{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:AN2ZG527QZYOPHYWDABJD46LLE","short_pith_number":"pith:AN2ZG527","schema_version":"1.0","canonical_sha256":"037593775f8670e79f16180291f3cb59239bb78d8f5f82dce06f1fd3935d67fc","source":{"kind":"arxiv","id":"2508.05628","version":1},"attestation_state":"computed","paper":{"title":"H-Net++: Hierarchical Dynamic Chunking for Tokenizer-Free Language Modelling in Morphologically-Rich Languages","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Mehrdad Zakershahrak, Samira Ghodratnama","submitted_at":"2025-08-07T17:59:01Z","abstract_excerpt":"Byte-level language models eliminate fragile tokenizers but face computational challenges in morphologically-rich languages (MRLs), where words span many bytes. We propose H-NET++, a hierarchical dynamic-chunking model that learns linguistically-informed segmentation through end-to-end training. Key innovations include: (1) a lightweight Transformer context-mixer (1.9M parameters) for cross-chunk attention, (2) a two-level latent hyper-prior for document-level consistency, (3) specialized handling of orthographic artifacts (e.g. Persian ZWNJ), and (4) curriculum-based training with staged sequ"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2508.05628","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-08-07T17:59:01Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"0266050de7328d0cf4ae83a65d649f219def29f99a9a4215daa5719a30ff5f79","abstract_canon_sha256":"99da4bbcf90ceda80e422a1f693df3ad30cab4923c582dc7b2b73ce3c939dcb3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:50:22.920939Z","signature_b64":"M+fc0npsLKO3oqGxOltayz4ZqQIPVnfnU9sVr6q/ixNjz3jfJ7qNhIL9I7kfy7Dul3mK6r0G1eSX5Py7HHbeDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"037593775f8670e79f16180291f3cb59239bb78d8f5f82dce06f1fd3935d67fc","last_reissued_at":"2026-07-05T11:50:22.920479Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:50:22.920479Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"H-Net++: Hierarchical Dynamic Chunking for Tokenizer-Free Language Modelling in Morphologically-Rich Languages","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Mehrdad Zakershahrak, Samira Ghodratnama","submitted_at":"2025-08-07T17:59:01Z","abstract_excerpt":"Byte-level language models eliminate fragile tokenizers but face computational challenges in morphologically-rich languages (MRLs), where words span many bytes. We propose H-NET++, a hierarchical dynamic-chunking model that learns linguistically-informed segmentation through end-to-end training. Key innovations include: (1) a lightweight Transformer context-mixer (1.9M parameters) for cross-chunk attention, (2) a two-level latent hyper-prior for document-level consistency, (3) specialized handling of orthographic artifacts (e.g. Persian ZWNJ), and (4) curriculum-based training with staged sequ"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.05628","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2508.05628/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2508.05628","created_at":"2026-07-05T11:50:22.920540+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.05628v1","created_at":"2026-07-05T11:50:22.920540+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.05628","created_at":"2026-07-05T11:50:22.920540+00:00"},{"alias_kind":"pith_short_12","alias_value":"AN2ZG527QZYO","created_at":"2026-07-05T11:50:22.920540+00:00"},{"alias_kind":"pith_short_16","alias_value":"AN2ZG527QZYOPHYW","created_at":"2026-07-05T11:50:22.920540+00:00"},{"alias_kind":"pith_short_8","alias_value":"AN2ZG527","created_at":"2026-07-05T11:50:22.920540+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.30080","citing_title":"Adaptive Targeted Dynamic Chunking for Tokenization-Free Hierarchical Model","ref_index":18,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AN2ZG527QZYOPHYWDABJD46LLE","json":"https://pith.science/pith/AN2ZG527QZYOPHYWDABJD46LLE.json","graph_json":"https://pith.science/api/pith-number/AN2ZG527QZYOPHYWDABJD46LLE/graph.json","events_json":"https://pith.science/api/pith-number/AN2ZG527QZYOPHYWDABJD46LLE/events.json","paper":"https://pith.science/paper/AN2ZG527"},"agent_actions":{"view_html":"https://pith.science/pith/AN2ZG527QZYOPHYWDABJD46LLE","download_json":"https://pith.science/pith/AN2ZG527QZYOPHYWDABJD46LLE.json","view_paper":"https://pith.science/paper/AN2ZG527","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.05628&json=true","fetch_graph":"https://pith.science/api/pith-number/AN2ZG527QZYOPHYWDABJD46LLE/graph.json","fetch_events":"https://pith.science/api/pith-number/AN2ZG527QZYOPHYWDABJD46LLE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AN2ZG527QZYOPHYWDABJD46LLE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AN2ZG527QZYOPHYWDABJD46LLE/action/storage_attestation","attest_author":"https://pith.science/pith/AN2ZG527QZYOPHYWDABJD46LLE/action/author_attestation","sign_citation":"https://pith.science/pith/AN2ZG527QZYOPHYWDABJD46LLE/action/citation_signature","submit_replication":"https://pith.science/pith/AN2ZG527QZYOPHYWDABJD46LLE/action/replication_record"}},"created_at":"2026-07-05T11:50:22.920540+00:00","updated_at":"2026-07-05T11:50:22.920540+00:00"}