FinBERT base + cross-bank pretraining substrate (samchain/BIS_speeches_97_23_MLM plus the gtfintechlab ECB / BoJ / BoE / BoC / RBA monetary-policy corpora reformatted as NSP pairs).

Training command

python -m app.data.continued_pretraining --substrate xbank_dapt --seed 11

Attribution

Base model: ProsusAI/finbert (Araci, 2019).
BIS speeches NSP pairs: samchain/BIS_speeches_97_23_MLM, redistributed from the Bank for International Settlements speeches archive.
Cross-bank monetary-policy corpora: gtfintechlab multi-axis sentence datasets for the ECB, Bank of Japan, Bank of England, Bank of Canada, and Reserve Bank of Australia.