In May 2025, one paper from the Natural Language Processing Group was accepted by the ACL 2025 Main Conference. ACL, or The Association for Computational Linguistics, is the most influential academic organization in the field of international computational linguistics. The ACL annual conference is the most important international event in computational linguistics and is the only CCF-recommended A-class conference in this area.。
The accepted paper is briefly introduced as follows:
1. LLaMA-Omni 2: LLM-based Real-time Spoken Chatbot with Autoregressive Streaming Speech Synthesis (Qingkai Fang, Yan Zhou, Shoutao Guo, Shaolei Zhang, Yang Feng)
Accepted by ACL 2025 Main Conference
Abstract: Real-time, intelligent, and natural speech interaction is a crucial part of next-generation human-computer interaction. Recent advances in technology have shown the great potential of large language models (LLMs) in building intelligent spoken chatbots. This paper introduces LLaMA-Omni 2, a series of speech language models (SpeechLMs) with parameter sizes ranging from 0.5 billion to 14 billion, capable of high-quality real-time speech interaction. LLaMA-Omni 2 is built upon the Qwen2.5 model series and integrates a speech encoder with an autoregressive streaming speech decoder. Despite being trained on only 200,000 multi-turn spoken dialogue samples, LLaMA-Omni 2 demonstrates strong performance across multiple benchmarks for spoken question answering and voice command following, outperforming previous leading speech language models such as GLM-4-Voice, which was trained on millions of hours of speech data.
