“Hmmm... checking against external sources is an interesting idea -- but using Google as a source of ground truth is a little bit tricky, given how often these days Google itself is spitting up confabulated AI-generated crud (or other low-quality stuff).”
Long-form factuality in large language models
Large language models (LLMs) often generate content that contains factual errors when responding to fact-seeking prompts on open-ended topics. To benchmark a model's long-form factuality in open domains, we first use GPT-4 to generate LongFact, a prompt set comprising thousands of questions spanning 38 topics. We then propose that LLM agents can be used as automated evaluators for long-form factuality through a method which we call Search-Augmented Factuality Evaluator (SAFE). SAFE utilizes an LLM to break down a long-form response into a set of individual facts and to evaluate the accuracy of each fact using a multi-step reasoning process comprising sending search queries to Google Search and determining whether a fact is supported by the search results. Furthermore, we propose extending F1 score as an aggregated metric for long-form factuality. To do so, we balance the percentage of supported facts in a response (precision) with the percentage of provided facts relative to a hyperparameter representing a user's preferred response length (recall). Empirically, we demonstrate that LLM agents can outperform crowdsourced human annotators - on a set of ~16k individual facts, SAFE agrees with crowdsourced human annotators 72% of the time, and on a random subset of 100 disagreement cases, SAFE wins 76% of the time. At the same time, SAFE is more than 20 times cheaper than human annotators. We also benchmark thirteen language models on LongFact across four model families (Gemini, GPT, Claude, and PaLM-2), finding that larger language models generally achieve better long-form factuality. LongFact, SAFE, and all experimental code are available at https://github.com/google-deepmind/long-form-factuality.
候选层:至少 2 个不同的人、在 2 个不同月份提到过,还没到主书架的 3 人 3 个月。没有做立场标注,所以只写“被提及”。
HN 书评
没有评论直接链到它;下面是它被提交的讨论串里回复最多的评论(不计入统计)“Using Google as an arbitrator of truth? What could go wrong? The core concept of breaking down statements to check for factuality is otherwise sound.”
“So using LLMs to fact-check LLMs with Google results? There are non-LLM fact-checkers out there [1] which aggregate search results from more diverse sources and which are shown to outperform LLMs [2]. [1] https://editor.factiverse.ai [2] https://arxiv.org/abs/2402.12147”
- Long-form factuality in large language models ↗TheIronYuppie · 2024-03-28 · 1 分
- Long-Form Factuality in Large Language Models ↗tosh · 2024-03-28 · 1 分
- Long-form factuality in large language models ↗rootforce · 2024-03-29 · 18 分
- Long-form factuality in large language models ↗rntn · 2024-04-01 · 2 分
- Long-form factuality in large language models ↗PaulHoule · 2024-04-06 · 26 分