implementation detail · filed under model architecture
Four-frame audio patches with within-patch bidirectional self-attention
Consecutive groups of four frames form audio patches processed with bidirectional self-attention confined to each patch.
- source
- 1
- models
- 2
- lab adopt it
- 1
- strongest
- core
How sources treat it
One count per evidence span, weakest treatment to strongest.
core 1
Documented in
Evidence
1 span quoted from the sources, strongest treatment first.
Every four consecutive frames are grouped into an audio patch and processed by a Transformer with bidirectional self-attention confined to that patch.
coremodel architecturein MiMo-V2.6 Audio Patch EncoderXiaomi
Filed alongside
Other methods under model architecture :: multimodal architecture.
Encoder-free multimodal architectureEarly fusion multimodal trainingDiscrete token encoding for audioHierarchical patch encoderMixed-modality trainingConfigurable visual token budgetJoint multimodal decoding in a shared hidden spaceNative multimodal understanding3×3 pixel-unshuffle downsamplingJoint vision-language pre-trainingRaw audio projection into the LLM embedding spaceVision Transformer (ViT)2×2 pixel-shuffle downsamplingAudio encoder initialized from MiMo-AudioAudio Transformer (AuT)Causal streaming ConvNet codec decoderData-parallel-first multimodal encodingDecoupled Encoder Process (DEP)Dedicated multimodal encodersDisaggregated encoder trainingDiscarding the LLM after vision-encoder trainingDual-format coordinate supervisionExplicit text-string timestampsFrom-scratch vision-encoder training with next-token prediction