NeurIPS Paper Releases Market-1T, a Trillion-Entry Stock Market Dataset

A NeurIPS paper released Market-1T, a dataset of roughly one trillion 1Hz stock observations spanning 2008–2025, to build financial world models. The work benchmarks 18 encoder strategies and proposes LeJEPA for market-state modeling, arguing joint-embedding approaches suit financial data better than autoregressive reconstruction.

2026-10-09 ~ 2026-10-09 · 4 related posts

• NeurIPS Paper Releases Market-1T, a ~1 Trillion-Observation Financial Dataset — randall_balestr · 2026-10-09
• Why financial data is ripe for JEPAs: latent-space market modeling in new NeurIPS paper — randall_balestr · 2026-10-09
• NeurIPS paper releases Market-1T: ~1 trillion 1Hz equity observations for world models — randall_balestr · 2026-10-09
• Market-1T: A Trillion 1Hz Equity Observations From 2008–2025, Open-Sourced — randall_balestr · 2026-10-09