{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:RWT45WHEDQ2T77UZTST5ALRBRS","short_pith_number":"pith:RWT45WHE","canonical_record":{"source":{"id":"2605.03229","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-04T23:46:40Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"a9199ad6290c362e71d7ebdeb0f59de454c47aeaa7e76e814a1cd0e7cca22e0a","abstract_canon_sha256":"f5f1f66b2cd960f80551f3f7b38a468e1641e83de9f26e9393a906e31f0a5c58"},"schema_version":"1.0"},"canonical_sha256":"8da7ced8e41c353ffe999ca7d02e218cb8181a55c59ecfcb1bb850111953d964","source":{"kind":"arxiv","id":"2605.03229","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.03229","created_at":"2026-06-09T02:08:43Z"},{"alias_kind":"arxiv_version","alias_value":"2605.03229v2","created_at":"2026-06-09T02:08:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.03229","created_at":"2026-06-09T02:08:43Z"},{"alias_kind":"pith_short_12","alias_value":"RWT45WHEDQ2T","created_at":"2026-06-09T02:08:43Z"},{"alias_kind":"pith_short_16","alias_value":"RWT45WHEDQ2T77UZ","created_at":"2026-06-09T02:08:43Z"},{"alias_kind":"pith_short_8","alias_value":"RWT45WHE","created_at":"2026-06-09T02:08:43Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:RWT45WHEDQ2T77UZTST5ALRBRS","target":"record","payload":{"canonical_record":{"source":{"id":"2605.03229","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-04T23:46:40Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"a9199ad6290c362e71d7ebdeb0f59de454c47aeaa7e76e814a1cd0e7cca22e0a","abstract_canon_sha256":"f5f1f66b2cd960f80551f3f7b38a468e1641e83de9f26e9393a906e31f0a5c58"},"schema_version":"1.0"},"canonical_sha256":"8da7ced8e41c353ffe999ca7d02e218cb8181a55c59ecfcb1bb850111953d964","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T02:08:43.619774Z","signature_b64":"E7ClZE8pNBaS461kdtsvkZ1FCIIkRjPpnQqPaZaQVLpIBebpGsUwFz/JJuND7C2J0QyQ+IaNqm80UI1njlwIDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8da7ced8e41c353ffe999ca7d02e218cb8181a55c59ecfcb1bb850111953d964","last_reissued_at":"2026-06-09T02:08:43.618522Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T02:08:43.618522Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.03229","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-09T02:08:43Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"U87fIq4uQzlPEr+RujBADnTcsuFhGWO8n+VdD9QVzG32qo1YIgAOYbMN/la9MsjEDT+Gd28TtG8kaU5HzvFDAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-09T21:37:55.799050Z"},"content_sha256":"da83d8c00576a92d730a4a5e4d43ae7bb24d496ba3fddc2cea82f539ece91a32","schema_version":"1.0","event_id":"sha256:da83d8c00576a92d730a4a5e4d43ae7bb24d496ba3fddc2cea82f539ece91a32"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:RWT45WHEDQ2T77UZTST5ALRBRS","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Sparse Memory Finetuning as a Low-Forgetting Alternative to LoRA and Full Finetuning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Sparse Memory Finetuning updates only the most heavily read rows in added key-value layers to gain task performance while preserving general capabilities better than LoRA or full finetuning.","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Anirudh Kanchi, Garv Shah, Prakhar Gupta, Satyam Goyal","submitted_at":"2026-05-04T23:46:40Z","abstract_excerpt":"Adapting a pretrained language model to a new task often hurts the general capabilities it already had, a problem known as catastrophic forgetting. Sparse Memory Finetuning (SMF) tries to avoid this by adding key-value memory layers to the model and, on each training step, updating only the small set of memory rows that the current batch reads most heavily. We re-implement SMF on Qwen-2.5-0.5B-Instruct and compare it with LoRA and full finetuning on MedMCQA, a 4-choice medical exam task, using WikiText perplexity and TriviaQA accuracy as forgetting probes. SMF improves MedMCQA by 2.5 percentag"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"SMF improves MedMCQA by 2.5 percentage points while keeping both forgetting probes within roughly 1 point of the base model, whereas LoRA and full finetuning achieve larger gains but with clear drift on both.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That selectively updating only the most heavily read memory rows during training is sufficient to acquire new task knowledge without unintended interference in unrelated general capabilities.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SMF improves MedMCQA accuracy by 2.5 points while keeping WikiText perplexity and TriviaQA accuracy within 1 point of the base model, outperforming LoRA and full finetuning on forgetting metrics.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Sparse Memory Finetuning updates only the most heavily read rows in added key-value layers to gain task performance while preserving general capabilities better than LoRA or full finetuning.