MambaGaze: Bidirectional Mamba with Explicit Missing Data Modeling for Cognitive Load Assessment from Eye-Gaze Tracking Data

2026-05-21 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Human-Computer Interaction · Depth: Expert, quick

Summary

MambaGaze is a novel framework designed for real-time cognitive load assessment using eye-gaze tracking data, addressing challenges of frequent data missingness and efficient long-range temporal dependency modeling. It integrates XMD encoding, which augments raw features with observation masks and time-deltas to explicitly model data uncertainty, with a bidirectional Mamba-2 architecture for capturing temporal dependencies with linear computational complexity. Evaluated using a leave-one-subject-out approach on the CLARE and CL-Drive datasets, MambaGaze achieved 76.8% and 73.1% accuracy, respectively, significantly outperforming CNN, Transformer, ResNet, and VGG baselines by 4-12 percentage points. Furthermore, edge deployment benchmarks on NVIDIA Jetson platforms demonstrated real-time inference capabilities at 43-68 FPS with power consumption below 7.5W, confirming its feasibility for wearable cognitive load monitoring in applications like driver vigilance or flight deck assistance.

Key takeaway

For Machine Learning Engineers developing real-time human-AI interaction systems, MambaGaze provides a robust approach for cognitive load assessment from noisy eye-tracking data. Its explicit modeling of data missingness via XMD encoding and efficient bidirectional Mamba-2 architecture addresses key challenges. You should consider MambaGaze's methodology for applications like driver vigilance or flight deck assistance, especially when deploying on edge platforms like NVIDIA Jetson, given its proven accuracy and low power consumption.

Key insights

MambaGaze uses XMD encoding and bidirectional Mamba-2 to assess cognitive load from eye-gaze data, handling missingness and long-range dependencies efficiently.

Principles

• Explicitly model data uncertainty.
• Capture long-range temporal dependencies.
• Achieve linear computational complexity.

Method

MambaGaze employs XMD encoding to augment features with observation masks and time-deltas, then uses bidirectional Mamba-2 to process these for cognitive load assessment.

In practice

• Real-time driver vigilance monitoring.
• Automated flight deck assistance.
• Wearable cognitive load monitoring.

Topics

• Cognitive Load Assessment
• Eye-Gaze Tracking
• MambaGaze
• Bidirectional Mamba
• Missing Data Modeling
• Edge AI

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, Robotics Engineer

Related on AIssential

• Coarse-to-fine Hierarchical Architecture with Sequential Mamba for Brain Reconstruction — CHASMBrain's dual-stream Mamba and coarse-to-fine fMRI encoding architecture achieves high predictive accuracy and reveals visual cortex functional specialization.
• Spatial-Aware Reduction Framework: Towards Efficient and Faithful Visual State Space Models — Token reduction in visual State Space Models requires spatial awareness to prevent performance collapse.
• Hip joint image quality screening based on the Diffusion Mamba model — A new dataset and quality framework enhance AI diagnostic development for rare orthopedic conditions.
• IEEE Transactions on Neural Networks and Learning Systems, Volume 37, Issue 2, February 2026 — This issue advances neural network theory and application across diverse domains, from medical imaging to robotics and trustworthy AI.
• ConvVitMamba: Efficient Multiscale Convolution, Transformer, and Mamba-Based Sequence modelling for Hyperspectral Image Classification — ConvVitMamba efficiently classifies hyperspectral images by combining multiscale CNNs, Vision Transformers, and Mamba-inspired sequence mixing.
• GEM: Generating LiDAR World Model via Deformable Mamba — GEM is a LiDAR world model using deformable Mamba to improve autonomous driving simulation by addressing point cloud disorder and object distinction.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.