Datasets & Models
Everything pushed to the KUTech Hugging Face org — the shared source of truth for the team.
KUTech/trading-llm-data
20k finance-reasoning SFT conversations (ODA-Fin-SFT-318k derived, English-filtered).
KUTech/trading-llm-market-data
Raw 1-minute OHLCV forex data, 7 pairs, 2015-2025, validated for gaps/duplicates.
KUTech/trading-llm-market-reasoning
Market-grounded reasoning conversations generated from real price action, no look-ahead leakage.
KUTech/trading-llm-training
Training, dataset-build, and validation scripts for the fine-tuning pipeline.
KUTech/trading-llm
LoRA adapter checkpoints, including the Colab Qwen2.5-0.5B fine-tune run.
KUTech/JOLASH
Project roadmap model card: reasoning model -> forecasting -> probability modeling -> risk management -> API deployment.
KUTech/trading-llm-validation-results
Real full-validation backtest results, pushed by backtest_signal_model.py --hf-results-repo. This dashboard's Backtest Results page fetches it live at build time.
KUTech/trading-llm-agent-state
Research Agent's prior-attempts state (pull-before/push-after), so it never re-proposes a (pair, timeframe, horizon) combo already tried.
KUTech/jolash-paper-trades
Live USDCHF paper-trading decision log (private) -- mirrored hourly from paper_trade_usdchf.py's launchd job. Powers the Live page.