Continuous Learning Pipeline

Supervised Fine-Tuning & DPO Dataset Synthesis

Active Model Registry

llama-3-forex-v1.2ACTIVE

Base: Llama-3-8B-Instruct

Brier Score0.08

llama-3-forex-v1.1

Base: Llama-3-8B-Instruct

Brier Score0.12

Synthesized Datasets (.jsonl)

core-macro-trends-v1

1840 Pairs2026-03-23
s3://datasets/macro.jsonl

risk-nfp-scenarios-v2

450 Pairs2026-03-20
s3://datasets/risk_nfp.jsonl

technical-breakouts-v1

3210 Pairs2026-03-15
s3://datasets/tech.jsonl