List your company
ResearchFormerly WizWand

Multi-talker Automatic Speech Recognition on Libri2Mix Noisy (dev)

8.4WERID-20.D
7.9610.9313.916.87Mar 11, 2026Mar 13, 2026Mar 16, 2026Mar 19, 2026Mar 22, 2026Mar 25, 2026Mar 28, 2026
Updated 7mo ago

Evaluation results

MethodLinks
ID-20.DLLM usage=With LLMs2026.038.4
ID-30.DLLM usage=With LLMs2026.038.6
SOP-Llama-3BLLM Usage=With LLMs2026.0310.5
SOP-Llama-3BLLM usage=With LLMs, Backbone=Llama-3B, Protocol=SOP2026.0310.5
ID-8LLM Usage=With LLMs, Method description=Proposed Encoder-only Model with distillation2026.0310.7
SOT-Llama-3BLLM Usage=With LLMs2026.0311.2
SOT-Llama-3BLLM usage=With LLMs, Backbone=Llama-3B, Protocol=SOT2026.0311.2
SOP-Llama-1BLLM Usage=With LLMs2026.0311.8
SOP-Llama-1BLLM usage=With LLMs, Backbone=Llama-1B, Protocol=SOP2026.0311.8
SOT-Llama-1BLLM Usage=With LLMs2026.0312.4
SOT-Llama-1BLLM usage=With LLMs, Backbone=Llama-1B, Protocol=SOT2026.0312.4
w/o decoder: CTCLLM Usage=Without LLMs, Encoder=SSL2026.0316.8
GEncSep w/o decoderLLM usage=Without LLMs, SSL for speech encoder=true, Alignment protocol=serialized CTC2026.0316.8
GEncSepLLM Usage=Without LLMs, Encoder=SSL2026.0317.2
GEncSepLLM usage=Without LLMs, SSL for speech encoder=true2026.0317.2
Training from ScratchLLM Usage=Without LLMs, Encoder=SSL2026.0319.4
Training from ScratchLLM usage=Without LLMs, SSL for speech encoder=true2026.0319.4