Research
Case Studies, Experiments & Publications:
Type:
Last UpdatedTypeTitleDescription
Apr 2026feature
The Checking SystemFact verification, bias detection, and censor auditing on live model outputs: how Aquin evaluates what the model says and what it avoids.Apr 2026feature
Feature BenchmarksInterpScore, FeaturePurityScore, and MUI: integrated benchmarks for evaluating SAE features on Llama 3.2 1B Instruct.Apr 2026feature
The Attribution SystemAquin's attribution pipeline: causal mediation analysis, SAE feature extraction and cross-token feature attribution on Llama 3.2 1B Instruct.
feature
The Checking System
Fact verification, bias detection, and censor auditing on live model outputs: how Aquin evaluates what the model says and what it avoids.
Last UpdatedTypeTitle
Apr 2026feature
The Checking SystemApr 2026feature
Feature BenchmarksApr 2026feature
The Attribution SystemNot sure if Aquin is right for you?
Aquin
