Zero-Mem / Zero-mem

Why you might care

Targets researchers reproducing the Zero-Mem paper: memory construction, routing, retrieval and calibration use no LLM tokens, with only encoder, NER, and final-QA reader calls costing compute. Model weights and datasets must be obtained separately, and the README is thin on usage details.