Card: the base reader's plan is PREREGISTRATION.md, written while it trained, before any test split was evaluated

Card: ONNX and in-browser use

Tokenizer files that load in transformers 4 and 5 and Transformers.js; card update

ONNX weights (fp32, and int8 for browsers) for onnxruntime and Transformers.js

EASE-Delta Reader: transformers-native claim-vs-evidence classifier

initial commit