{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:75CWPNKQXMP5DM6EA6CIGINBNY","short_pith_number":"pith:75CWPNKQ","canonical_record":{"source":{"id":"2605.13652","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-13T15:11:37Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"b761ae48dffb740662929dc60450dcc022e1b609a0c695cc77bb0181259b9197","abstract_canon_sha256":"bbe06d6966ca0f1da8317ca4cb7edc912392e2916f0a5f55be6b1dc8db123c29"},"schema_version":"1.0"},"canonical_sha256":"ff4567b550bb1fd1b3c407848321a16e22e6c4cd7a2bbee5cd426b5aa1bdb53a","source":{"kind":"arxiv","id":"2605.13652","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.13652","created_at":"2026-05-18T02:44:17Z"},{"alias_kind":"arxiv_version","alias_value":"2605.13652v1","created_at":"2026-05-18T02:44:17Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13652","created_at":"2026-05-18T02:44:17Z"},{"alias_kind":"pith_short_12","alias_value":"75CWPNKQXMP5","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"75CWPNKQXMP5DM6E","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"75CWPNKQ","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:75CWPNKQXMP5DM6EA6CIGINBNY","target":"record","payload":{"canonical_record":{"source":{"id":"2605.13652","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-13T15:11:37Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"b761ae48dffb740662929dc60450dcc022e1b609a0c695cc77bb0181259b9197","abstract_canon_sha256":"bbe06d6966ca0f1da8317ca4cb7edc912392e2916f0a5f55be6b1dc8db123c29"},"schema_version":"1.0"},"canonical_sha256":"ff4567b550bb1fd1b3c407848321a16e22e6c4cd7a2bbee5cd426b5aa1bdb53a","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T02:44:17.449523Z","signature_b64":"BRJg/UquIqXSjv9zjkYVRyo8dCmeJyJiep99RRc5BPw9c6e8di3Rx039iR+vqOH6+MdH/+rt4CG3+8NQRycnCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ff4567b550bb1fd1b3c407848321a16e22e6c4cd7a2bbee5cd426b5aa1bdb53a","last_reissued_at":"2026-05-18T02:44:17.448889Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T02:44:17.448889Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.13652","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-18T02:44:17Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"YhVbeYzT/f16He98szV1jb3kvtWd5RX2pfBVK6X8cdfXVBpaOVFuo8Wzjj8e/tjtnPRXu4pP6SENWDawaWO2Aw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-01T23:46:28.655116Z"},"content_sha256":"74c24ed2ef8782a27fdfacb9f4a16e643d752a11e5003be2848ea9b540aa9fa9","schema_version":"1.0","event_id":"sha256:74c24ed2ef8782a27fdfacb9f4a16e643d752a11e5003be2848ea9b540aa9fa9"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:75CWPNKQXMP5DM6EA6CIGINBNY","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Beyond Perplexity: A Geometric and Spectral Study of Low-Rank Pre-Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Low-rank pre-training methods reach geometrically distinct loss basins than full-rank training even at matched perplexity.","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Anna Rumshisky, Namrata Shivagunde, Sherin Muckatira, Vijeta Deshpande","submitted_at":"2026-05-13T15:11:37Z","abstract_excerpt":"Pre-training large language models is dominated by the memory cost of storing full-rank weights, gradients, and optimizer states. Low-rank pre-training has emerged to address this, and the space of methods has grown rapidly. A central question remains open: do low-rank methods produce models that generalize comparably to full-rank training, or does the rank constraint fundamentally alter the solutions reached? Existing comparisons rely almost entirely on validation perplexity from single-seed runs, often carried forward from prior literature. Yet perplexity is a poor proxy for solution quality"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"We show that low-rank methods are not equivalent to full-rank training, nor to one another, even when validation perplexity is close. Full-rank training settles into a sharper basin than low-rank methods along random directions, while the reverse holds for the top-1 PCA direction.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the 16 metrics across 1-D loss landscape, interpolation, spectral structure, and activation similarity sufficiently characterize meaningful differences in solution quality, generalization, and downstream performance.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Low-rank pre-training methods converge to geometrically and spectrally distinct basins from full-rank training and from each other, even at similar validation perplexity.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Low-rank pre-training methods reach geometrically distinct loss basins than full-rank training even at matched perplexity.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"60bb05392f9d202279cbc3d85ee7309e1c5d2f0501b524da177fce0a21016081"},"source":{"id":"2605.13652","kind":"arxiv","version":1},"verdict":{"id":"f085b09f-888c-4faf-aa35-51a843a0f09e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T19:16:23.841652Z","strongest_claim":"We show that low-rank methods are not equivalent to full-rank training, nor to one another, even when validation perplexity is close. Full-rank training settles into a sharper basin than low-rank methods along random directions, while the reverse holds for the top-1 PCA direction.","one_line_summary":"Low-rank pre-training methods converge to geometrically and spectrally distinct basins from full-rank training and from each other, even at similar validation perplexity.