Merging Facts, Crafting Fallacies: Evaluating the Contradictory Nature of Aggregated Factual Claims in Long-Form Generations
This addresses a critical flaw in evaluating LLM outputs for applications requiring high factual precision, though it is an incremental improvement to existing metrics.
The paper tackles the problem that existing factuality metrics overestimate the accuracy of long-form LLM generations when facts are combined with entity ambiguity, and introduces D-FActScore which reduces this overestimation by over 10% compared to prior metrics.
Long-form generations from large language models (LLMs) contain a mix of factual and non-factual claims, making evaluating factuality difficult. Prior works evaluate the factuality of a long paragraph by decomposing it into multiple facts, verifying those facts independently, and aggregating the results. Such methods assume that combining factual claims forms a factual paragraph. The above assumption can be violated: we show that strong open-source models like Llama-chat can generate paragraphs that contain verifiable facts, but the facts are combined into a non-factual paragraph due to entity ambiguity. We further reveal that existing factuality metrics, including FActScore and citation recall, cannot properly evaluate these non-factual paragraphs and overestimate their factuality. To address this, we introduce an enhanced metric, D-FActScore, specifically designed for content with ambiguous entities. We evaluate the D-FActScores of people biographies generated by retrieval-augmented LLMs. We show that D-FActScore can better assess the factuality of paragraphs with entity ambiguity than FActScore. We also find that four widely used open-source LLMs tend to mix information of distinct entities to form non-factual paragraphs, making their D-FActScore much lower than FActScore by over 10%.