64, Judge=Gemini 2.5 Flash2026.0477.3
PyramidKVModel=Qwen2.5-Omni-7B, KV cache budget=40%, max_gen_tokens=64, Judge=Gemini 2.5 Flash2026.0476.7
SnapKVModel=Qwen2.5-Omni-7B, KV cache budget=40%, max_gen_tokens=64, Judge=Gemini 2.5 Flash2026.0476
AdaKVModel=Qwen2.5-Omni-7B, KV cache budget=40%, max_gen_tokens=64, Judge=Gemini 2.5 Flash2026.0476
AudioKVModel=Qwen2.5-Omni-3B, KV cache budget=40%, max_gen_tokens=64, Judge=Gemini 2.5 Flash2026.0475
AdaKVModel=Qwen2.5-Omni-3B, KV cache budget=40%, max_gen_tokens=64, Judge=Gemini 2.5 Flash2026.0474.7
PyramidKVModel=Qwen2.5-Omni-3B, KV cache budget=40%, max_gen_tokens=64, Judge=Gemini 2.5 Flash2026.0474.7
FreezeEmpathModality=S2S2026.0474.67
LLaMA-Omni2-7BModality=S2T2026.0473.67
SPEECH-OMNI-LITEModel Parameters=8B2026.0373.33
LLaMA-Omni2Model Parameters=7B2026.0372
SnapKVModel=Qwen2.5-Omni-3B, KV cache budget=40%, max_gen_tokens=64, Judge=Gemini 2.5 Flash2026.0472
Fun-Audio-Chat-8BModality=S2S2026.0472
OpenS2SModality=S2T2026.0470.67
Step-Audio2-MiniModality=S2T2026.0466.67
LLaMA-Omni2-7BModality=S2S2026.0466.67
Baichuan-Omni-1.5Model Parameters=7B2026.0365.67
Kimi-AudioModality=S2S2026.0464.67
Step-Audio2-MiniModality=S2S2026.0464.33
GLM-4-VoiceModel Parameters=9B2026.0363.67
OpenS2SModality=S2S2026.0459
Freeze-OmniModel Parameters=7B2026.0358.67