{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/the-goldilocks-principle-reading-childrens","title":"The Goldilocks Principle: Reading Children's Books with Explicit Memory Representations","arxiv_id":"1511.02301","date":"2015-11-07","proceeding":null,"authors":["Felix Hill","Antoine Bordes","Sumit Chopra","Jason Weston"],"abstract":"We introduce a new test of how well language models capture meaning in\nchildren's books. Unlike standard language modelling benchmarks, it\ndistinguishes the task of predicting syntactic function words from that of\npredicting lower-frequency words, which carry greater semantic content. We\ncompare a range of state-of-the-art models, each with a different way of\nencoding what has been previously read. We show that models which store\nexplicit representations of long-term contexts outperform state-of-the-art\nneural language models at predicting semantic content words, although this\nadvantage is not observed for syntactic function words. Interestingly, we find\nthat the amount of text encoded in a single memory representation is highly\ninfluential to the performance: there is a sweet-spot, not too big and not too\nsmall, between single words and full sentences that allows the most meaningful\ninformation in a text to be effectively retained and recalled. Further, the\nattention over such window-based memories can be trained effectively through\nself-supervision. We then assess the generality of this principle by applying\nit to the CNN QA benchmark, which involves identifying named entities in\nparaphrased summaries of news articles, and achieve state-of-the-art\nperformance.","url_abs":"http://arxiv.org/abs/1511.02301v4","url_pdf":"http://arxiv.org/pdf/1511.02301v4.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"the-goldilocks-principle-reading-childrens","repo_url":"https://github.com/facebookresearch/ParlAI","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok","spdx":"MIT"}},{"paper_slug":"the-goldilocks-principle-reading-childrens","repo_url":"https://github.com/joe-prog/https-github.com-facebookresearch-ParlAI","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"gone","observed_at":"2026-09-18","how":"tree_404+repo_404"}},{"paper_slug":"the-goldilocks-principle-reading-childrens","repo_url":"https://github.com/wujsAct/DeepLearningModels","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":{"status":"ok"}}],"tasks":[{"task_slug":"articles","task_name":"Articles"},{"task_slug":"language-modelling","task_name":"Language Modelling"}],"methods":[],"datasets_introduced":[{"slug":"cbt","name":"CBT","full_name":"Children’s Book Test"},{"slug":"children-s-book-test","name":"Children's Book Test","full_name":""}],"methods_introduced":[],"results":[],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1511.02301","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}