网页信息提取能力哪家强？GPT4、Claude、perplexity、kimi、通义千问大比拼

现在想把网页上的书名都提取出来，测试了国内外几个主流AI大模型，提示词：

具体结果如下：

GPT4表现很棒，但是只提取了20个书名，原页面中有40个书名；

根据提供的网页内容，这里列出了一些"Step into Reading"系列中儿童小说的书名:

Show me the Honey (Dr. Seuss/Cat in the Hat) by Tish Rabe, Christopher Moroney (Illustrator)

Danger: Dinosaurs! (Jurassic World) by Courtney Carbone, Random House (Illustrator)

Quest for the Heart (Disney Moana) by RH Disney, RH Disney (Illustrator)

Dinosaur Rescue! (Jurassic World) by Kristen L. Depken, Random House (Illustrator)

Dinosaurs in the Wild! by Dennis R. Shealy, Random House (Illustrator)

Dolphins! (Step into Reading) by Sharon Bokoske, Richard Courtney (Illustrator)

Poppy and Branch's Big Adventure by Mona Miller

Arthur's Reading Trick (Step into Reading) by Marc Brown

Elsa's Epic Journey (Disney Frozen 2) by Susan Amerikaner, Disney Storybook Art Team (Illustrator)

The Best Mistake Ever! And Other Stories by Richard Scarry

The Missing Tooth (Step into Reading) by Joanna Cole

Scooby's Scary Christmas! by Random House

Hooray for Hair! (Dr. Seuss/Cat in the Hat) by Tish Rabe, Tom Brannon (Illustrator)

The Incredible Elastigirl (Disney/Pixar The Incredibles 2) by Natasha Bouchard, Disney Storybook Art Team (Illustrator)

Arthur in New York (Step into Reading) by Marc Brown

Arthur and the School Pet by Marc Brown

Journey into the Mind (Disney/Pixar Inside Out) by RH Disney, RH Disney (Illustrator)

Driven to Win! (Disney/Pixar Cars 3) by RH Disney, RH Disney (Illustrator)

Poppy's Party (DreamWorks Trolls) by Frank Berrios, Gabriella Matta (Illustrator), Fabio Laguna (Illustrator), Francesco Legramandi (Illustrator)

Wonder Woman for President/Rule the School! (DC Super Hero Girls) by Shea Fontana, Dario Brizuela (Illustrator)

这个列表包含了部分书名，以及每本书的作者和插画师。

perplexity.ai在胡编：

Claude3也是瞎编高手：

Kimichat无法读取网页，但给出了解决方法

通义千问也是无法访问网页

豆包直接说不行

综上，GPT4的网页信息提取能力最突出，结果最令人满意，其他基本没有提取出来。更令人担心的是，有些AI模型会像模像样的给出看起来很像的假结果。