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the 16 metrics across 1-D loss landscape, interpolation, spectral structure, and activation similarity sufficiently characterize meaningful differences in solution quality, generalization, and downstream performance.","pith_extraction_headline":"Low-rank pre-training methods reach geometrically distinct loss basins than full-rank training even at matched perplexity."},"references":{"count":30,"sample":[{"doi":"","year":2023,"title":"A modern look at the relationship between sharpness and generaliza- tion","work_id":"f4cfd80b-20d4-4ee1-9cff-8db4420b01b6","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Understanding pre-training and fine-tuning from loss landscape perspectives","work_id":"eea59928-7026-41ae-92a7-e32519a2cd03","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"Fira: Can we achieve full-rank training of llms under low-rank constraint?ArXiv, abs/2410.01623, 2024","work_id":"bc618a36-505b-4655-8771-ed9a465c53df","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2020,"title":"Linear mode connectivity and the lottery ticket hypothesis","work_id":"6d8c5789-2503-4bb7-ac1d-85f2d561f194","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2024,"title":"The language model evaluation harness, 07 2024","work_id":"9148854a-4b09-4c5d-b1a4-5ee9cccc772c","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":30,"snapshot_sha256":"efb7cf3025648a1cb0c0cbee46227db8e907fb559456449435169a5a7d965018","internal_anchors":1},"formal_canon":{"evidence_count":2,"snapshot_sha256":"10bee0cf672c63e78885db17e77a8090f4c3d7d443fd57aee1dcb8ea00c3794d"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"f085b09f-888c-4faf-aa35-51a843a0f09e"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-18T02:44:17Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"3AT3H0Vo2FleSD/Tvj1ZRfMW2fRqB1d/qIi1pffw7z2NKROCdwqq1lH6uUNp6Isza06t5dw6kHpN7NtjR8TmBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-01T23:46:28.655638Z"},"content_sha256":"0b8481f13740c0e6513e7810d706858c34effc3c52b1b648aa38ec5e066e8ff3","schema_version":"1.0","event_id":"sha256:0b8481f13740c0e6513e7810d706858c34effc3c52b1b648aa38ec5e066e8ff3"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/75CWPNKQXMP5DM6EA6CIGINBNY/bundle.json","state_url":"https://pith.science/pith/75CWPNKQXMP5DM6EA6CIGINBNY/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/75CWPNKQXMP5DM6EA6CIGINBNY/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-06-01T23:46:28Z","links":{"resolver":"https://pith.science/pith/75CWPNKQXMP5DM6EA6CIGINBNY","bundle":"https://pith.science/pith/75CWPNKQXMP5DM6EA6CIGINBNY/bundle.json","state":"https://pith.science/pith/75CWPNKQXMP5DM6EA6CIGINBNY/state.json","well_known_bundle":"https://pith.science/.well-known/pith/75CWPNKQXMP5DM6EA6CIGINBNY/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:75CWPNKQXMP5DM6EA6CIGINBNY","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":"bbe06d6966ca0f1da8317ca4cb7edc912392e2916f0a5f55be6b1dc8db123c29","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-13T15:11:37Z","title_canon_sha256":"b761ae48dffb740662929dc60450dcc022e1b609a0c695cc77bb0181259b9197"},"schema_version":"1.0","source":{"id":"2605.13652","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.13652","created_at":"2026-05-18T02:44:17Z"},{"alias_kind":"arxiv_version","alias_value":"2605.13652v1","created_at":"2026-05-18T02:44:17Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.13652","created_at":"2026-05-18T02:44:17Z"},{"alias_kind":"pith_short_12","alias_value":"75CWPNKQXMP5","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"75CWPNKQXMP5DM6E","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"75CWPNKQ","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:0b8481f13740c0e6513e7810d706858c34effc3c52b1b648aa38ec5e066e8ff3","target":"graph","created_at":"2026-05-18T02:44:17Z","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":"We show that low-rank methods are not equivalent to full-rank training, nor to one another, even when validation perplexity is close. Full-rank training settles into a sharper basin than low-rank methods along random directions, while the reverse holds for the top-1 PCA direction."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the 16 metrics across 1-D loss landscape, interpolation, spectral structure, and activation similarity sufficiently characterize meaningful differences in solution quality, generalization, and downstream performance."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Low-rank pre-training methods converge to geometrically and spectrally distinct basins from full-rank training and from each other, even at similar validation perplexity."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Low-rank pre-training methods reach geometrically distinct loss basins than full-rank training even at matched perplexity."