{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:QUPZXDZBDFCL7W52IZH4EGNZPT","short_pith_number":"pith:QUPZXDZB","schema_version":"1.0","canonical_sha256":"851f9b8f211944bfdbba464fc219b97cfd89c2bcc04ebc179f14f5b4d3c98efa","source":{"kind":"arxiv","id":"2503.22547","version":1},"attestation_state":"computed","paper":{"title":"Bridging the Dimensional Chasm: Uncover Layer-wise Dimensional Reduction in Transformers through Token Correlation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Hua Xing Zhu, Ming-Xing Luo, Qing-Hong Cao, Zeyu Li, Zhuo-Yang Song","submitted_at":"2025-03-28T15:47:30Z","abstract_excerpt":"The geometric evolution of token representations in large language models (LLMs) presents a fundamental paradox: while human language inherently organizes semantic information in low-dimensional spaces ($\\sim 10^1$ dimensions), modern LLMs employ high-dimensional embeddings ($\\sim 10^3$ dimensions) processed through Transformer architectures. To resolve this paradox, this work bridges this conceptual gap by developing a geometric framework that tracks token dynamics across Transformers layers. Through layer-wise analysis of intrinsic dimensions across multiple architectures, we reveal an expan"},"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":"2503.22547","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-03-28T15:47:30Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"0b43fa3f4be38eed13471538e452b05dc6e6cc92c1b00af4eb84114476e38f1d","abstract_canon_sha256":"d242cd08809757e96cbdff69952edf26a945bbbe4c0ec76cc4e07c77395ca1ec"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:41:06.570576Z","signature_b64":"eif1oahCX5nLXNBTwV09SxrJVhhGrORMkLIgGcVf8At6bB04sMpKPrjkFhVK1tcON6QzrZ8sWsbW0dNYOckDBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"851f9b8f211944bfdbba464fc219b97cfd89c2bcc04ebc179f14f5b4d3c98efa","last_reissued_at":"2026-07-05T10:41:06.570164Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:41:06.570164Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Bridging the Dimensional Chasm: Uncover Layer-wise Dimensional Reduction in Transformers through Token Correlation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Hua Xing Zhu, Ming-Xing Luo, Qing-Hong Cao, Zeyu Li, Zhuo-Yang Song","submitted_at":"2025-03-28T15:47:30Z","abstract_excerpt":"The geometric evolution of token representations in large language models (LLMs) presents a fundamental paradox: while human language inherently organizes semantic information in low-dimensional spaces ($\\sim 10^1$ dimensions), modern LLMs employ high-dimensional embeddings ($\\sim 10^3$ dimensions) processed through Transformer architectures. To resolve this paradox, this work bridges this conceptual gap by developing a geometric framework that tracks token dynamics across Transformers layers. Through layer-wise analysis of intrinsic dimensions across multiple architectures, we reveal an expan"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.22547","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/2503.22547/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":"2503.22547","created_at":"2026-07-05T10:41:06.570221+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.22547v1","created_at":"2026-07-05T10:41:06.570221+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.22547","created_at":"2026-07-05T10:41:06.570221+00:00"},{"alias_kind":"pith_short_12","alias_value":"QUPZXDZBDFCL","created_at":"2026-07-05T10:41:06.570221+00:00"},{"alias_kind":"pith_short_16","alias_value":"QUPZXDZBDFCL7W52","created_at":"2026-07-05T10:41:06.570221+00:00"},{"alias_kind":"pith_short_8","alias_value":"QUPZXDZB","created_at":"2026-07-05T10:41:06.570221+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.30813","citing_title":"Gradient Smoothing: Coupling Layer-wise Updates for Improved Optimization","ref_index":31,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QUPZXDZBDFCL7W52IZH4EGNZPT","json":"https://pith.science/pith/QUPZXDZBDFCL7W52IZH4EGNZPT.json","graph_json":"https://pith.science/api/pith-number/QUPZXDZBDFCL7W52IZH4EGNZPT/graph.json","events_json":"https://pith.science/api/pith-number/QUPZXDZBDFCL7W52IZH4EGNZPT/events.json","paper":"https://pith.science/paper/QUPZXDZB"},"agent_actions":{"view_html":"https://pith.science/pith/QUPZXDZBDFCL7W52IZH4EGNZPT","download_json":"https://pith.science/pith/QUPZXDZBDFCL7W52IZH4EGNZPT.json","view_paper":"https://pith.science/paper/QUPZXDZB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.22547&json=true","fetch_graph":"https://pith.science/api/pith-number/QUPZXDZBDFCL7W52IZH4EGNZPT/graph.json","fetch_events":"https://pith.science/api/pith-number/QUPZXDZBDFCL7W52IZH4EGNZPT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QUPZXDZBDFCL7W52IZH4EGNZPT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QUPZXDZBDFCL7W52IZH4EGNZPT/action/storage_attestation","attest_author":"https://pith.science/pith/QUPZXDZBDFCL7W52IZH4EGNZPT/action/author_attestation","sign_citation":"https://pith.science/pith/QUPZXDZBDFCL7W52IZH4EGNZPT/action/citation_signature","submit_replication":"https://pith.science/pith/QUPZXDZBDFCL7W52IZH4EGNZPT/action/replication_record"}},"created_at":"2026-07-05T10:41:06.570221+00:00","updated_at":"2026-07-05T10:41:06.570221+00:00"}