EXAONE Finance is a time‑series foundation model tailored for financial forecasting. It removes the quadratic cost of self‑attention and uses two linear‑time operators: a causal 1‑D convolution for temporal mixing and a group‑aware pooling MLP for variate mixing. During pre‑training a masked context augmentation exposes the model to contiguous missing spans, improving robustness to the intermittent observations typical of market data. The model is pretrained on a large financial corpus that spans equities, foreign exchange, commodities, crypto‑assets, fixed‑income and macro‑economic indicators. On the FinVerse benchmark, which covers diverse asset classes, EXAONE Finance achieves state‑of‑the‑art results across point‑forecast accuracy, cross‑sectional asset ranking and portfolio profitability, ranking first in all three tiers.
Review: By simplifying the architecture and explicitly handling missing data, the model proves highly practical for finance and highlights the promise of large‑scale financial pre‑training.