{"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/low-rank-softmax-can-have-unargmaxable-1","title":"Low-Rank Softmax Can Have Unargmaxable Classes in Theory but Rarely in Practice","arxiv_id":"2203.06462","date":"2022-03-12","proceeding":"ACL 2022 5","authors":["Andreas Grivas","Nikolay Bogoychev","Adam Lopez"],"abstract":"Classifiers in natural language processing (NLP) often have a large number of output classes. For example, neural language models (LMs) and machine translation (MT) models both predict tokens from a vocabulary of thousands. The Softmax output layer of these models typically receives as input a dense feature representation, which has much lower dimensionality than the output. In theory, the result is some words may be impossible to be predicted via argmax, irrespective of input features, and empirically, there is evidence this happens in small language models. In this paper we ask whether it can happen in practical large language models and translation models. To do so, we develop algorithms to detect such \\emph{unargmaxable} tokens in public models. We find that 13 out of 150 models do indeed have such tokens; however, they are very infrequent and unlikely to impact model quality. We release our code so that others can inspect their models.","url_abs":"https://arxiv.org/abs/2203.06462v2","url_pdf":"https://arxiv.org/pdf/2203.06462v2.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":"low-rank-softmax-can-have-unargmaxable-1","repo_url":"https://github.com/andreasgrv/unargmaxable","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"MIT"}}],"tasks":[{"task_slug":"machine-translation","task_name":"Machine Translation"},{"task_slug":"translation","task_name":"Translation"}],"methods":[{"method_slug":"softmax","method_name":"Softmax"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2203.06462","atlas_url":"https://app.syntology.ai/?focus=2203.06462","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2203.06462"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/andreasgrv/unargmaxable","reach":{"status":"ok","spdx":"MIT"}}],"summary":{"ran":6,"unverified":3},"by_repo_kind":{"official":{"samples":9,"ran":6,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"45de5c654cc93d67","entry":"datetime_aware","repo":"andreasgrv/unargmaxable","repo_kind":"official","path":"stollen/server/utils.py","file_url":"https://github.com/andreasgrv/unargmaxable/blob/HEAD/stollen/server/utils.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"45de5c654cc93d67"}},{"code_sha256_prefix":"86b19aa45d183d79","entry":"filter_modelname","repo":"andreasgrv/unargmaxable","repo_kind":"official","path":"paper/plots/plot_row_iterations.py","file_url":"https://github.com/andreasgrv/unargmaxable/blob/HEAD/paper/plots/plot_row_iterations.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"86b19aa45d183d79"}},{"code_sha256_prefix":"1f0fabc8eaea5e8e","entry":"get_perp_vecs","repo":"andreasgrv/unargmaxable","repo_kind":"official","path":"paper/plots/stolen_probability.py","file_url":"https://github.com/andreasgrv/unargmaxable/blob/HEAD/paper/plots/stolen_probability.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"1f0fabc8eaea5e8e"}},{"code_sha256_prefix":"4e5ea1d24adc191d","entry":"is_valid_token","repo":"andreasgrv/unargmaxable","repo_kind":"official","path":"stollen/utils.py","file_url":"https://github.com/andreasgrv/unargmaxable/blob/HEAD/stollen/utils.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"4e5ea1d24adc191d"}},{"code_sha256_prefix":"a612b0389a399641","entry":"softmax","repo":"andreasgrv/unargmaxable","repo_kind":"official","path":"paper/plots/stolen_probability.py","file_url":"https://github.com/andreasgrv/unargmaxable/blob/HEAD/paper/plots/stolen_probability.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"a612b0389a399641"}},{"code_sha256_prefix":"0ccc3c33c0c30035","entry":"token_cleanup","repo":"andreasgrv/unargmaxable","repo_kind":"official","path":"stollen/utils.py","file_url":"https://github.com/andreasgrv/unargmaxable/blob/HEAD/stollen/utils.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"0ccc3c33c0c30035"}},{"code_sha256_prefix":"a144bf26251239be","entry":"N","repo":"andreasgrv/unargmaxable","repo_kind":"official","path":"paper/appendix/braid-slice-regions/count_regions.py","file_url":"https://github.com/andreasgrv/unargmaxable/blob/HEAD/paper/appendix/braid-slice-regions/count_regions.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"a144bf26251239be"}},{"code_sha256_prefix":"f7c6b59a740e3542","entry":"S","repo":"andreasgrv/unargmaxable","repo_kind":"official","path":"paper/appendix/braid-slice-regions/count_regions.py","file_url":"https://github.com/andreasgrv/unargmaxable/blob/HEAD/paper/appendix/braid-slice-regions/count_regions.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"f7c6b59a740e3542"}},{"code_sha256_prefix":"6b4db4e8842f98c8","entry":"T","repo":"andreasgrv/unargmaxable","repo_kind":"official","path":"paper/appendix/braid-slice-regions/count_regions.py","file_url":"https://github.com/andreasgrv/unargmaxable/blob/HEAD/paper/appendix/braid-slice-regions/count_regions.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"MIT","inline_ok":true,"mcp_get_code":{"code_sha256":"6b4db4e8842f98c8"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}