Dali
The open verification layer for AI
Dali is the open verification layer for AI: it creates, scores, and preserves evidence so AI-assisted outputs can be independently verified, exchanged, and replayed. Legal AI is the proving ground.
Open Artifacts
Dali Citation Benchmark (Seed Sample)
· available
5 hand-curated cases, 14 authorities — methodology review sample (not the full run).
Full evaluation run
· available
524 citations · 3 models · 5 jurisdiction tracks — on GitHub data/results/.
Evidence Infrastructure
· available (early-stage v0.2.1)
Open verification engine, methodologies, and reproducible evaluation workflows.
Quick Start
from datasets import load_dataset
dataset = load_dataset("yenklabs/open-evidence-corpus")GitHub · Contribute · yenklabs.com
Research models (roadmap)
Lightweight, reproducible research models — not foundation models. All planned.
| Release | Model | Purpose | Status |
|---|---|---|---|
| v0.1 | Dali Verification Taxonomy Classifier | Predict standardized verification outcome labels | planned |
| v0.2 | Dali Citation Risk Classifier | Estimate citation verification risk from evidence metadata | planned |
| v0.3 | Dali Authority Matching Baseline | Reproducible baseline for authority matching experiments | planned |
| v0.4 | Dali Proposition Support Classifier | Classify proposition support relationships for legal authorities | planned |
Ecosystem
- Datasets — Open Evidence Corpus, Citation Benchmark seed sample, Verification Taxonomy · available
- Full evaluation run — 524 citations on github.com/yenklabs/Dali/data/results · available
- Models — Taxonomy Classifier, Citation Risk, Authority Matching, Proposition Support · planned
- Spaces — Evidence Explorer, Benchmark Dashboard, Citation Verification Demo · planned
- Future datasets — Evaluation Prompts, Replay Corpus, Evidence Artifacts · planned
Dali is the open verification layer for AI: it creates, scores, and preserves evidence so AI-assisted outputs can be independently verified, exchanged, and replayed.