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"f936b72255b9524f1282e03c8b02357a9ae1149b2ebc7e71d4028d621d89a9c2"},"source":{"id":"2605.03229","kind":"arxiv","version":2},"verdict":{"id":"919f7bab-02fa-49fa-9ecc-267ca47dbda6","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T17:47:39.561915Z","strongest_claim":"SMF improves MedMCQA by 2.5 percentage points while keeping both forgetting probes within roughly 1 point of the base model, whereas LoRA and full finetuning achieve larger gains but with clear drift on both.","one_line_summary":"SMF improves MedMCQA accuracy by 2.5 points while keeping WikiText perplexity and TriviaQA accuracy within 1 point of the base model, outperforming LoRA and full finetuning on forgetting metrics.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That selectively updating only the most heavily read memory rows during training is sufficient to acquire new task knowledge without unintended interference in unrelated general capabilities.","pith_extraction_headline":"Sparse Memory Finetuning updates only the most heavily read rows in added key-value layers to gain task performance while preserving general capabilities better than LoRA or full finetuning."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.03229/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-20T14:35:19.247524Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_title_agreement","ran_at":"2026-05-20T01:31:21.794774Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T15:34:04.280909Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"6dc37e4ce03a15165ed85bcb415ba044c98eea186f08319dd73876c48ed8e5a9"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":2,"snapshot_sha256":"58dce1ef058b469a5bb57b12b9a65cce0e644b5409789449ba5ae62c061a9f55"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"919f7bab-02fa-49fa-9ecc-267ca47dbda6"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-09T02:08:43Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ASo2evOFVuXP+uxHvNQ4y4DzrB5wBTxfpHsBotxx1he6gqrqUeeLhjlBzTyD15uurP2SFz5z5YU4yDkxx6ucBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-09T21:37:55.800169Z"},"content_sha256":"3bf56f2ae93d84ddd0cf9451e36765314acd536212e0af1f1440f643f3d36848","schema_version":"1.0","event_id":"sha256:3bf56f2ae93d84ddd0cf9451e36765314acd536212e0af1f1440f643f3d36848"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/RWT45WHEDQ2T77UZTST5ALRBRS/bundle.json","state_url":"https://pith.science/pith/RWT45WHEDQ2T77UZTST5ALRBRS/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/RWT45WHEDQ2T77UZTST5ALRBRS/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":"2026-06-09T21:37:55Z","links":{"resolver":"https://pith.science/pith/RWT45WHEDQ2T77UZTST5ALRBRS","bundle":"https://pith.science/pith/RWT45WHEDQ2T77UZTST5ALRBRS/bundle.json","state":"https://pith.science/pith/RWT45WHEDQ2T77UZTST5ALRBRS/state.json","well_known_bundle":"https://pith.science/.well-known/pith/RWT45WHEDQ2T77UZTST5ALRBRS/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:RWT45WHEDQ2T77UZTST5ALRBRS","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":"f5f1f66b2cd960f80551f3f7b38a468e1641e83de9f26e9393a906e31f0a5c58","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-04T23:46:40Z","title_canon_sha256":"a9199ad6290c362e71d7ebdeb0f59de454c47aeaa7e76e814a1cd0e7cca22e0a"},"schema_version":"1.0","source":{"id":"2605.03229","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.03229","created_at":"2026-06-09T02:08:43Z"},{"alias_kind":"arxiv_version","alias_value":"2605.03229v2","created_at":"2026-06-09T02:08:43Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.03229","created_at":"2026-06-09T02:08:43Z"},{"alias_kind":"pith_short_12","alias_value":"RWT45WHEDQ2T","created_at":"2026-06-09T02:08:43Z"},{"alias_kind":"pith_short_16","alias_value":"RWT45WHEDQ2T77UZ","created_at":"2026-06-09T02:08:43Z"},{"alias_kind":"pith_short_8","alias_value":"RWT45WHE","created_at":"2026-06-09T02:08:43Z"}],"graph_snapshots":[{"event_id":"sha256:3bf56f2ae93d84ddd0cf9451e36765314acd536212e0af1f1440f643f3d36848","target":"graph","created_at":"2026-06-09T02:08:43Z","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":"SMF improves MedMCQA by 2.5 percentage points while keeping both forgetting probes within roughly 1 point of the base model, whereas LoRA and full finetuning achieve larger gains but with clear drift on both."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That selectively updating only the most heavily read memory rows during training is sufficient to acquire new task knowledge without unintended interference in unrelated general capabilities."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"SMF improves MedMCQA accuracy by 2.5 points while keeping WikiText perplexity and TriviaQA accuracy within 1 point of the base model, outperforming LoRA and full finetuning on forgetting metrics."