FinDendro - Cutoff 2022 - Before: leaderboard
Dated financial fact recall and prediction under explicit time prompts. FinDendro v1 Table 8 uses candidate-likelihood Task 1 accuracy on controlled facts up to and after each training cutoff year. Each year has separate Before and After boards so compared models share the same fact cohort. Exact repository identities are pinned in Table 11. This is likelihood scoring, not instruction-prompt accuracy. Question access: Primary result reviewed; exact scored release availability not confirmed.
Metric: Accuracy (fraction). Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 2 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | TiMaGPT2-2022 (Benchmark Sweep Run: source=FinDendro; candidate-likelihood scoring; cutoff=2022) | 0.27 | |
| 2 | ChronoGPT-Instruct-2022 (Benchmark Sweep Run: source=FinDendro; candidate-likelihood scoring; cutoff=2022) | 0.27 |
Interactive version: theaggregate.ai/benchmark?slug=findendro-cutoff-2022-before · How It Works · Data refreshed daily, snapshot 2026-11-06.