[GitHub Trending] shiyu-coder/Kronos
6.3 relevance
Score Breakdown
technical depth 8
novelty 8
actionability 2
community 6
strategic 6
personal 7
Scored daily by a customisable AI persona to surface the most relevant engineering leadership news.
New foundation model for finance, technically deep but not immediately actionable.
Summary
Kronos, the first open-source foundation model for financial candlesticks, is trained on data from 45+ global exchanges and uses a two-stage framework: a specialized tokenizer quantizes OHLCV data into hierarchical discrete tokens, then an autoregressive Transformer is pre-trained. Released in sizes from 4.1M to 499.2M parameters (large is closed-source), it has been accepted by AAAI 2026 and includes a live demo for BTC/USDT forecasting.
Author
shiyu-coder