Text Classification
setfit
Joblib
Safetensors
English
modernbert
xgboost
intent-classification
adversarial-defense
Instructions to use atekrugis/intent-router-v11-modernbert with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- setfit
How to use atekrugis/intent-router-v11-modernbert with setfit:
from setfit import SetFitModel model = SetFitModel.from_pretrained("atekrugis/intent-router-v11-modernbert") - Notebooks
- Google Colab
- Kaggle
Phase 11 Champion: Intent-Router-v11-ModernBERT
This is a high-precision Intent Router designed to handle Adversarial Logic Traps. It uses a two-stage hybrid architecture to achieve robust performance.
Architecture: The Dual-Brain
- Backbone: answerdotai/ModernBERT-base fine-tuned via SetFit on Diamond v11.
- Classification Head: Surgically calibrated XGBoost Forest (500 estimators).
Performance (Phase 11 Gauntlet)
- Overall Accuracy: 95.00%
- Logic Handling: Successfully solves Bat and Ball, Sally's Brothers, and 3D Rotation traps.
- Latency: Optimized for A100 throughput using ModernBERT native Flash Attention.
Quick Start (Inference)
Load the ModernBERT backbone and the calibrated XGBoost head (v11_forest_head.joblib).
Example usage: model = SetFitModel.from_pretrained('atekrugis/intent-router-v11-modernbert') forest_clf = joblib.load('v11_forest_head.joblib')
- Downloads last month
- 2