The FACTS Leaderboard: A Comprehensive Benchmark for Large Language Model Factuality

TL;DR

The FACTS Leaderboard introduces a comprehensive, multi-faceted benchmark suite to evaluate large language models' factual accuracy across diverse scenarios, including multimodal, parametric, search, and grounding tasks.

Contribution

It presents a new holistic benchmark suite with automated scoring for assessing factuality in various contexts, enhancing evaluation robustness.

Findings

Provides a balanced assessment of models' factuality

Includes four diverse sub-leaderboards for comprehensive evaluation

Employs automated judges for scalable scoring

Abstract

We introduce The FACTS Leaderboard, an online leaderboard suite and associated set of benchmarks that comprehensively evaluates the ability of language models to generate factually accurate text across diverse scenarios. The suite provides a holistic measure of factuality by aggregating the performance of models on four distinct sub-leaderboards: (1) FACTS Multimodal, which measures the factuality of responses to image-based questions; (2) FACTS Parametric, which assesses models' world knowledge by answering closed-book factoid questions from internal parameters; (3) FACTS Search, which evaluates factuality in information-seeking scenarios, where the model must use a search API; and (4) FACTS Grounding (v2), which evaluates whether long-form responses are grounded in provided documents, featuring significantly improved judge models. Each sub-leaderboard employs automated judge models to score model responses, and the final suite score is an average of the four components, designed to provide a robust and balanced assessment of a model's overall factuality. The FACTS Leaderboard Suite will be actively maintained, containing both public and private splits to allow for external participation while guarding its integrity. It can be found at https://www.kaggle.com/benchmarks/google/facts .