XForecast Challenge

Multimodal, News-Grounded Financial Forecasting

Financial markets are a demanding stress test for machine learning: extreme non-stationarity, low signal-to-noise ratios, and adversarial dynamics. Hosted at KDD 2026 Finance Day, the challenge invites participants to combine financial time-series with unstructured news text to forecast future stock prices under a realistic multimodal setting.

Join on Kaggle View Details

At a Glance

Prize Pool

$5,000

Timeline

Jun–Jul 2026

Dataset

FinTexTS

Universe

Top 100 Stocks

The Challenge

A Kaggle competition on multimodal, news-grounded financial forecasting.

Overview

The challenge explores how modern AI models can combine financial time-series with unstructured news text to improve stock forecasting. Participants receive historical price data for the top 100 stocks by market capitalization, paired with news articles associated with each stock and trading date from the FinTexTS dataset.

Models must leverage both past price dynamics and the associated news to predict closing prices four weeks ahead. For example, given all price and news data up to December 1, 2023, a model is asked to predict the stock price on December 29, 2023.

By promoting news-grounded forecasting, the challenge aims to bridge structured financial time-series with unstructured textual information, advancing multimodal and representation-learning methods for informative and robust financial prediction systems.

Key Details