The HAAS Score: 6 dimensions, transparent methodology

Every model in our evaluation roster is selected through a documented, four-criteria methodology:

Our evaluation system combines three methodological layers, each grounded in peer-reviewed research:

Swiss-Bench (D1 Performance): Our proprietary benchmark uses a three-phase ground truth construction pipeline inspired by the data curation methodology of OpenScholar (Asai et al., Nature, 2026): expert drafting from primary statutory sources, adversarial verification, and quality-gated filtering. A 100-item subset was independently validated by a Swiss legal expert (MLaw, University of Fribourg), achieving 100% Legal Accuracy and 0% rated Incorrect. Model responses are scored by a blind three-judge LLM panel across three dimensions (Legal Accuracy, Citation Accuracy, Completeness) with majority-vote aggregation, following the multi-judge ensemble methodology of Zheng et al. (NeurIPS, 2023). The complete Swiss-Bench methodology is documented in our published ArXiv paper (Uenal, 2026).

Compl-AI (D2 Robustness, D4 Compliance): EU AI Act compliance scoring adapts the COMPL-AI framework (ETH Zurich / INSAIT, 2024), mapping regulatory principles to technical requirements. Recognized by the OECD.

Inspect AI (D3 Safety): Safety and adversarial testing builds on the UK AI Security Institute evaluation framework, adopted by leading AI labs.

Documentation (D6): Transparency assessment follows the Model Cards framework (Mitchell et al., 2019), operationalized as a structured checklist aligned with EU AI Act Article 11 technical documentation requirements.

Our holistic evaluation philosophy follows HELM (Stanford CRFM, peer-reviewed in TMLR). Swiss legal translation evaluation (D5) builds on methodology validated by Niklaus et al. (EMNLP 2023, ACL 2025) covering 180,000+ Swiss legal translation pairs. Related benchmarking work includes MMLU-Redux (Gema et al., NAACL 2025), CUAD (Hendrycks et al., NeurIPS 2021), and LegalBench (Guha et al., NeurIPS 2023). In total, our methodology draws on 100+ peer-reviewed publications.

Every evaluation follows a documented methodology with deterministic scoring. While LLM outputs have inherent variability, our structured scoring framework (temperature 0, fixed prompts, multi-judge majority vote) maximises consistency. You receive detailed benchmark results, scoring breakdowns, and full methodology documentation, sufficient to verify and understand every finding.

This is not an opinion. It’s evidence.

Helvetic AI has no commercial relationships with any AI model provider. No referral fees, no vendor partnerships, no pay-for-score agreements. Every model is evaluated with the same system, the same benchmarks, and the same scoring methodology.

1. Asai, A. et al. “Citation correctness in large language models.” Nature, 2026.
2. Uenal, F. “Swiss-Bench SBP-002: A Frontier Model Comparison on Swiss Legal and Regulatory Tasks.” ArXiv, 2026.
3. Dobreva, R. et al. “Compliance assessment of LLMs against EU AI Act requirements.” ETH Zürich / INSAIT, 2024.
4. Liang, P. et al. “Holistic Evaluation of Language Models (HELM).” TMLR, 2023. (Stanford CRFM)
5. UK AI Security Institute. “Evaluation framework for AI systems.” 2024.
6. Niklaus, J. et al. “MultiLegalPile: a 689GB multilingual legal corpus.” EMNLP, 2023.
7. Niklaus, J. et al. “Swiss legal translation evaluation: 180,000+ translation pairs.” ACL, 2025.
8. Gema, A.P. et al. “MMLU-Redux: Fixing expert-written evaluation sets.” NAACL, 2025.
9. Hendrycks, D. et al. “CUAD: An expert-annotated NLP dataset for legal contract review.” NeurIPS, 2021.
10. Guha, N. et al. “LegalBench: A collaboratively built benchmark for measuring legal reasoning.” NeurIPS, 2023.
11. OECD. “AI risk management and governance frameworks.” OECD AI Policy Observatory, 2024.
12. Zheng, L. et al. “Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena.” NeurIPS, 2023.
13. Mitchell, M. et al. “Model Cards for Model Reporting.” FAT*, 2019.