Skip to content
DraconiticResearch
Search Sources API
203,203 studies31 sourcesUpdated 27 min ago
Peer-reviewedCC BY 4.0PLOS One

Near-optimal P300 speller performance using large language models: A multi-model analysis with performance bounds

  • Nithin Parthasarathy1
  • James Soetedjo2
  • Saarang Panchavati3
  • Nitya Parthasarathy4
  • Dongwoo Lee5
  • Corey Arnold6
  • Nader Pouratian
  • William Speier
6 affiliations
  1. 1Department of Computer Science, University of Illinois at Urbana-Champaign, Urbana, Illinois, United States of America
  2. 2Department of Bioengineering, University of Washington, Seattle, Washington, United States of America
  3. 3Department of Radiological Sciences, University of California, Los Angeles, California, United States of America
  4. 4PE Investments, Boston, Massachusetts, United States of America
  5. 5Department of Computer Science, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States of America
  6. 6Department of Neurological Surgery, University of Texas Southwestern Medical Center, Dallas, Texas, United States of America
Open at sourcePDFDOICite

Abstract

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 ~45% (within-subject training) and ~75% (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.

Related

  1. Informed forecasting: Leveraging auxiliary knowledge to boost large language model performance on time series forecasting
    Mohammadmahdi Ghasemloo, Alireza Moradi
    PLOS OnePeer-reviewedCC BY 4.0Full textBioinformatics & computational biologyNeurosciencePLOS
  2. Bridging trust and performance in intelligent systems: Hybrid explainable AI approaches for interpreting large language models
    Arul Selvam P., Tamije Selvy P.
    PLOS OnePeer-reviewedCC BY 4.0Full textNeuroscienceEpidemiology & public healthPLOS
  3. Performance of large language models in the Chinese National Medical Licensing Examinations and Beyond: Scoping review
    Hui Zong, Jiao Wang, Jiaxue Cha +4
    PLOS Digital HealthPeer-reviewedCC BY 4.0Full textNeuroscienceOncologyPLOS
  4. Benchmarking Large Language Models for Assay-Aware Drug-Target Binding Affinity Prediction
    N. Gupta, A. Verma, P. Khandelwal +1
    bioRxivPreprintAll rights reservedFull textBioinformatics & computational biologybioRxiv
  5. BrainAgent: A Large Language Model-Driven Multi-Agent Framework for Autonomous Brain Signal Understanding
    Yangxuan Zhou, Sha Zhao, Jiquan Wang +2
    arXiv (Cornell University)PreprintLicence not statedFull text5 citedNeuroscienceOpenAlex

Cite

BibTeXRIS
BibTeX
@article{parthasarathy2026nearoptimal,
  title = {{Near-optimal P300 speller performance using large language models: A multi-model analysis with performance bounds}},
  author = {Nithin Parthasarathy and James Soetedjo and Saarang Panchavati and Nitya Parthasarathy and Dongwoo Lee and Corey Arnold and Nader Pouratian and William Speier},
  journal = {PLOS One},
  year = {2026},
  month = oct,
  doi = {10.1371/journal.pone.0349281},
  url = {https://doi.org/10.1371/journal.pone.0349281},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - Near-optimal P300 speller performance using large language models: A multi-model analysis with performance bounds
AU  - Nithin Parthasarathy
AU  - James Soetedjo
AU  - Saarang Panchavati
AU  - Nitya Parthasarathy
AU  - Dongwoo Lee
AU  - Corey Arnold
AU  - Nader Pouratian
AU  - William Speier
PY  - 2026
DA  - 2026/10/01
JO  - PLOS One
DO  - 10.1371/journal.pone.0349281
UR  - https://doi.org/10.1371/journal.pone.0349281
AB  - 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 ~45% (within-subject training) and ~75% (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.
SN  - 1932-6203
ER  -