NeurIPS Paper Releases Market-1T, a ~1 Trillion-Observation Financial Dataset
randall_balestr · x · 2026-10-09
The NeurIPS paper 'Towards Financial World Modeling' (authors include Randall Balestriero) lays a foundation for financial world models:
- Market-1T dataset: nearly one trillion observations across US equities from 2008–2025 at 1 Hz resolution.
- Rigorous evaluation protocol: assessing representations both on finance prediction tasks and via probes of latent structure.
- Systematic study: comparing 18 encoder-training strategies across nearly two decades of market regimes.
Key finding: encoders with similar predictive performance can organize market state very differently. The authors argue financial data — noisy, non-stationary, with unknown ground-truth states — is a natural fit for JEPA-style world models like DINO-WM, V-JEPA 2, and LeWM.
Related event: NeurIPS Paper Releases Market-1T, a Trillion-Entry Stock Market Dataset(4 posts)→
More from Venture
- Union Square Ventures shrinks partnership, sizes new funds at $900M — nmasc_ · 2026-10-09
- The operator behind WisprFlow's paid ads shares how he'd rebuild the program in 30 days — VibeMarketer_ · 2026-10-09
- Monetizing proprietary data through MCP: pay-per-call pricing for agents — ValosSantanos · 2026-10-09
- "Journalists shouldn't report ARR numbers — it's a fake science," and an AI professor agrees — DavidSKrueger · 2026-10-09
- Catalyst launches autonomous AI trading agent backed by Jake Paul — jzlegion · 2026-10-09
- When compute debt outpaces user cash: margins flow from software to physical asset owners — LexSokolin · 2026-10-09