Skip to content
Log in
Sponsored Artificial Intelligence: A Modern Approach, Global Edition Artificial Intelligence: A Modern Approach, Global Edition

Author

Meixuan Ren

View on Semantic Scholar

2 papers indexed here

We haven’t gathered this author’s papers yet. Follow them and we’ll fetch their work.

Not the right person? Other researchers publish under this name.

Open access Jul 2026

Electroclinical features and dynamic network connectivity of orbitofrontal epilepsy.

OBJECTIVE Orbitofrontal epilepsy (OFE) is rare and underrecognized. This study aims to characterize electroclinical features and dynamic network connectivity of OFE and subtypes. METHODS We retrospectively analyzed 17 patients with OFE who underwent stereoelectroencephalographic (SEEG) electrode implantation and epilepsy surgery at our center. Thirteen patients with available ictal SEEG recordings were further categorized into lateral (LOFE, n = 6), medial (MOFE, n = 5), and posterior (POFE, n = 2) subtypes. Clinical semiology, scalp electroencephalographic (EEG), neuroimaging, and SEEG data were analyzed. Directed functional connectivity was evaluated in 22 seizures from 13 patients using pairwise nonlinear regression across ictal stages. RESULTS The mean seizure duration was 68.4 s. Semiology was heterogeneous, with impaired consciousness being the most common manifestation (47.1%), frequently evolving into elementary motor phenomena (62.5%) or automatisms (37.5%). Hyperkinetic behavior (23.5%) and focal to bilateral tonic-clonic seizures (29.4%) were also observed. This heterogeneity may relate to propagation to frontal, cingulate, insular, and temporal regions. In 13 patients with available ictal SEEG recordings, EEG seizure activity preceded clinical onset, with a mean electroclinical delay of 28.9 s. LOFE most often showed nonspecific auras and early motor manifestations; MOFE was associated with autonomic auras and automatisms; and POFE frequently presented with fear auras. Functional connectivity increased from early to late ictal periods and remained elevated postictally. Connectivity differences between LOFE and MOFE were confined to the seizure onset and early propagation phases, with LOFE showing stronger lateral frontal interactions and MOFE stronger medial frontal-cingulate/insula interactions. SEEG resolved frequent mislocalizations by scalp EEG/imaging. Most patients achieved favorable outcomes (Engel class I: 84.6%, mean follow-up = 52.1 months, 11/13), and focal cortical dysplasia was the most common pathology (76.9%, 10/13). SIGNIFICANCE OFE subtypes tend to show different semiological features and preferential connectivity patterns, potentially related to differential engagement of frontal-limbic networks. Targeted SEEG is associated with accurate localization and generally favorable outcomes in this cohort.

Meixuan Ren, Cuiping Xu, N. Pan et al. · 0 citations
Review Open access Aug 2026

A Dual-Teacher Distilled MoE Agent for Complex Industrial Document Analysis

Consistency verification of drilling reports is critical for engineering quality control because a single data item may be distributed across reports with different formats, units, and page structures. Existing retrieval-augmented generation methods remain sensitive to retrieval and parsing errors in such documents, whereas ultra-large models impose substantial local computing and memory costs. This study proposes a lightweight tool-augmented framework based on dual-teacher distillation and sparse mixture-of-experts (MoE) modeling. Qwen3-235B-A22B serves as the primary teacher and Qwen3-30B-A3B as the assistant teacher. Their tool-use and task-planning capabilities are transferred to a sparse MoE student upgraded from a Qwen3-1.7B dense backbone through trajectory pruning, sample decomposition, and token-level Kullback–Leibler (KL) distillation. The student adopts an eight-expert Top-2 routing architecture. Experiments on 1000 drilling reports containing 30,127 verification instances show an F1 score of 58.0 ± 0.5%, with file-level, location-level, and exact-match accuracies of 66.5%, 55.2%, and 45.0%, respectively. The model contains 9.1B total parameters and 2.8B activated parameters, and reaches a latency of 12.1 ms per forward pass and a memory footprint of 18.4 GB under bfloat16 (BF16) precision. The reported F1 score characterizes the end-to-end verification task rather than an autonomous safety decision capability. The framework is intended to support evidence localization, anomaly prioritization, and expert review in local deployment settings.

Enli Zhang, Qiang Kang, Ruilong Tang et al. · 0 citations