{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:7746NDQ6F2TDP2S5PQXIAQW2UO","short_pith_number":"pith:7746NDQ6","canonical_record":{"source":{"id":"2605.26097","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-25T17:54:34Z","cross_cats_sorted":[],"title_canon_sha256":"fbbf9a1f048801e6cd8acc4dff69234c8eeb51b71a85b48878d8cfebc32c7de7","abstract_canon_sha256":"aa04cf1a381db67298a6324fe3c739ee5b8006e18dde441ec7e0507abc123f78"},"schema_version":"1.0"},"canonical_sha256":"fff9e68e1e2ea637ea5d7c2e8042daa3be8b76c1275bbdb77ec6ac9a4ac7be45","source":{"kind":"arxiv","id":"2605.26097","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.26097","created_at":"2026-05-26T02:05:27Z"},{"alias_kind":"arxiv_version","alias_value":"2605.26097v1","created_at":"2026-05-26T02:05:27Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.26097","created_at":"2026-05-26T02:05:27Z"},{"alias_kind":"pith_short_12","alias_value":"7746NDQ6F2TD","created_at":"2026-05-26T02:05:27Z"},{"alias_kind":"pith_short_16","alias_value":"7746NDQ6F2TDP2S5","created_at":"2026-05-26T02:05:27Z"},{"alias_kind":"pith_short_8","alias_value":"7746NDQ6","created_at":"2026-05-26T02:05:27Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:7746NDQ6F2TDP2S5PQXIAQW2UO","target":"record","payload":{"canonical_record":{"source":{"id":"2605.26097","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-25T17:54:34Z","cross_cats_sorted":[],"title_canon_sha256":"fbbf9a1f048801e6cd8acc4dff69234c8eeb51b71a85b48878d8cfebc32c7de7","abstract_canon_sha256":"aa04cf1a381db67298a6324fe3c739ee5b8006e18dde441ec7e0507abc123f78"},"schema_version":"1.0"},"canonical_sha256":"fff9e68e1e2ea637ea5d7c2e8042daa3be8b76c1275bbdb77ec6ac9a4ac7be45","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T02:05:27.319818Z","signature_b64":"QKheUuIhaIVSaJrNWUYg8NKYb2X6SOInCHUy6ch54ugS8tZP1CaLQAUH1DRU46sWdimnf95oRXRVs5x9mmoJBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fff9e68e1e2ea637ea5d7c2e8042daa3be8b76c1275bbdb77ec6ac9a4ac7be45","last_reissued_at":"2026-05-26T02:05:27.319080Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T02:05:27.319080Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.26097","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":"2026-05-26T02:05:27Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"09TsMqLosvYGprQfj4hJkqQsk1FzgLdxyQWt/MDymgwj4uIv/eIiTI8jddYc7YRRn2+tJj+ozY987VW4sHN6Dg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-02T06:01:14.282778Z"},"content_sha256":"1832628158d4cc26e3f5055f5b6f8e20d416fad33823ed1abd4c9e05312c708c","schema_version":"1.0","event_id":"sha256:1832628158d4cc26e3f5055f5b6f8e20d416fad33823ed1abd4c9e05312c708c"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:7746NDQ6F2TDP2S5PQXIAQW2UO","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Forgetting in Language Models: Capacity, Optimization, and Self-Generated Replay","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Andrew Gordon Wilson, Dongkyu Cho, Martin Marek, Pavel Izmailov, Rumi Chunara, Shikai Qiu","submitted_at":"2026-05-25T17:54:34Z","abstract_excerpt":"Models trained on a new task typically degrade on prior tasks, a phenomenon known as forgetting. Traditionally, mitigating forgetting has required replaying stored exemplars from prior tasks, which is often impractical. By contrast, language models can sample from their own training distribution, and we show that these self-generated samples serve as effective replay data, nearly eliminating forgetting. We find that forgetting nonetheless persists when the model has little remaining capacity: models pretrained close to saturation cannot absorb new information without overwriting prior knowledg"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.26097","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/2605.26097/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-26T02:05:27Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"MpMdbB2Er4/HdnRvS2wIj650PksL/9mbcpMFdHqSa9DVw/pUjmp1bluzFgz1XktNxafkZA+ipRT/AV/jaKdsBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-02T06:01:14.283167Z"},"content_sha256":"b1363419ea46a6f7c0ad3f3b56460336608322a84c2e004d2c998de5a86cca41","schema_version":"1.0","event_id":"sha256:b1363419ea46a6f7c0ad3f3b56460336608322a84c2e004d2c998de5a86cca41"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/7746NDQ6F2TDP2S5PQXIAQW2UO/bundle.json","state_url":"https://pith.science/pith/7746NDQ6F2TDP2S5PQXIAQW2UO