Optimize and train foundation models using IBM's FMS

Measure BERT model performance using WASM and WebGPU

Evaluate LLM over-refusal rates with OR-Bench