gowtham-15/

AACB-v1.0

Authors: [Gowtham P]Date: March 2026DOI: [10.5281/zenodo.19418848]

The AI Alignment & Constraint Benchmark (AACB) is a reproducible evaluation suite designed to assess large language models across four dimensions:

Creativity Instruction Following Bias Mitigation Hallucination Avoidance

It consists of 8 diverse prompts with explicit constraints, a 5‑point scoring rubric, and human‑annotated responses from… See the full description on the dataset page: https://huggingface.co/datasets/gowtham-15/AACB-v1.0.