/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/7746NDQ6F2TDP2S5PQXIAQW2UO/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":"2026-06-02T06:01:14Z","links":{"resolver":"https://pith.science/pith/7746NDQ6F2TDP2S5PQXIAQW2UO","bundle":"https://pith.science/pith/7746NDQ6F2TDP2S5PQXIAQW2UO/bundle.json","state":"https://pith.science/pith/7746NDQ6F2TDP2S5PQXIAQW2UO/state.json","well_known_bundle":"https://pith.science/.well-known/pith/7746NDQ6F2TDP2S5PQXIAQW2UO/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:7746NDQ6F2TDP2S5PQXIAQW2UO","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":"aa04cf1a381db67298a6324fe3c739ee5b8006e18dde441ec7e0507abc123f78","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-25T17:54:34Z","title_canon_sha256":"fbbf9a1f048801e6cd8acc4dff69234c8eeb51b71a85b48878d8cfebc32c7de7"},"schema_version":"1.0","source":{"id":"2605.26097","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.26097","created_at":"2026-05-26T02:05:27Z"},{"alias_kind":"arxiv_version","alias_value":"2605.26097v1","created_at":"2026-05-26T02:05:27Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.26097","created_at":"2026-05-26T02:05:27Z"},{"alias_kind":"pith_short_12","alias_value":"7746NDQ6F2TD","created_at":"2026-05-26T02:05:27Z"},{"alias_kind":"pith_short_16","alias_value":"7746NDQ6F2TDP2S5","created_at":"2026-05-26T02:05:27Z"},{"alias_kind":"pith_short_8","alias_value":"7746NDQ6","created_at":"2026-05-26T02:05:27Z"}],"graph_snapshots":[{"event_id":"sha256:b1363419ea46a6f7c0ad3f3b56460336608322a84c2e004d2c998de5a86cca41","target":"graph","created_at":"2026-05-26T02:05:27Z","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":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2605.26097/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Models trained on a new task typically degrade on prior tasks, a phenomenon known as forgetting. Traditionally, mitigating forgetting has required replaying stored exemplars from prior tasks, which is often impractical. By contrast, language models can sample from their own training distribution, and we show that these self-generated samples serve as effective replay data, nearly eliminating forgetting. We find that forgetting nonetheless persists when the model has little remaining capacity: models pretrained close to saturation cannot absorb new information without overwriting prior knowledg","authors_text":"Andrew Gordon Wilson, Dongkyu Cho, Martin Marek, Pavel Izmailov, Rumi Chunara, Shikai Qiu","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-25T17:54:34Z","title":"Forgetting in Language Models: Capacity, Optimization, and Self-Generated Replay"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.26097","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:1832628158d4cc26e3f5055f5b6f8e20d416fad33823ed1abd4c9e05312c708c","target":"record","created_at":"2026-05-26T02:05:27Z","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":"aa04cf1a381db67298a6324fe3c739ee5b8006e18dde441ec7e0507abc123f78","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-05-25T17:54:34Z","title_canon_sha256":"fbbf9a1f048801e6cd8acc4dff69234c8eeb51b71a85b48878d8cfebc32c7de7"},"schema_version":"1.0","source":{"id":"2605.26097","kind":"arxiv","version":1}},"canonical_sha256":"fff9e68e1e2ea637ea5d7c2e8042daa3be8b76c1275bbdb77ec6ac9a4ac7be45","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"fff9e68e1e2ea637ea5d7c2e8042daa3be8b76c1275bbdb77ec6ac9a4ac7be45","first_computed_at":"2026-05-26T02:05:27.319080Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-26T02:05:27.319080Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"QKheUuIhaIVSaJrNWUYg8NKYb2X6SOInCHUy6ch54ugS8tZP1CaLQAUH1DRU46sWdimnf95oRXRVs5x9mmoJBQ==","signature_status":"signed_v1","signed_at":"2026-05-26T02:05:27.319818Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.26097","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:1832628158d4cc26e3f5055f5b6f8e20d416fad33823ed1abd4c9e05312c708c","sha256:b1363419ea46a6f7c0ad3f3b56460336608322a84c2e004d2c998de5a86cca41"],"state_sha256":"3999c0ee0b6743864fd55eb2d2b1fc9d302165e25a2450b838d2d03e7535069a"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6LXBkT/Nb+lYNSdifAF16M+viErle8Wv0F9ZtGXfSINFmahSkiU9fWSZTIL1/k5+BuAszQaUj0BwCkLnkSyvDA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-02T06:01:14.285508Z","bundle_sha256":"3d1477cdc0bbc4950e17b46aedffe7f9de1134c21f0af3ba0e8ecb7fba82cbc5"}}