2026-10-14T03:46:25Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34049579 2026-10-01T17:51:15Z category_146 category_15 category_734 category_931 portal_5 item_type_6 month_year_10_2026

<p>Kneser–Ney smoothing formulation.</p> Nithin Parthasarathy (25158080) James Soetedjo (25158083) Saarang Panchavati (8205312) Nitya Parthasarathy (25158086) Dongwoo Lee (1511272) Corey Arnold (475067) Nader Pouratian (475068) William Speier (475066) Space Science Neuroscience Biological Sciences not elsewhere classified Information Systems not elsewhere classified using extensive simulations requiring assistive technologies remains unclear whether progressive neurodegenerative disease neural signal decoding indicating diminishing returns establish upper bounds despite architectural differences amyotrophic lateral sclerosis assisted p300 spellers severely impairs communication translating eeg responses reduce calibration requirements unified decoding framework p300 speller brain considered decoding framework systematically characterized within slow typing speed operate within 5 div >< p theoretical performance bound large language models improvements generalize across improving bci communication specific performance bounds subject classifier training p300 speller typing speed language models enables communication theoretical perspective specific calibration eeg data performing models language modeling practical framework achievable performance subject training work provides wide range significantly improve several high results suggest restore interaction recent work primary bottleneck predicting words practical adoption <div><p>Amyotrophic lateral sclerosis (ALS), a progressive neurodegenerative disease, severely impairs communication, requiring assistive technologies that restore interaction. The P300 speller brain-computer interface (BCI) enables communication by translating EEG responses into text; however, its practical adoption is limited by slow typing speed and the need for subject-specific calibration. Recent work has demonstrated that large language models (LLMs), such as GPT-2, can significantly improve the performance of the P300 speller by predicting words. However, it remains unclear whether these gains are model-specific or represent a broader trend across language models. Furthermore, the framework-specific performance bounds of LLM-assisted P300 spellers have not been systematically characterized within a unified decoding framework. In this study, we address these gaps through a systematic multi-model theoretical analysis framework. We evaluate a wide range of language models and introduce an idealized LLM to establish upper bounds on achievable performance. In addition, we incorporate cross-subject classifier training to reduce calibration requirements and assess generalization across subjects. Using extensive simulations on EEG data from 78 subjects, we demonstrate that the evaluated models consistently achieve substantial improvements in typing speed, with gains of up to (within-subject training) and (across-subject training) over conventional approaches. More importantly, we show that several high-performing models, despite architectural differences, operate within 5% of the theoretical performance bound, indicating diminishing returns from further model scaling. These improvements generalize across both within-subject and across-subject classifiers. Our results suggest that LLM-assisted P300 spellers are approaching oracle-based upper bounds within the considered decoding framework, shifting the primary bottleneck from language modeling to neural signal decoding. This work provides both a practical framework for improving BCI communication and a theoretical perspective on its achievable limits.</p></div> 2026-10-01T17:51:12Z Text Journal contribution 10.1371/journal.pone.0349281.s001 https://figshare.com/articles/journal_contribution/_p_Kneser_Ney_smoothing_formulation_p_/34049579 CC BY 4.0