benchmark

Political Even-Handedness Evaluation

benchmarkactiveprovisionalpolitical-even-handedness-evaluation-65524efd·1 events·first seen 15d ago

Aliases: Political Even-Handedness Evaluation

Co-occurring entities

claude.ai Claude Sonnet 4.5 Grok 4 Ideological Turing Test Gemini-2.5-Pro Llama GPT-5.5 Anthropic

More like this (12)

political bias evaluation Persuasion Index Persuasion Index: A Theory-Guided Framework for Persuasion Analysis From Self to Other: Evaluating Demographic Perspective-Taking in LLM Hate Speech Annotation Political Consistency Training (PCT)geopolitical bias A Resource for Enthymeme Detection in Controversial Political Discourse proactive assistance evaluation Skill-RM: Unifying Heterogeneous Evaluation Criteria via Agent Skill Helpfulness Consistency Sentiment Consistency ValueEval

Recent events (1)

7Anthropic News·15d ago·source ↗

Anthropic Publishes Political Even-Handedness Evaluation for Claude, Open-Sources Methodology

Anthropic has released a detailed account of how it trains and evaluates Claude for political even-handedness, including character traits instilled via reinforcement learning since early 2024 and a new automated evaluation methodology. The evaluation tests thousands of prompts across hundreds of political stances and benchmarks Claude Sonnet 4.5 against GPT-5, Llama 4, Grok 4, and Gemini 2.5 Pro, finding Claude comparable to Grok 4 and Gemini 2.5 Pro and more even-handed than GPT-5 and Llama 4. Anthropic is open-sourcing the evaluation framework to encourage shared industry standards for measuring political bias. The post also discloses the specific system prompt language used on Claude.ai to enforce even-handed behavior.

Frontier Model Releases Evaluation and Benchmarking claude.ai Claude Sonnet 4.5 Grok 4 +8 more