2 papers
cs.LG2026
Veriphi: Attack-Guided Neural Network Verification with Dataset-Dependent Training Methods
Pratik Deshmukh, Kartik Arya, Vasili Savin
We present Veriphi, a GPU-accelerated neural network verification system that combines fast adversarial attacks with formal bound certification using alpha,beta-CROWN methods. Thro…
cs.LG2026
On Semantic Loss Fine-Tuning Approach for Preventing Model Collapse in Causal Reasoning
Pratik Deshmukh, Atirek Gupta
Standard fine-tuning of transformer models on causal reasoning tasks leads to catastrophic model collapse, where models learn trivial solutions such as always predicting "Yes" or "…