YirongSun/LLaSO-Align

LLaSO是一个用于大规模语音语言建模的全开放端到端框架，包含三个资源：LLaSO-Align语音文本对齐语料库、LLaSO-Instruct多任务指令微调数据集和LLaSO-Eval标准化评估的基准。该数据集旨在统一数据、评估和建模，支持所有主要的文本和音频组合，涵盖语言、语义和副语言任务。LLaSO is a fully open end-to-end stack for large-scale speech-language modeling, including three resources: LLaSO-Align speech-text alignment corpus, LLaSO-Instruct multi-task instruction tuning dataset, and LLaSO-Eval benchmark for standardized evaluation. The dataset aims to unify data, evaluation, and modeling, supporting all major text-audio combinations across linguistic, semantic, and paralinguistic tasks.