Awesome-Speech-Language-Model
A curated list of papers, code, and resources for speech language models and end-to-end speech dialogue systems, covering understanding, benchmarks, and full-duplex modeling.
Awesome-Speech-Language-Model organizes research on universal speech, audio, and music understanding and on end-to-end speech dialogue systems into model and benchmark sections for each area, plus a section on full-duplex modeling and a survey section. Listed understanding models include LTU, SALMONN, Qwen-Audio, Audio Flamingo, Qwen2-Audio, WavLLM, DiVA, and SenseVoice, with benchmarks such as Dynamic-SUPERB, AIR-Bench, SD-Eval, AudioBench, SALMon, and MMAU/MMAR.
Listed end-to-end speech dialogue systems include SpeechGPT, an audio-native voice mode from a major model provider, PSLM, VITA, Mini-Omni, and LLaMA-Omni, each tagged with its publication venue such as EMNLP, ICLR, ICML, ACL, or arXiv.