{"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/arxiv-2601-02031","title":"Output Embedding Centering for Stable LLM Pretraining","arxiv_id":"2601.02031","date":"2026-01-05","proceeding":null,"authors":["Felix Stollenwerk","Anna Lokrantz","Niclas Hertzberg"],"abstract":"Pretraining of large language models is not only expensive but also prone to certain training instabilities. A specific instability that often occurs at the end of training is output logit divergence. The most widely used mitigation strategies, z-loss and logit soft-capping, merely address the symptoms rather than the underlying cause of the problem. In this paper, we analyze the instability from the perspective of the output embeddings' geometry and identify anisotropic embeddings as its source. Based on this, we propose output embedding centering (OEC) as a new mitigation strategy, and demonstrate that it suppresses output logit divergence. OEC can be implemented in two different ways: as a deterministic operation called $μ$-centering, or a regularization method called $μ$-loss. Our experiments show that both variants outperform z-loss in terms of training stability, while being on par with logit soft-capping. This holds true both in the presence and the absence of weight tying. As a secondary result, we find that $μ$-loss is significantly less sensitive to regularization hyperparameter tuning than z-loss.","url_abs":"https://arxiv.org/abs/2601.02031","url_pdf":"https://arxiv.org/pdf/2601.02031","source":{"archive":null,"snapshot":"2025-07-28","note":"not in the Papers with Code archive (frozen at the snapshot)","row_kind":"graph","title_abstract_authors_date":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)"},"code_links":[],"tasks":[],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2601.02031","atlas_url":"https://app.syntology.ai/?focus=2601.02031","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2601.02031"}},"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":[{"mentioned_in_github":null,"is_official":null,"provenance":"deterministic:regex_extraction","mentioned_in_paper":null,"url":"https://github.com/flxst/output-embedding-centering","reach":{"status":"ok"}}],"summary":{"unverified":15},"by_repo_kind":{"found_in_text":{"samples":15,"ran":0,"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":15,"samples":[{"code_sha256_prefix":"f1a8e6c5cfd036a4","entry":"compute_average_acc","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"nanoGPT/analysis/norms.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/nanoGPT/analysis/norms.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":"f1a8e6c5cfd036a4"}},{"code_sha256_prefix":"746bf8b091dcf89b","entry":"extract_embeddings","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"nanoGPT/analysis/embeddings.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/nanoGPT/analysis/embeddings.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":"746bf8b091dcf89b"}},{"code_sha256_prefix":"897588fa0f3168dc","entry":"extract_quantity_wortsman","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"results/plots.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/results/plots.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":"897588fa0f3168dc"}},{"code_sha256_prefix":"9ac7e32856158250","entry":"get_global_batch_size","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"nanoGPT/config/helper_functions.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/nanoGPT/config/helper_functions.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":"9ac7e32856158250"}},{"code_sha256_prefix":"78ae4a6f3dd55d19","entry":"get_hardcoded_loss","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"nanoGPT/analysis/script_analyze.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/nanoGPT/analysis/script_analyze.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":"78ae4a6f3dd55d19"}},{"code_sha256_prefix":"956b879e10f36c60","entry":"get_input_embeddings_torch","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"nanoGPT/analysis/embeddings.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/nanoGPT/analysis/embeddings.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":"956b879e10f36c60"}},{"code_sha256_prefix":"973304852624bd51","entry":"get_label","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"results/plots.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/results/plots.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":"973304852624bd51"}},{"code_sha256_prefix":"2fb3e1f3ddc2017c","entry":"get_last_checkpoint","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"nanoGPT/config/script_create_config_val.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/nanoGPT/config/script_create_config_val.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":"2fb3e1f3ddc2017c"}},{"code_sha256_prefix":"865e92f2beff05ab","entry":"get_losses","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"nanoGPT/analysis/script_analyze.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/nanoGPT/analysis/script_analyze.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":"865e92f2beff05ab"}},{"code_sha256_prefix":"478a0eb4deb0c02c","entry":"get_prefix","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"nanoGPT/config/helper_functions.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/nanoGPT/config/helper_functions.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":"478a0eb4deb0c02c"}},{"code_sha256_prefix":"ca2fd1f5b0990a35","entry":"get_section_experiment","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"nanoGPT/config/script_create_config.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/nanoGPT/config/script_create_config.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":"ca2fd1f5b0990a35"}},{"code_sha256_prefix":"ce30c5f3c3ea52e0","entry":"get_section_types","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"nanoGPT/config/script_create_config.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/nanoGPT/config/script_create_config.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":"ce30c5f3c3ea52e0"}},{"code_sha256_prefix":"5b50be5fb856492f","entry":"get_test_losses_and_perplexities","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"nanoGPT/analysis/script_analyze.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/nanoGPT/analysis/script_analyze.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":"5b50be5fb856492f"}},{"code_sha256_prefix":"49b57d61738877be","entry":"isotropy_run","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"nanoGPT/analysis/isotropy.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/nanoGPT/analysis/isotropy.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":"49b57d61738877be"}},{"code_sha256_prefix":"238a79e0b798853f","entry":"load_isotropy","repo":"flxst/output-embedding-centering","repo_kind":"found_in_text","path":"nanoGPT/analysis/norms.py","file_url":"https://github.com/flxst/output-embedding-centering/blob/HEAD/nanoGPT/analysis/norms.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":"238a79e0b798853f"}}]},"arxiv_metadata":{"licence":"arXiv metadata, CC0 1.0 (https://info.arxiv.org/help/license)","fields":["title","abstract","authors","date"],"primary_category":"cs.LG","source":"arxiv_2026.jsonl"},"syntology_extracted_results":{"kind":"leaderboard_placements","source":"Syntology's leaderboard-shaped extractor over the paper's own arXiv-HTML tables: a model pointed at a cell, the number was read from that cell and checked against the board's metric, dataset, split and scale, and an independent check accepted the entry; not reviewed by the paper's authors or the archive's editors","extractor_model":"global.anthropic.claude-sonnet-4-5-20250929-v1:0","verifier_model":null,"prompt_sha":"fa63d4bb9d755694","coverage":{"sentence":"Syntology has checked 6,885 of the 9,623 papers on this site that are newer than the archive; results from the others appear after they are checked.","papers_newer_than_archive":9623,"papers_checked":6885},"entries":[],"not_placed":{"boards":0,"rejected_by_independent_check":0,"refused_by_a_rule":0,"check_did_not_answer":0,"proposed_without_a_cell":0,"declined_by_site":0}}}