Tag
This paper proposes a unified optimization framework to explain misclassifications and assess classifier robustness by sparse, interpretable instance alterations and a Tolerance Region Confusion Matrix.
A user reports that Fable 5's new classifiers misrouted 75% of a coding session to Opus, flagging routine coding as a cybersecurity risk and causing unexpectedly high costs.
Proposes a tree-based ensemble classifier called KFDA Forest that applies kernel Fisher discriminant analysis for rotation, using bootstrap and random variable subsets to promote diversity, achieving improved classification accuracy.
The author replaced an LLM classifier with a simple set of if-statements and found the client preferred the rule-based approach, highlighting the value of simplicity over complex AI.
The article introduces a technique that extracts hidden states from an LLM at the last prompt token to perform classification without text generation, using a small MLP to read the model's internal decision, enabling fast and cheap zero-shot classifiers.
This paper introduces HRVConformer, a hybrid Convolution-Transformer architecture for classifying neonatal hypoxic-ischemic encephalopathy directly from raw heart rate signals, achieving an AUC of 83.23% and outperforming baseline models like ResNet50 and Transformer.
Trained a prompt injection classifier using ml-intern and DeepSeek V4 Flash, achieving 99% F1 with DistilBERT, optimized to ONNX int8 (~65MB) and deployable in the browser via Transformers.js v3.