{"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/grokking-of-hierarchical-structure-in-vanilla","title":"Grokking of Hierarchical Structure in Vanilla Transformers","arxiv_id":"2305.18741","date":"2023-05-30","proceeding":null,"authors":["Shikhar Murty","Pratyusha Sharma","Jacob Andreas","Christopher D. Manning"],"abstract":"For humans, language production and comprehension is sensitive to the hierarchical structure of sentences. In natural language processing, past work has questioned how effectively neural sequence models like transformers capture this hierarchical structure when generalizing to structurally novel inputs. We show that transformer language models can learn to generalize hierarchically after training for extremely long periods -- far beyond the point when in-domain accuracy has saturated. We call this phenomenon \\emph{structural grokking}. On multiple datasets, structural grokking exhibits inverted U-shaped scaling in model depth: intermediate-depth models generalize better than both very deep and very shallow transformers. When analyzing the relationship between model-internal properties and grokking, we find that optimal depth for grokking can be identified using the tree-structuredness metric of \\citet{murty2023projections}. Overall, our work provides strong evidence that, with extended training, vanilla transformers discover and use hierarchical structure.","url_abs":"https://arxiv.org/abs/2305.18741v1","url_pdf":"https://arxiv.org/pdf/2305.18741v1.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":"grokking-of-hierarchical-structure-in-vanilla","repo_url":"https://github.com/murtyshikhar/structural-grokking","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2305.18741","atlas_url":"https://app.syntology.ai/?focus=2305.18741","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2305.18741"}},"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/murtyshikhar/structural-grokking","reach":null}],"summary":{"ran_honours":1,"unverified":3},"by_repo_kind":{"official":{"samples":4,"ran":1,"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":4,"samples":[{"code_sha256_prefix":"3e7de7b860103b81","entry":"get_norm","repo":"murtyshikhar/structural-grokking","repo_kind":"official","path":"transformer_lm_sci.py","file_url":"https://github.com/murtyshikhar/structural-grokking/blob/HEAD/transformer_lm_sci.py","link_basis":"first_harvest_node","language":"python","status":"ran_honours","verification_level":1,"contract_check":"HONOURS","metamorphic_tier":"well_formed","behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"3e7de7b860103b81"}},{"code_sha256_prefix":"68c87a2411e211e5","entry":"get_base_transformer_lm","repo":"murtyshikhar/structural-grokking","repo_kind":"official","path":"train_transformers.py","file_url":"https://github.com/murtyshikhar/structural-grokking/blob/HEAD/train_transformers.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"68c87a2411e211e5"}},{"code_sha256_prefix":"1227a753d0bd93d8","entry":"get_base_transformer_model","repo":"murtyshikhar/structural-grokking","repo_kind":"official","path":"train_transformers.py","file_url":"https://github.com/murtyshikhar/structural-grokking/blob/HEAD/train_transformers.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"1227a753d0bd93d8"}},{"code_sha256_prefix":"0b7a33664b0e5940","entry":"process","repo":"murtyshikhar/structural-grokking","repo_kind":"official","path":"transformer_lm_sci.py","file_url":"https://github.com/murtyshikhar/structural-grokking/blob/HEAD/transformer_lm_sci.py","link_basis":"first_harvest_node","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"NONE","inline_ok":false,"mcp_get_code":{"code_sha256":"0b7a33664b0e5940"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}