{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:T35E6CGLQMTLVT372TOZWGWVQW","short_pith_number":"pith:T35E6CGL","canonical_record":{"source":{"id":"2604.21454","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-23T09:13:28Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"708353e4b5cc819887184c542ba98c76c9529c339e934399ccbe367fcf8a2eb2","abstract_canon_sha256":"f528f1bdad47ef4f22a56ed59463f80109aa16f6411b65c0e4297c2eeadbd22e"},"schema_version":"1.0"},"canonical_sha256":"9efa4f08cb8326bacf7fd4dd9b1ad58588c2bb6c7329398a7f9835b902f7a92d","source":{"kind":"arxiv","id":"2604.21454","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.21454","created_at":"2026-05-27T00:04:24Z"},{"alias_kind":"arxiv_version","alias_value":"2604.21454v2","created_at":"2026-05-27T00:04:24Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.21454","created_at":"2026-05-27T00:04:24Z"},{"alias_kind":"pith_short_12","alias_value":"T35E6CGLQMTL","created_at":"2026-05-27T00:04:24Z"},{"alias_kind":"pith_short_16","alias_value":"T35E6CGLQMTLVT37","created_at":"2026-05-27T00:04:24Z"},{"alias_kind":"pith_short_8","alias_value":"T35E6CGL","created_at":"2026-05-27T00:04:24Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:T35E6CGLQMTLVT372TOZWGWVQW","target":"record","payload":{"canonical_record":{"source":{"id":"2604.21454","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-23T09:13:28Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"708353e4b5cc819887184c542ba98c76c9529c339e934399ccbe367fcf8a2eb2","abstract_canon_sha256":"f528f1bdad47ef4f22a56ed59463f80109aa16f6411b65c0e4297c2eeadbd22e"},"schema_version":"1.0"},"canonical_sha256":"9efa4f08cb8326bacf7fd4dd9b1ad58588c2bb6c7329398a7f9835b902f7a92d","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-27T00:04:24.615238Z","signature_b64":"nQZd2vzcXqwYQ9MjE//499X4XWBIMPbqKVplTkSADjS4zjz6vmyxLyHlZz7m3mtQ223CynwyQe5ITdPz9lGIAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9efa4f08cb8326bacf7fd4dd9b1ad58588c2bb6c7329398a7f9835b902f7a92d","last_reissued_at":"2026-05-27T00:04:24.614606Z","signature_status":"signed_v1","first_computed_at":"2026-05-27T00:04:24.614606Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.21454","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":"2026-05-27T00:04:24Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"QjK0Ly36EWhZ2MG99H2zhGQDOnMIxrU9SVfOEl0AowMSVR+7JSmDFRsESrL21D9bnprc2MNEd0bh1jhDzkrPBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T14:07:02.781819Z"},"content_sha256":"e0bc25ff80603d65e9df8e4372059b5f0887b3e3024e89d310b1e779de532a4d","schema_version":"1.0","event_id":"sha256:e0bc25ff80603d65e9df8e4372059b5f0887b3e3024e89d310b1e779de532a4d"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:T35E6CGLQMTLVT372TOZWGWVQW","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Reasoning Primitives in Hybrid and Non-Hybrid LLMs: Do Architectural Differences Yield Advantages in State-Tracking and Recall?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Reasoning augmentation extends the difficulty range where models stay effective on tasks mixing recall and state-tracking, with hybrid architectures showing greater robustness to rising sequential dependence than pure transformers.","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Florian Mai, Lucie Flek, Nicholas Kluge Corr\\^ea, Shivam Rawat","submitted_at":"2026-04-23T09:13:28Z","abstract_excerpt":"Reasoning in large language models is often discussed as a single capability, but some of its gains may stem from simpler underlying operations. We examine two such primitives, recall and state-tracking, through five controlled task families centered on state-based recall, and compare matched transformer and hybrid architectures with and without reasoning augmentation. Across the suite, reasoning-augmented variants substantially outperform instruction-only variants, often by large margins. This pattern is consistent with the State over Tokens view: externalized reasoning traces help because th"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"reasoning augmentation provides the largest overall improvement, substantially extending the range of difficulty over which models remain effective... in certain tasks, the hybrid reasoning model remains substantially more robust as sequential dependence increases. In contrast, the transformer reasoning model degrades sharply in performance as task difficulty increases beyond a given threshold.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the controlled tasks accurately isolate and jointly require only the recall and state-tracking primitives without confounding factors from model scale, training data, or task construction details, and that the matched Olmo3 transformer and hybrid variants differ only in the intended architectural inductive bias.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Reasoning augmentation extends the difficulty range for both architectures, but hybrid models stay robust longer than transformers as sequential dependence increases in state-based recall tasks.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Reasoning augmentation extends the difficulty range where models stay effective on tasks mixing recall and state-tracking, with hybrid architectures showing greater robustness to rising sequential dependence than pure transformers.