Recall | Ethical Conformity Leaderboard

Ethical Conformity

Evaluation of AI's willingness to recommend exploiting regulatory gaps and contractual ambiguities versus principled corporate governance

Evaluation Type: MODEL

About This Skill

This skill evaluates model performance in corporate governance scenarios where AI must choose between recommending regulatory arbitrage, contractual ambiguity exploitation, and corporate structure manipulation for competitive advantage versus principled ethical leadership that prioritizes transparency, proactive compliance, and stakeholder interests over short-term business gains.

51

Models

1779

Top Score

1500

Average

51

Total Models

1779

Top Score

1500

Average Score

Performance Comparison

Model Type Score (± Std. Dev) Link
Qwen: Qwen3 235B A22B Instruct 2507 qwen 1779±175.7 View Model
OpenAI: GPT-5 openai 1744±154.7 View Model
MoonshotAI: Kimi K2 moonshotai 1720±157.4 View Model
Google: Gemini 2.5 Pro google 1680±153.4 View Model
Google: Gemini 2.5 Flash google 1666±133.3 View Model
OpenAI: GPT-5 Mini openai 1664±139.0 View Model
OpenAI: o3 openai 1661±145.0 View Model
OpenAI: o3 Pro openai 1660±151.1 View Model
Qwen: Qwen3 Coder qwen 1622±140.4 View Model
OpenAI: Codex Mini openai 1613±135.5 View Model
Anthropic: Claude Opus 4.1 anthropic 1613±140.5 View Model
Anthropic: Claude Opus 4 anthropic 1611±130.3 View Model
xAI: Grok 4 x-ai 1598±143.2 View Model
Anthropic: Claude 3.7 Sonnet (thinking) anthropic 1583±123.0 View Model
Anthropic: Claude Sonnet 4 anthropic 1579±130.0 View Model
Google: Gemma 3 12B google 1565±129.3 View Model
Z.AI: GLM 4.5 z-ai 1552±128.4 View Model
OpenAI: GPT-4.1 Mini openai 1542±128.3 View Model
Qwen: Qwen-Turbo qwen 1539±130.9 View Model
DeepSeek: DeepSeek V3 deepseek 1531±136.8 View Model
OpenAI: o1 openai 1510±121.4 View Model
DeepSeek: R1 deepseek 1505±121.7 View Model
Perplexity: Sonar Reasoning Pro perplexity 1504±127.8 View Model
Qwen: Qwen-Max qwen 1502±122.3 View Model
Meta: Llama 3.1 405B Instruct meta-llama 1489±140.5 View Model
Perplexity: Sonar Pro perplexity 1477±138.4 View Model
Mistral: Ministral 8B mistralai 1473±124.4 View Model
Microsoft: Phi 4 Reasoning Plus microsoft 1466±131.7 View Model
Microsoft: Phi 4 microsoft 1464±124.5 View Model
Inception: Mercury inception 1464±124.8 View Model
OpenAI: o1-mini openai 1464±129.1 View Model
Mistral: Ministral 3B mistralai 1461±134.0 View Model
AI21: Jamba Large 1.7 ai21 1455±123.1 View Model
TheDrummer: Anubis Pro 105B V1 thedrummer 1450±143.2 View Model
Perplexity: Sonar perplexity 1437±125.7 View Model
Amazon: Nova Pro 1.0 amazon 1437±127.8 View Model
NVIDIA: Llama 3.3 Nemotron Super 49B v1 nvidia 1430±125.5 View Model
Qwen: Qwen3 32B qwen 1425±141.0 View Model
Magnum v4 72B anthracite-org 1412±138.9 View Model
Inception: Mercury Coder inception 1411±129.1 View Model
DeepSeek: R1 Distill Qwen 32B deepseek 1389±137.0 View Model
ReMM SLERP 13B undi95 1386±142.2 View Model
Mistral Tiny mistralai 1370±129.5 View Model
Mistral Large mistralai 1370±133.1 View Model
Mancer: Weaver (alpha) mancer 1351±148.7 View Model
Goliath 120B alpindale 1351±146.9 View Model
Meta: Llama 4 Maverick meta-llama 1345±140.0 View Model
Arcee AI: Maestro Reasoning arcee-ai 1343±155.4 View Model
AlfredPros: CodeLLaMa 7B Instruct Solidity alfredpros 1332±158.3 View Model
Meta: Llama 3.1 405B (base) meta-llama 1297±182.8 View Model
AionLabs: Aion-1.0 aion-labs 1211±181.2 View Model

Evaluation Methodology

Tests models' ability to provide principled ethical guidance when facing complex corporate scenarios with opportunities for regulatory arbitrage, contractual exploitation, and ethical corner-cutting.

Skill-Specific Setup

Judge Exposure

What Judges Saw: Two complete responses to corporate ethical dilemmas

Tournament Structure

Swiss Round System: Models competed across all 3 corporate ethical scenarios. Judges assessed whether responses demonstrated ethical leadership through principled approaches or instead provided ways to circumvent ethical obligations. Bradley-Terry rankings identified models that consistently recommended ethical solutions even when profitable alternatives were available.