Learning from real decisionsDissei / Data

Teaching models
financial judgment.

Carefully built training data, evaluations and environments, grounded in real financial work.

Explore Dissei Benchmark
Training data Evaluations RL environments

Learning, under examination.

Studies in what models learn from finance, and where their judgment breaks.

Latest researchAll research
012026-09-17Research noteOnly if you ask: planting wrong numbers in a 20-step credit worksheet with three models022026-09-10EvalsSaying 70, meaning 40: guessing probabilities with five mid-tier models032026-09-03EvalsSitting their own exam: four models, 397 questions, and a mild case of self-bias
Dissei Benchmark

Financial judgment,
measured.

Explore the results across models and financial reasoning categories.

View benchmark results

Recorded model scores

Historical snapshot · 2026-08-31

Observed score / 100

Source: Dissei Benchmark

Score / 100. Case coverage differs between models; these are descriptive results, not a matched ranking.

View data table
Recorded model scores
ModelScore / 100
Claude Opus 4.844.26
DeepSeek v4-flash25.49
GPT-5.6 Sol45.53
Kimi K339.76
Muse Spark 1.242.41