PhaseCoder: Microphone Geometry-Agnostic Spatial Audio Understanding for Multimodal LLMs
This addresses the limitation of current audio processing in AI by enabling deployment across diverse devices, though it is incremental in combining existing transformer methods with spatial audio.
The paper tackled the problem of spatial audio understanding in multimodal LLMs by introducing PhaseCoder, a geometry-agnostic encoder that achieved state-of-the-art results on localization benchmarks and enabled LLMs to perform complex spatial reasoning tasks.
Current multimodal LLMs process audio as a mono stream, ignoring the rich spatial information essential for embodied AI. Existing spatial audio models, conversely, are constrained to fixed microphone geometries, preventing deployment across diverse devices. We present PhaseCoder, a transformer-only spatial audio encoder that is agnostic to microphone geometry. PhaseCoder takes raw multichannel audio and microphone coordinates as inputs to perform localization and produces robust spatial embeddings. We demonstrate that Gemma 3n LLM can be fine-tuned to reason over "Spatial Audio Tokens" produced by PhaseCoder. We show our encoder achieves state-of-the-art results on microphone-invariant localization benchmarks and, for the first time, enables an LLM to perform complex spatial reasoning and targeted transcription tasks from an arbitrary microphone array.