Our latest model training achieves exceptional accuracy with under 4.5% validation error rate, powered by our expanded dataset of 34+ million eye fixations from 1.77+ million eye-tracking tests.
We're thrilled to announce a major breakthrough in our AI attention prediction technology. Our latest model training has achieved unprecedented accuracy levels, with validation error rates below 4.5% across all attention components.
🎯 Record-Breaking Performance Metrics
The latest training results from July 2025 demonstrate exceptional improvement:
- Overall Training Error: 3.85% MAE
- Overall Validation Error: 4.52% MAE
- Hold Component: 0.043 MAE (sustained attention prediction)
- Speed Component: 0.052 MAE (first fixation timing)
- Reach Component: 0.040 MAE (participant coverage)
📊 Comprehensive Training Dataset
These remarkable accuracy improvements are powered by our comprehensive training dataset:
- 33,998,738 Eye Fixations: Massive dataset for robust prediction accuracy
- 1,774,117 Eye-Tracking Tests: Extensive real-world validation data
- 12,053 Distinct Items: Comprehensive coverage across industries and design types
- Diverse Viewing Patterns: Unprecedented diversity in attention behaviors
🔬 Industry-Leading Accuracy
With error rates consistently below 5.5% across all attention components, our model now delivers:
- Best-in-class sustained attention prediction (Hold component)
- Exceptional immediate attention capture accuracy (Speed component)
- Outstanding participant coverage forecasting (Reach component)
💼 Real-World Impact for Businesses
This level of accuracy translates to more reliable design optimization decisions, reduced testing costs, and higher confidence in pre-launch design validation. Businesses can now trust our predictions for critical design decisions with research-grade confidence.
Experience the power of our enhanced accuracy for your design projects. Book a demo to see how these improvements can optimize your specific use cases.