{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:ERBMSKCVHF5ETVX2DJW5NHOQD3","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"95cf7dcc087fa1067e2a6e12016390653396469c8ef71be08c5541b0a847e0e6","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-13T12:31:04Z","title_canon_sha256":"fe3fc4f0fe3e5638c42b357229d5dddc318e3b413ba0ca2cab87bff14dadd35a"},"schema_version":"1.0","source":{"id":"2605.13436","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.13436","created_at":"2026-05-18T02:44:47Z"},{"alias_kind":"arxiv_version","alias_value":"2605.13436v1","created_at":"2026-05-18T02:44:47Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13436","created_at":"2026-05-18T02:44:47Z"},{"alias_kind":"pith_short_12","alias_value":"ERBMSKCVHF5E","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"ERBMSKCVHF5ETVX2","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"ERBMSKCV","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:41ba1eb446413de618e03fd0d655bc0d7fe5c10734bacdd05ee8ea9c5fc6dfba","target":"graph","created_at":"2026-05-18T02:44:47Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"Across tasks, the best results are typically obtained when stochastic tokenization is applied during both pretraining and fine-tuning, whereas applying BPE dropout only during fine-tuning can underperform deterministic tokenization in smaller-data settings."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the downsampled subsets of high-resource languages and the chosen evaluation tasks sufficiently represent truly low-resource scenarios, and that the modest morphological alignment gains explain the performance benefits."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Stochastic tokenization with BPE dropout during both pretraining and fine-tuning outperforms deterministic tokenization or fine-tuning-only dropout on low-resource NLP tasks."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Stochastic tokenization during both pretraining and fine-tuning yields the best results in low-resource NLP tasks."}],"snapshot_sha256":"c79f1748cd71246fa7fa1d7261e3aa5ce914459802ef982c38de64bce6e8276d"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Subword regularization methods such as BPE dropout are typically applied only during fine-tuning, while pretraining is usually done with deterministic tokenization. This creates a potential segmentation mismatch between pretraining and fine-tuning. We investigate whether applying BPE dropout during pretraining improves downstream performance in low-resource NLP. We train monolingual and bilingual BERT models on downsampled subsets of English, German, French, Spanish, Kiswahili, and isiXhosa, and evaluate them on XNLI, PAWS-X, PAN-X, and MasakhaNER 2.0. Across tasks, the best results are typica","authors_text":"Marcel Dunaiski, Ruan Visser, Trienko Grobler","cross_cats":["cs.LG"],"headline":"Stochastic tokenization during both pretraining and fine-tuning yields the best results in low-resource NLP tasks.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-13T12:31:04Z","title":"Pretraining Language Models with Subword Regularization: An Empirical Study of BPE Dropout in Low-Resource NLP"},"references":{"count":34,"internal_anchors":1,"resolved_work":34,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"[Adelaniet al., 2022 ] David Ifeoluwa Adelani, Graham Neubig, Sebastian Ruder, Shruti Rijhwani, Michael Beuk- man, Chester Palen-Michel, Constantine Lignos, Je- sujoba O. Alabi, Shamsuddeen H. Muhamma","work_id":"67c4ec17-bbd8-4240-bd97-3ed5e3b53551","year":2022},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"[Arnett and Bergen, 2025] Catherine Arnett and Ben- jamin K","work_id":"04ee1166-7fe1-4f96-9b36-381954561a32","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Morphynet: A large multilin- gual database of derivational and inflectional morphol- ogy","work_id":"f48a1b4e-0a78-4d41-889d-14a5f4d316dc","year":2021},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"BPE-knockout: Pruning pre-existing BPE tokenisers with backwards-compatible morpho- logical semi-supervision","work_id":"12aa29ab-7de1-4364-a9c4-963a01b0bb96","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"[Cognettaet al., 2024 ] Marco Cognetta, Vil ´em Zouhar, and Naoaki Okazaki","work_id":"a9e124dd-e469-477d-94f1-f97b809339de","year":2024}],"snapshot_sha256":"7a67ea21a534b828b28cca2517bca675cafe6d48ba93e209a7f06cd6148f3bbc"},"source":{"id":"2605.13436","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-14T19:25:56.908331Z","id":"aab51e3e-c161-4e4d-8984-7eb1b006def7","model_set":{"reader":"grok-4.3"},"one_line_summary":"Stochastic tokenization with BPE dropout during both pretraining and fine-tuning outperforms deterministic tokenization or fine-tuning-only dropout on low-resource NLP tasks.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Stochastic tokenization during both pretraining and fine-tuning yields the best results in low-resource NLP tasks.","strongest_claim":"Across tasks, the best results are typically obtained when stochastic tokenization is applied during both pretraining and fine-tuning, whereas applying BPE dropout only during fine-tuning can underperform deterministic tokenization in smaller-data settings.","weakest_assumption":"That the downsampled subsets of high-resource languages and the chosen evaluation tasks sufficiently represent truly low-resource scenarios, and that the modest morphological alignment gains explain the performance benefits."}},"verdict_id":"aab51e3e-c161-4e4d-8984-7eb1b006def7"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:904e2c99990256d6ac4bb2ca9896e735304a0f5813947e5eed7b1461fa8c8370","target":"record","created_at":"2026-05-18T02:44:47Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"95cf7dcc087fa1067e2a6e12016390653396469c8ef71be08c5541b0a847e0e6","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-13T12:31:04Z","title_canon_sha256":"fe3fc4f0fe3e5638c42b357229d5dddc318e3b413ba0ca2cab87bff14dadd35a"},"schema_version":"1.0","source":{"id":"2605.13436","kind":"arxiv","version":1}},"canonical_sha256":"2442c92855397a49d6fa1a6dd69dd01ed208fffb1703e0dd7cb1e5486fb175bb","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2442c92855397a49d6fa1a6dd69dd01ed208fffb1703e0dd7cb1e5486fb175bb","first_computed_at":"2026-05-18T02:44:47.101101Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T02:44:47.101101Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"JD4J6P/l1VAlJ8JdqQ9o3dr6/hWGrRFHDIHRijlT/JHPhmMHyyzTfHeAbkTpZOi55mb4VR8wwC/HqNAjN+26CA==","signature_status":"signed_v1","signed_at":"2026-05-18T02:44:47.101528Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.13436","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:904e2c99990256d6ac4bb2ca9896e735304a0f5813947e5eed7b1461fa8c8370","sha256:41ba1eb446413de618e03fd0d655bc0d7fe5c10734bacdd05ee8ea9c5fc6dfba"],"state_sha256":"9d7eb7a07e9b40b8bafbb007bc15fc1ff2cd97501418456c27762efc5bcd84c2"}