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Sparse Memory Finetuning updates only the most heavily read rows in added key-value layers to gain task performance while preserving general capabilities better than LoRA or full finetuning."}],"snapshot_sha256":"f936b72255b9524f1282e03c8b02357a9ae1149b2ebc7e71d4028d621d89a9c2"},"formal_canon":{"evidence_count":2,"snapshot_sha256":"58dce1ef058b469a5bb57b12b9a65cce0e644b5409789449ba5ae62c061a9f55"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-20T14:35:19.247524Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-20T01:31:21.794774Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T15:34:04.280909Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.03229/integrity.json","findings":[],"snapshot_sha256":"6dc37e4ce03a15165ed85bcb415ba044c98eea186f08319dd73876c48ed8e5a9","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Adapting a pretrained language model to a new task often hurts the general capabilities it already had, a problem known as catastrophic forgetting. Sparse Memory Finetuning (SMF) tries to avoid this by adding key-value memory layers to the model and, on each training step, updating only the small set of memory rows that the current batch reads most heavily. We re-implement SMF on Qwen-2.5-0.5B-Instruct and compare it with LoRA and full finetuning on MedMCQA, a 4-choice medical exam task, using WikiText perplexity and TriviaQA accuracy as forgetting probes. SMF improves MedMCQA by 2.5 percentag","authors_text":"Anirudh Kanchi, Garv Shah, Prakhar Gupta, Satyam Goyal","cross_cats":["cs.LG"],"headline":"Sparse Memory Finetuning updates only the most heavily read rows in added key-value layers to gain task performance while preserving general capabilities better than LoRA or full finetuning.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-04T23:46:40Z","title":"Sparse Memory Finetuning as a Low-Forgetting Alternative to LoRA and Full Finetuning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.03229","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-08T17:47:39.561915Z","id":"919f7bab-02fa-49fa-9ecc-267ca47dbda6","model_set":{"reader":"grok-4.3"},"one_line_summary":"SMF improves MedMCQA accuracy by 2.5 points while keeping WikiText perplexity and TriviaQA accuracy within 1 point of the base model, outperforming LoRA and full finetuning on forgetting metrics.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Sparse Memory Finetuning updates only the most heavily read rows in added key-value layers to gain task performance while preserving general capabilities better than LoRA or full finetuning.","strongest_claim":"SMF improves MedMCQA by 2.5 percentage points while keeping both forgetting probes within roughly 1 point of the base model, whereas LoRA and full finetuning achieve larger gains but with clear drift on both.","weakest_assumption":"That selectively updating only the most heavily read memory rows during training is sufficient to acquire new task knowledge without unintended interference in unrelated general capabilities."}},"verdict_id":"919f7bab-02fa-49fa-9ecc-267ca47dbda6"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:da83d8c00576a92d730a4a5e4d43ae7bb24d496ba3fddc2cea82f539ece91a32","target":"record","created_at":"2026-06-09T02:08:43Z","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":"f5f1f66b2cd960f80551f3f7b38a468e1641e83de9f26e9393a906e31f0a5c58","cross_cats_sorted":["cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-05-04T23:46:40Z","title_canon_sha256":"a9199ad6290c362e71d7ebdeb0f59de454c47aeaa7e76e814a1cd0e7cca22e0a"},"schema_version":"1.0","source":{"id":"2605.03229","kind":"arxiv","version":2}},"canonical_sha256":"8da7ced8e41c353ffe999ca7d02e218cb8181a55c59ecfcb1bb850111953d964","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"8da7ced8e41c353ffe999ca7d02e218cb8181a55c59ecfcb1bb850111953d964","first_computed_at":"2026-06-09T02:08:43.618522Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-09T02:08:43.618522Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"E7ClZE8pNBaS461kdtsvkZ1FCIIkRjPpnQqPaZaQVLpIBebpGsUwFz/JJuND7C2J0QyQ+IaNqm80UI1njlwIDQ==","signature_status":"signed_v1","signed_at":"2026-06-09T02:08:43.619774Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.03229","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:da83d8c00576a92d730a4a5e4d43ae7bb24d496ba3fddc2cea82f539ece91a32","sha256:3bf56f2ae93d84ddd0cf9451e36765314acd536212e0af1f1440f643f3d36848"],"state_sha256":"448e402645d972e082470f7a0e527fa03e9ebd61e2a5a12d022153ab505a4e10"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"EUw8ZKjF76/fo1Nmt+Ugx/Um2irrVOY4657VMv+eWcqomMg7ATqcrAKzOXN8B/Ouc4TCPZHiJksb+wBc0mdDAg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-09T21:37:55.805267Z","bundle_sha256":"601f641ef5a5da85d98a3da1b1a1bd6881c09b0a45b329c9beebe13c06df0e58"}}