{"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/unmasking-clever-hans-predictors-and","title":"Unmasking Clever Hans Predictors and Assessing What Machines Really Learn","arxiv_id":"1902.10178","date":"2019-02-26","proceeding":null,"authors":["Sebastian Lapuschkin","Stephan Wäldchen","Alexander Binder","Grégoire Montavon","Wojciech Samek","Klaus-Robert Müller"],"abstract":"Current learning machines have successfully solved hard application problems,\nreaching high accuracy and displaying seemingly \"intelligent\" behavior. Here we\napply recent techniques for explaining decisions of state-of-the-art learning\nmachines and analyze various tasks from computer vision and arcade games. This\nshowcases a spectrum of problem-solving behaviors ranging from naive and\nshort-sighted, to well-informed and strategic. We observe that standard\nperformance evaluation metrics can be oblivious to distinguishing these diverse\nproblem solving behaviors. Furthermore, we propose our semi-automated Spectral\nRelevance Analysis that provides a practically effective way of characterizing\nand validating the behavior of nonlinear learning machines. This helps to\nassess whether a learned model indeed delivers reliably for the problem that it\nwas conceived for. Furthermore, our work intends to add a voice of caution to\nthe ongoing excitement about machine intelligence and pledges to evaluate and\njudge some of these recent successes in a more nuanced manner.","url_abs":"http://arxiv.org/abs/1902.10178v1","url_pdf":"http://arxiv.org/pdf/1902.10178v1.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":"unmasking-clever-hans-predictors-and","repo_url":"https://github.com/sebastian-lapuschkin/lrp_toolbox","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"tf","reach":{"status":"ok","spdx":"NOASSERTION"}}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1902.10178","atlas_url":"https://app.syntology.ai/?focus=1902.10178","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}