}],"snapshot_sha256":"60bb05392f9d202279cbc3d85ee7309e1c5d2f0501b524da177fce0a21016081"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"10bee0cf672c63e78885db17e77a8090f4c3d7d443fd57aee1dcb8ea00c3794d"},"paper":{"abstract_excerpt":"Pre-training large language models is dominated by the memory cost of storing full-rank weights, gradients, and optimizer states. Low-rank pre-training has emerged to address this, and the space of methods has grown rapidly. A central question remains open: do low-rank methods produce models that generalize comparably to full-rank training, or does the rank constraint fundamentally alter the solutions reached? Existing comparisons rely almost entirely on validation perplexity from single-seed runs, often carried forward from prior literature. Yet perplexity is a poor proxy for solution quality","authors_text":"Anna Rumshisky, Namrata Shivagunde, Sherin Muckatira, Vijeta Deshpande","cross_cats":["cs.AI","cs.CL"],"headline":"Low-rank pre-training methods reach geometrically distinct loss basins than full-rank training even at matched perplexity.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-13T15:11:37Z","title":"Beyond Perplexity: A Geometric and Spectral Study of Low-Rank Pre-Training"},"references":{"count":30,"internal_anchors":1,"resolved_work":30,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"A modern look at the relationship between sharpness and generaliza- tion","work_id":"f4cfd80b-20d4-4ee1-9cff-8db4420b01b6","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Understanding pre-training and fine-tuning from loss landscape perspectives","work_id":"eea59928-7026-41ae-92a7-e32519a2cd03","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":3,"title":"Fira: Can we achieve full-rank training of llms under low-rank constraint?ArXiv, abs/2410.01623, 2024","work_id":"bc618a36-505b-4655-8771-ed9a465c53df","year":2024},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"Linear mode connectivity and the lottery ticket hypothesis","work_id":"6d8c5789-2503-4bb7-ac1d-85f2d561f194","year":2020},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"The language model evaluation harness, 07 2024","work_id":"9148854a-4b09-4c5d-b1a4-5ee9cccc772c","year":2024}],"snapshot_sha256":"efb7cf3025648a1cb0c0cbee46227db8e907fb559456449435169a5a7d965018"},"source":{"id":"2605.13652","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-14T19:16:23.841652Z","id":"f085b09f-888c-4faf-aa35-51a843a0f09e","model_set":{"reader":"grok-4.3"},"one_line_summary":"Low-rank pre-training methods converge to geometrically and spectrally distinct basins from full-rank training and from each other, even at similar validation perplexity.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Low-rank pre-training methods reach geometrically distinct loss basins than full-rank training even at matched perplexity.","strongest_claim":"We show that low-rank methods are not equivalent to full-rank training, nor to one another, even when validation perplexity is close. Full-rank training settles into a sharper basin than low-rank methods along random directions, while the reverse holds for the top-1 PCA direction.","weakest_assumption":"That the 16 metrics across 1-D loss landscape, interpolation, spectral structure, and activation similarity sufficiently characterize meaningful differences in solution quality, generalization, and downstream performance."}},"verdict_id":"f085b09f-888c-4faf-aa35-51a843a0f09e"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:74c24ed2ef8782a27fdfacb9f4a16e643d752a11e5003be2848ea9b540aa9fa9","target":"record","created_at":"2026-05-18T02:44:17Z","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":"bbe06d6966ca0f1da8317ca4cb7edc912392e2916f0a5f55be6b1dc8db123c29","cross_cats_sorted":["cs.AI","cs.CL"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-13T15:11:37Z","title_canon_sha256":"b761ae48dffb740662929dc60450dcc022e1b609a0c695cc77bb0181259b9197"},"schema_version":"1.0","source":{"id":"2605.13652","kind":"arxiv","version":1}},"canonical_sha256":"ff4567b550bb1fd1b3c407848321a16e22e6c4cd7a2bbee5cd426b5aa1bdb53a","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"ff4567b550bb1fd1b3c407848321a16e22e6c4cd7a2bbee5cd426b5aa1bdb53a","first_computed_at":"2026-05-18T02:44:17.448889Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-18T02:44:17.448889Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"BRJg/UquIqXSjv9zjkYVRyo8dCmeJyJiep99RRc5BPw9c6e8di3Rx039iR+vqOH6+MdH/+rt4CG3+8NQRycnCQ==","signature_status":"signed_v1","signed_at":"2026-05-18T02:44:17.449523Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.13652","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:74c24ed2ef8782a27fdfacb9f4a16e643d752a11e5003be2848ea9b540aa9fa9","sha256:0b8481f13740c0e6513e7810d706858c34effc3c52b1b648aa38ec5e066e8ff3"],"state_sha256":"c9e6b55eaef12d99b78a330fb955ec07018efece4d342936c3ae29430424df36"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"98yn1wq+rjyg2JgY1tZRSxE475IEHC1Fe/MQzD59OJtxMaJFN6m7DZcFyTgTDOHoq5GnSwv19jO3Y7te+wUXCQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-01T23:46:28.657920Z","bundle_sha256":"91858d00af08d1bd50fe47f2d968b5e69b033c6499c048e0651a6d30500915af"}}