Why financial data is ripe for JEPAs: latent-space market modeling in new NeurIPS paper
randall_balestr · x · 2026-10-09
In their NeurIPS paper 'Towards Financial World Modeling', the authors argue standard autoregressive/generative models get trapped reconstructing raw price ticks, wasting capacity on unpredictable microstructure noise. JEPAs instead predict in latent space, capturing underlying market states. The accompanying Market-1T dataset and benchmarks resolve the lack of financial data and evaluation.
Related event: NeurIPS Paper Releases Market-1T, a Trillion-Entry Stock Market Dataset(4 posts)→
More from Research
- Inherit-MAS cuts multi-agent token use by up to 34.6% with evolution-inspired inheritance — Songtao Wei · 2026-10-09
- Zero human labels: auto-generated soccer tracking dataset hits 62.5 HOTA, beating fine-tuned FairMOT — RexDouglass · 2026-10-09
- Why AI doesn't actually read words: from BPE subwords to byte-level models like BLT and H-Net — jbhuang0604 · 2026-10-09
- NVIDIA details HSTU recommender inference stack with up to 5.93x lower latency — PyTorch · 2026-10-09
- Preprint: LLMs store numbers as curves and helices, but compute comparisons differently — tweetsatpreet · 2026-10-09
- Best model was cheapest: open-weights model ran 669 clinical decisions for 1.7 cents — antoine_chaffin · 2026-10-09