Financial Risk Signal Prediction

Surface risk signals from public filings automatically

Train models to extract and predict risk signals from SEC filings, earnings calls, and financial news. By treating future risk events as labels, models learn which filing language predicts actual adverse outcomes.

+11.6%
Brier Skill Score on 6,109 SEC risk queries
↗ Blog post
64.7%
lower calibration error than GPT-5
↗ Blog post

Example prediction questions

The kinds of questions a model trained on your data can answer.


Key results

Benchmark comparisons against frontier models

SEC Risk Prediction: Brier Score, Skill, and Calibration

Fine-tuned Qwen3-32B achieves Brier Skill Score +11.6% with ECE of 0.029 across 6,109 SEC risk queries — 64.7% lower calibration error than GPT-5 (ECE 0.081). The model learns to distinguish boilerplate legal language from meaningful signals preceding adverse outcomes.

Four charts: Brier score, Brier Skill Score, ECE, and calibration reliability diagram comparing trained Qwen3-32B vs. GPT-5 vs. Naive Baseline

Explore

Primary write-ups and artifacts for this solution.

Ready to build your own expert?

Leverage your own raw data or use public sources. No labeling required.