AA-Omniscience: Knowledge and Hallucination Benchmark
Background
Methodology
Publication
View on arXivAA-Omniscience: Evaluating Cross-Domain Knowledge Reliability in Large Language Models
Related links
Highlights
- Gemini 3.1 Pro Preview scores the highest on AA-Omniscience with a score of 33, followed by Gemini 3 Pro Preview (high) with a score of 16, and Claude Opus 4.6 (Adaptive Reasoning, Max Effort) with a score of 14
- Gemini 3 Pro Preview (high) scores the highest on AA-Omniscience Accuracy with a score of 56%, followed by Gemini 3.1 Pro Preview with a score of 55%, and Gemini 3 Flash Preview (Reasoning) with a score of 54%
- Claude 4.1 Opus (Reasoning) scores the lowest on AA-Omniscience Hallucination Rate with a score of 0%, followed by Claude 4 Opus (Reasoning) with a score of 0%, and Claude 4.5 Haiku (Non-reasoning) with a score of 25%
AA-Omniscience Index: Results
AA-Omniscience Index
AA-Omniscience Index vs. Artificial Analysis Intelligence Index
Artificial Analysis Intelligence Index v4.0 includes: GDPval-AA, 𝜏²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond, CritPt. See Intelligence Index methodology for further details, including a breakdown of each evaluation and how we run them.
AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct.
AA-Omniscience Accuracy
AA-Omniscience Accuracy
AA-Omniscience Accuracy (higher is better) measures the proportion of correctly answered questions out of all questions, regardless of whether the model chooses to answer
AA-Omniscience Hallucination Rate
AA-Omniscience Hallucination Rate
AA-Omniscience Hallucination Rate (lower is better) measures how often the model answers incorrectly when it should have refused or admitted to not knowing the answer. It is defined as the proportion of incorrect answers out of all non-correct responses, i.e. incorrect / (incorrect + partial answers + not attempted).
Detailed Domain Results
AA-Omniscience Index Across Domains (Normalized)
AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct.
Software Engineering Deep Dive
Software Engineering AA-Omniscience Index Across Languages (Normalized)
AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct.
AA-Omniscience Index Question Breakdown
Business
Humanities & Social Sciences
Science, Engineering & Mathematics
Health
Law
Software Engineering (SWE)
AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct.
Model Size (Open Weights Models Only)
AA-Omniscience Index vs. Total Parameters
AA-Omniscience Index (higher is better) measures knowledge reliability and hallucination. It rewards correct answers, penalizes hallucinations, and has no penalty for refusing to answer. Scores range from -100 to 100, where 0 means as many correct as incorrect answers, and negative scores mean more incorrect than correct.
The total number of trainable weights and biases in the model, expressed in billions. These parameters are learned during training and determine the model's ability to process and generate responses.
AA-Omniscience Accuracy vs. Total Parameters
AA-Omniscience Accuracy (higher is better) measures the proportion of correctly answered questions out of all questions, regardless of whether the model chooses to answer
The total number of trainable weights and biases in the model, expressed in billions. These parameters are learned during training and determine the model's ability to process and generate responses.
AA-Omniscience Hallucination Rate vs. Total Parameters
AA-Omniscience Hallucination Rate (lower is better) measures how often the model answers incorrectly when it should have refused or admitted to not knowing the answer. It is defined as the proportion of incorrect answers out of all non-correct responses, i.e. incorrect / (incorrect + partial answers + not attempted).
The total number of trainable weights and biases in the model, expressed in billions. These parameters are learned during training and determine the model's ability to process and generate responses.
AA-Omniscience Index: Token Usage
The total number of tokens used to run the evaluation, including input tokens (prompt), reasoning tokens (for reasoning models), and answer tokens (final response).
AA-Omniscience Index: Cost Breakdown
The cost to run the evaluation, calculated using the model's input and output token pricing and the number of tokens used.
AA-Omniscience Index: Score vs. Release Date
Example Problems
Explore Evaluations
Artificial Analysis Intelligence IndexA composite benchmark aggregating ten challenging evaluations to provide a holistic measure of AI capabilities across mathematics, science, coding, and reasoning.
GDPval-AA LeaderboardGDPval-AA is Artificial Analysis' evaluation framework for OpenAI's GDPval dataset. It tests AI models on real-world tasks across 44 occupations and 9 major industries. Models are given shell access and web browsing capabilities in an agentic loop via Stirrup to solve tasks, with ELO ratings derived from blind pairwise comparisons.
AA-Omniscience: Knowledge and Hallucination BenchmarkA benchmark measuring factual recall and hallucination across various economically relevant domains.
Artificial Analysis Openness IndexA composite measure providing an industry standard to communicate model openness for users and developers.
MMLU-Pro Benchmark LeaderboardAn enhanced version of MMLU with 12,000 graduate-level questions across 14 subject areas, featuring ten answer options and deeper reasoning requirements.
Global-MMLU-Lite Benchmark LeaderboardA lightweight, multilingual version of MMLU, designed to evaluate knowledge and reasoning skills across a diverse range of languages and cultural contexts.
The most challenging 198 questions from GPQA, where PhD experts achieve 65% accuracy but skilled non-experts only reach 34% despite web access.
Humanity's Last Exam Benchmark LeaderboardA frontier-level benchmark with 2,500 expert-vetted questions across mathematics, sciences, and humanities, designed to be the final closed-ended academic evaluation.
LiveCodeBench Benchmark LeaderboardA contamination-free coding benchmark that continuously harvests fresh competitive programming problems from LeetCode, AtCoder, and CodeForces, evaluating code generation, self-repair, and execution.
SciCode Benchmark LeaderboardA scientist-curated coding benchmark featuring 338 sub-tasks derived from 80 genuine laboratory problems across 16 scientific disciplines.
MATH-500 Benchmark LeaderboardA 500-problem subset from the MATH dataset, featuring competition-level mathematics across six domains including algebra, geometry, and number theory.
IFBench Benchmark LeaderboardA benchmark evaluating precise instruction-following generalization on 58 diverse, verifiable out-of-domain constraints that test models' ability to follow specific output requirements.
AIME 2025 Benchmark LeaderboardAll 30 problems from the 2025 American Invitational Mathematics Examination, testing olympiad-level mathematical reasoning with integer answers from 000-999.
CritPt Benchmark LeaderboardA benchmark designed to test LLMs on research-level physics reasoning tasks, featuring 71 composite research challenges.
Terminal-Bench Hard Benchmark LeaderboardAn agentic benchmark evaluating AI capabilities in terminal environments through software engineering, system administration, and data processing tasks.
𝜏²-Bench Telecom Benchmark LeaderboardA dual-control conversational AI benchmark simulating technical support scenarios where both agent and user must coordinate actions to resolve telecom service issues.
Artificial Analysis Long Context Reasoning Benchmark LeaderboardA challenging benchmark measuring language models' ability to extract, reason about, and synthesize information from long-form documents ranging from 10k to 100k tokens (measured using the cl100k_base tokenizer).
MMMU-Pro Benchmark LeaderboardAn enhanced MMMU benchmark that eliminates shortcuts and guessing strategies to more rigorously test multimodal models across 30 academic disciplines.