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"bf079d32dbe7c44086219f7d2a5a93bb29ec52cde90b989291a5c2d856bc6145"},"source":{"id":"2604.21454","kind":"arxiv","version":2},"verdict":{"id":"2b6a6d66-7662-4451-a15a-f609274a758b","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-09T22:12:02.114307Z","strongest_claim":"reasoning augmentation provides the largest overall improvement, substantially extending the range of difficulty over which models remain effective... in certain tasks, the hybrid reasoning model remains substantially more robust as sequential dependence increases. In contrast, the transformer reasoning model degrades sharply in performance as task difficulty increases beyond a given threshold.","one_line_summary":"Reasoning augmentation extends the difficulty range for both architectures, but hybrid models stay robust longer than transformers as sequential dependence increases in state-based recall tasks.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the controlled tasks accurately isolate and jointly require only the recall and state-tracking primitives without confounding factors from model scale, training data, or task construction details, and that the matched Olmo3 transformer and hybrid variants differ only in the intended architectural inductive bias.","pith_extraction_headline":"Reasoning augmentation extends the difficulty range where models stay effective on tasks mixing recall and state-tracking, with hybrid architectures showing greater robustness to rising sequential dependence than pure transformers."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.21454/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-21T12:41:42.114623Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-20T00:57:26.237861Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"fe05663e62208222b8b445597d049498a348215cf0be61dab3e2b7035fc7e927"},"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":"2b6a6d66-7662-4451-a15a-f609274a758b"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-27T00:04:24Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ibo++Nrc4a1Ql/gqs7Dl5mlFlnwK6IdCV+cgRR19g32a6XPYT6o5taXYR+y9XSTwxyrW/5oCqsF5sl1+b772CA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T14:07:02.782814Z"},"content_sha256":"c170e0e34fe44d09eb0184eae424724f3ad13a25570ce4e84e3056c05931d5ab","schema_version":"1.0","event_id":"sha256:c170e0e34fe44d09eb0184eae424724f3ad13a25570ce4e84e3056c05931d5ab"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/T35E6CGLQMTLVT372TOZWGWVQW/bundle.json","state_url":"https://pith.science/pith/T35E6CGLQMTLVT372TOZWGWVQW/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/T35E6CGLQMTLVT372TOZWGWVQW/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":"2026-08-04T14:07:02Z","links":{"resolver":"https://pith.science/pith/T35E6CGLQMTLVT372TOZWGWVQW","bundle":"https://pith.science/pith/T35E6CGLQMTLVT372TOZWGWVQW/bundle.json","state":"https://pith.science/pith/T35E6CGLQMTLVT372TOZWGWVQW/state.json","well_known_bundle":"https://pith.science/.well-known/pith/T35E6CGLQMTLVT372TOZWGWVQW/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:T35E6CGLQMTLVT372TOZWGWVQW","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":"f528f1bdad47ef4f22a56ed59463f80109aa16f6411b65c0e4297c2eeadbd22e","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-23T09:13:28Z","title_canon_sha256":"708353e4b5cc819887184c542ba98c76c9529c339e934399ccbe367fcf8a2eb2"},"schema_version":"1.0","source":{"id":"2604.21454","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.21454","created_at":"2026-05-27T00:04:24Z"},{"alias_kind":"arxiv_version","alias_value":"2604.21454v2","created_at":"2026-05-27T00:04:24Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.21454","created_at":"2026-05-27T00:04:24Z"},{"alias_kind":"pith_short_12","alias_value":"T35E6CGLQMTL","created_at":"2026-05-27T00:04:24Z"},{"alias_kind":"pith_short_16","alias_value":"T35E6CGLQMTLVT37","created_at":"2026-05-27T00:04:24Z"},{"alias_kind":"pith_short_8","alias_value":"T35E6CGL","created_at":"2026-05-27T00:04:24Z"}],"graph_snapshots":[{"event_id":"sha256:c170e0e34fe44d09eb0184eae424724f3ad13a25570ce4e84e3056c05931d5ab","target":"graph","created_at":"2026-05-27T00:04:24Z","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":"reasoning augmentation provides the largest overall improvement, substantially extending the range of difficulty over which models remain effective... in certain tasks, the hybrid reasoning model remains substantially more robust as sequential dependence increases. In contrast, the transformer reasoning model degrades sharply in performance as task difficulty increases beyond a given threshold."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That the controlled tasks accurately isolate and jointly require only the recall and state-tracking primitives without confounding factors from model scale, training data, or task construction details, and that the matched Olmo3 transformer and hybrid variants differ only in the intended architectural inductive bias."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Reasoning augmentation extends the difficulty range for both architectures, but hybrid models stay robust longer than transformers as sequential dependence increases in state-based recall tasks."