Releases·Model Release
SNU researcher releases Sori-1B, a 1B audio-language model trained from scratch without text-only pretraining
Sori-1B is a 1B-parameter audio-grounded language model released on Hugging Face by a single SNU researcher. Its decoder is trained entirely from scratch on audio-paired text, with no text-only pretraining or pretrained-LM initialization, so answers ground in audio rather than text-only priors. The release includes evidence that typical audio models lean heavily on those priors.
6 engagement1 sourceMon, Aug 31, 2026, 02:48 AM
AI Overview
- Reuses NVIDIA's frozen Audio Flamingo Next audio encoder while training the decoder from scratch on audio-paired text
- Shows Audio Flamingo 3 retains ~74% of its above-chance MMAU margin even when audio is replaced with silence, indicating text-prior reliance
- Built by a single Seoul National University researcher; released via Hugging Face (snkii/Sori-1B)
Related
Releases·Model Release
Mizar: A 159M-Parameter Audio-Language Model for Audio Understanding
Researchers introduce Mizar, a 159.3M-parameter audio-language model built from a CED-Small audio encoder connected to SmolLM2-135M via a frequency-merging mapper. A three-stage training recipe on ReasonAQA, AudioMCQ, and AVQA data pushes it past all prior ALMs under 200M parameters on MMAU, MMAR, and ADQA-clean. Code and checkpoints are public, and the model runs on-device with roughly one-second CPU latency.
Researchers introduce Mizar, a 159.3M-parameter audio-language model built from a CED-Small audio encoder connected to SmolLM2-135M via a frequency-merging mapper. A three-stage training recipe on ReasonAQA, AudioMCQ, and AVQA data pushes it past all prior ALMs under 200M parameters on MMAU, MMAR, and ADQA-clean. Code and checkpoints are public, and the model runs on-device with roughly one-second CPU latency.
0 engagement1 sourcearxivWed, Sep 23, 2026, 04:21 PM
Releases·Model Release
Samsone: A Family of Open Small Audio Language Models for On-Device Inference
A new arXiv paper introduces Samsone, a family of open Small Audio Language Models (SALMs) for edge computing, with sizes of 134M, 99M, and 356M parameters. Samsone-134M sets a state-of-the-art for its size class while remaining competitive with far larger models. The team releases training code, weights, mobile-optimized checkpoints, and an open-source Android app for real-time on-device inference.
A new arXiv paper introduces Samsone, a family of open Small Audio Language Models (SALMs) for edge computing, with sizes of 134M, 99M, and 356M parameters. Samsone-134M sets a state-of-the-art for its size class while remaining competitive with far larger models. The team releases training code, weights, mobile-optimized checkpoints, and an open-source Android app for real-time on-device inference.
0 engagement1 sourcearxivFri, Sep 18, 2026, 12:03 PM
Releases·Model Release
ai-sage releases GigaChat Audio 10B, an audio-native LLM with temporal grounding
ai-sage released GigaChat Audio 10B, an audio-native LLM built on GigaChat 3.1 Lightning. It uses a Conformer speech encoder and modality adapter to feed audio embeddings into a Mixture-of-Experts decoder, enabling audio question answering, classification, temporal grounding, tool use, and text-only tasks.
ai-sage released GigaChat Audio 10B, an audio-native LLM built on GigaChat 3.1 Lightning. It uses a Conformer speech encoder and modality adapter to feed audio embeddings into a Mixture-of-Experts decoder, enabling audio question answering, classification, temporal grounding, tool use, and text-only tasks.
87 engagement1 sourceredditSun, Jul 26, 2026, 09:59 AM
Releases·Model Release
FireRedTeam releases FireRedAudio, a 9B-parameter audio language model
FireRedTeam released FireRedAudio, a general-purpose audio language model built on a shared 9B-parameter LLM with decoupled continuous representations for understanding and generation. The model is available on HuggingFace with code and demos, enabling tasks like speech recognition, synthesis, and audio understanding in one framework.
FireRedTeam released FireRedAudio, a general-purpose audio language model built on a shared 9B-parameter LLM with decoupled continuous representations for understanding and generation. The model is available on HuggingFace with code and demos, enabling tasks like speech recognition, synthesis, and audio understanding in one framework.
70 engagement1 sourceredditFri, Aug 21, 2026, 04:05 PM
Developments·Paper
Fusion Embedding family adds audio to vision-language retrieval
A new paper introduces Fusion Embedding, a family of models that extend frozen vision-language embedding backbones to handle audio, enabling a single embedding space for text, image, video, and audio. The first generation (fusion-embedding-1) trains only a 16.4M-parameter connector between a frozen audio tower and the frozen base, achieving competitive audio-text retrieval without updating the base model.
A new paper introduces Fusion Embedding, a family of models that extend frozen vision-language embedding backbones to handle audio, enabling a single embedding space for text, image, video, and audio. The first generation (fusion-embedding-1) trains only a 16.4M-parameter connector between a frozen audio tower and the frozen base, achieving competitive audio-text retrieval without updating the base model.
0 engagement1 sourcearxivTue, Jul 21, 2026, 03:25 AM