Reduces the RAM requirement for speech neuroprosthesis CTC decoding from 320 GB to 10 GB without sacrificing accuracy.
March 17, 2026
Original Paper
LightBeam: An Accurate and Memory-Efficient CTC Decoder for Speech Neuroprostheses
arXiv · 2603.14002
The Takeaway
By replacing large WFST-based N-gram decoders with a delayed-fusion LLM approach, this tool makes high-performance brain-to-text interfaces deployable on standard consumer hardware. It enables real-time, high-accuracy communication for medical applications where massive server clusters were previously mandatory.
From the abstract
A promising pathway for restoring communication in patients with dysarthria and anarthria is speech neuroprostheses, which directly decode speech from cortical neural activity. Two benchmarks, Brain-to-Text '24 and '25, released intracranial recordings from patients with dysarthria along with a baseline algorithm trained with Connectionist Temporal Classification (CTC). Despite significant innovation on these benchmarks, all leading published prior work relies on a WFST-based CTC decoder that re