arXiv q-bio.NC authors propose REALM for LFP-based brain-computer decoding

arXiv q-bio.NC authors propose REALM for LFP-based brain-computer decoding

Peicheng Wu and coauthors propose REALM, a method for decoding behavior from local field potentials (LFPs), electrical signals recorded from brain tissue. They argue that LFPs may suit wireless brain-computer interfaces because they require less energy and bandwidth than spike recordings, but LFP decoders have often been less accurate or unsuitable for real-time use. REALM uses a bidirectional teacher model trained across recording sessions, then transfers its representations to a smaller causal student that can process signals without relying on future data. The authors report that REALM had the highest mean accuracy among the compared decoders in both label-free and fine-tuned tests, while using fewer parameters than CrossModalDistill’s published student and less pretraining time. They say the approach used LFPs alone throughout. Escape Meridian interpretation: These results suggest a possible route to real-time LFP decoding with lower resource demands, but the abstract does not establish performance in deployed or wireless implants. How well does REALM perform in real-world, long-term use with wireless or implantable brain-computer interfaces?

Publication: 38aefb1e-7671-454b-85c8-87e4dbf35f67

attribution — pass: The reading attributes REALM and its results to Peicheng Wu and coauthors and keeps performance claims as author reports. The title's 'arXiv q-bio.NC authors' follows the publisher metadata. The paper's primary category is cs.LG, with a q-bio.NC cross-listing, but this does not misattribute the work.

Authors:Peicheng Wu, Zhenyu Bu, Runze Ma, Lin Du

summary — pass: The reading explains LFPs as electrical signals recorded from brain tissue and causal processing as not using future data. It states the motivation (power and bandwidth in wireless iBCIs) and the teacher-to-student transfer. Some terms, such as 'label-free' and 'parameters', are not explained, but they do not block a general reader from understanding the central idea. The question follows from what the abstract does not establish.

REALM transfers non-causal representational knowledge from a pretrained, multi-session bidirectional LFP teacher to a causal student model.

date — pass: originalPublished 2026-05-14 matches the v1 submission date. The later revisions (through v4 on 27 Oct 2026) and the feed date are correctly not used.

[Submitted on 14 May 2026 (v1), last revised 27 Oct 2026 (this version, v4)]

support — pass: The claims about highest mean accuracy, fewer parameters, shorter pretraining time and LFP-only use match the abstract. The Escape Meridian interpretation is labeled and hedged. It correctly notes that the abstract does not demonstrate performance in deployed or wireless implants.

with less than half the parameters of CrossModalDistill's published student and one-tenth of its pretraining time

media — pass: The selected media is the verified direct arXiv abstract page for this paper, which is the most direct way to access the source.