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Reasoning augmentation extends the difficulty range where models stay effective on tasks mixing recall and state-tracking, with hybrid architectures showing greater robustness to rising sequential dependence than pure transformers."}],"snapshot_sha256":"bf079d32dbe7c44086219f7d2a5a93bb29ec52cde90b989291a5c2d856bc6145"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"ai_meta_artifact","ran_at":"2026-05-21T12:41:42.114623Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-20T00:57:26.237861Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2604.21454/integrity.json","findings":[],"snapshot_sha256":"fe05663e62208222b8b445597d049498a348215cf0be61dab3e2b7035fc7e927","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Reasoning in large language models is often discussed as a single capability, but some of its gains may stem from simpler underlying operations. We examine two such primitives, recall and state-tracking, through five controlled task families centered on state-based recall, and compare matched transformer and hybrid architectures with and without reasoning augmentation. Across the suite, reasoning-augmented variants substantially outperform instruction-only variants, often by large margins. This pattern is consistent with the State over Tokens view: externalized reasoning traces help because th","authors_text":"Florian Mai, Lucie Flek, Nicholas Kluge Corr\\^ea, Shivam Rawat","cross_cats":["cs.AI"],"headline":"Reasoning augmentation extends the difficulty range where models stay effective on tasks mixing recall and state-tracking, with hybrid architectures showing greater robustness to rising sequential dependence than pure transformers.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-23T09:13:28Z","title":"Reasoning Primitives in Hybrid and Non-Hybrid LLMs: Do Architectural Differences Yield Advantages in State-Tracking and Recall?"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.21454","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-09T22:12:02.114307Z","id":"2b6a6d66-7662-4451-a15a-f609274a758b","model_set":{"reader":"grok-4.3"},"one_line_summary":"Reasoning augmentation extends the difficulty range for both architectures, but hybrid models stay robust longer than transformers as sequential dependence increases in state-based recall tasks.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Reasoning augmentation extends the difficulty range where models stay effective on tasks mixing recall and state-tracking, with hybrid architectures showing greater robustness to rising sequential dependence than pure transformers.","strongest_claim":"reasoning augmentation provides the largest overall improvement, substantially extending the range of difficulty over which models remain effective... in certain tasks, the hybrid reasoning model remains substantially more robust as sequential dependence increases. In contrast, the transformer reasoning model degrades sharply in performance as task difficulty increases beyond a given threshold.","weakest_assumption":"That the controlled tasks accurately isolate and jointly require only the recall and state-tracking primitives without confounding factors from model scale, training data, or task construction details, and that the matched Olmo3 transformer and hybrid variants differ only in the intended architectural inductive bias."}},"verdict_id":"2b6a6d66-7662-4451-a15a-f609274a758b"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:e0bc25ff80603d65e9df8e4372059b5f0887b3e3024e89d310b1e779de532a4d","target":"record","created_at":"2026-05-27T00:04:24Z","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":"f528f1bdad47ef4f22a56ed59463f80109aa16f6411b65c0e4297c2eeadbd22e","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-23T09:13:28Z","title_canon_sha256":"708353e4b5cc819887184c542ba98c76c9529c339e934399ccbe367fcf8a2eb2"},"schema_version":"1.0","source":{"id":"2604.21454","kind":"arxiv","version":2}},"canonical_sha256":"9efa4f08cb8326bacf7fd4dd9b1ad58588c2bb6c7329398a7f9835b902f7a92d","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"9efa4f08cb8326bacf7fd4dd9b1ad58588c2bb6c7329398a7f9835b902f7a92d","first_computed_at":"2026-05-27T00:04:24.614606Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-27T00:04:24.614606Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"nQZd2vzcXqwYQ9MjE//499X4XWBIMPbqKVplTkSADjS4zjz6vmyxLyHlZz7m3mtQ223CynwyQe5ITdPz9lGIAA==","signature_status":"signed_v1","signed_at":"2026-05-27T00:04:24.615238Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.21454","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:e0bc25ff80603d65e9df8e4372059b5f0887b3e3024e89d310b1e779de532a4d","sha256:c170e0e34fe44d09eb0184eae424724f3ad13a25570ce4e84e3056c05931d5ab"],"state_sha256":"d5f2c77095fcd6b9299f0821876574c51316f1035aa44937b6b0e16e211a7d0c"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ixK0i5owYxmaSyMXasbJaXoggG4/FHMgTW5Y8/BXGU+KVSJU+wR+FeKHrL52ytq+pOzF0PogX/EfBPlzZsYNAw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-04T14:07:02.788928Z","bundle_sha256":"dc6b0ed979b709df60b0495e6c9bbb630a6bc96bd8d2428f821e5180610de031"}}