NeuroFusion Reports Top Preliminary ForecastBench Score for Financial AI
The AMW Read
The preliminary benchmark result is an incremental update for a tracked AI-fintech player, with potential relevance to financial-research workflows but no demonstrated segment-wide impact yet.
NeuroFusion Reports Top Preliminary ForecastBench Score for Financial AI
South Korean fintech AI startup NeuroFusion said it ranked first among 464 models on ForecastBench's preliminary leaderboard, reporting a Brier Index score of 67.5. The company said the benchmark, run by the Forecasting Research Institute and presented at ICLR 2025, poses questions whose outcomes occur after submission, limiting the possibility of training on already-known answers. NeuroFusion said its score surpassed reported reference points for human superforecasters at 63.6 and Google DeepMind's top model at 65.0.
The result is notable because it tests probabilistic forecasting against unresolved real-world events rather than a static benchmark set. NeuroFusion attributes the performance to the operating stack behind Valley AI, its financial-investment platform, and CereFin, its investment-research agent: macroeconomic and market time-series processing, probabilistic reasoning over news and geopolitical events, and analysis of implied market consensus. Per the AI Market Watch index, NeuroFusion is a 2021-founded AI-in-fintech company, although the index tracks roughly 5,000 companies and is coverage rather than a census.
For builders, the claim underscores that financial AI differentiation may depend less on a general model alone than on calibration, data preparation, and workflows that combine structured market signals with qualitative evidence. For investors, the preliminary status matters: the central diligence question is whether the reported leaderboard lead persists over more forecast cycles and translates into measurable value for Valley AI's users rather than remaining a benchmark-specific result.
