{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:PFHXQCNERNF35JBXQJB2ZRTIYW","short_pith_number":"pith:PFHXQCNE","canonical_record":{"source":{"id":"2605.28304","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-27T11:00:26Z","cross_cats_sorted":[],"title_canon_sha256":"90569763f9fc3d0900aa195b18baf0ba6d4ea661ffa3e241deb0685aa4b076a1","abstract_canon_sha256":"068fd7890b1efb0e682140afe9a05f6bb4fc613fca527fed5e8aa52b2f53ffa5"},"schema_version":"1.0"},"canonical_sha256":"794f7809a48b4bbea4378243acc668c5b4e0618d174a0640d585b378e74fa23b","source":{"kind":"arxiv","id":"2605.28304","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.28304","created_at":"2026-05-28T01:05:05Z"},{"alias_kind":"arxiv_version","alias_value":"2605.28304v1","created_at":"2026-05-28T01:05:05Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.28304","created_at":"2026-05-28T01:05:05Z"},{"alias_kind":"pith_short_12","alias_value":"PFHXQCNERNF3","created_at":"2026-05-28T01:05:05Z"},{"alias_kind":"pith_short_16","alias_value":"PFHXQCNERNF35JBX","created_at":"2026-05-28T01:05:05Z"},{"alias_kind":"pith_short_8","alias_value":"PFHXQCNE","created_at":"2026-05-28T01:05:05Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:PFHXQCNERNF35JBXQJB2ZRTIYW","target":"record","payload":{"canonical_record":{"source":{"id":"2605.28304","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-27T11:00:26Z","cross_cats_sorted":[],"title_canon_sha256":"90569763f9fc3d0900aa195b18baf0ba6d4ea661ffa3e241deb0685aa4b076a1","abstract_canon_sha256":"068fd7890b1efb0e682140afe9a05f6bb4fc613fca527fed5e8aa52b2f53ffa5"},"schema_version":"1.0"},"canonical_sha256":"794f7809a48b4bbea4378243acc668c5b4e0618d174a0640d585b378e74fa23b","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:05:05.827852Z","signature_b64":"PcUsnmBLCJvtACY9FUHI9uzxTo5HOUQXRPRGYIAXlboVfMMMCBycBtuzcFJFavT4E+CXoAdOKbBqMV4HGf+fBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"794f7809a48b4bbea4378243acc668c5b4e0618d174a0640d585b378e74fa23b","last_reissued_at":"2026-05-28T01:05:05.826279Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:05:05.826279Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.28304","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":"2026-05-28T01:05:05Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"uTlb03ExC4dppDFwXDSQvTIZLiRgAYAdZK83g6XrXrwSucsDh/0iRuTH87WscueRF/H8Sv57qH01nOXWwleKCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-20T02:40:55.667045Z"},"content_sha256":"86db4b83ad8e9cb1548607da98b5380e531ca3057ff00398f9ab50a08f06672e","schema_version":"1.0","event_id":"sha256:86db4b83ad8e9cb1548607da98b5380e531ca3057ff00398f9ab50a08f06672e"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:PFHXQCNERNF35JBXQJB2ZRTIYW","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Compositional Generalization in Autoregressive Models via Logit Composition","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Aakash Kumar, Emanuele Natale, Maria Sofia Bucarelli","submitted_at":"2026-05-27T11:00:26Z","abstract_excerpt":"Composing autoregressive models remains a core challenge in understanding how large language models can combine behaviors or skills learned across tasks. We introduce a new and principled composition strategy for autoregressive systems, inspired by composition methods developed for diffusion models. Under a factorized-conditionals assumption, we show that the resulting composition is projective: each component model preserves control over its own designated subspace of the output distribution avoiding interference between models. This property is further preserved under smooth reparameterizati"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.28304","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/2605.28304/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":"2026-05-28T01:05:05Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"/U3ss7otOqnL5dH4gyWjQoke15G7O0gYe3f2IFK5XFvuVIvuTzMnXNbSq/WVpti7R3fm4+v19Z9cziOSu6w2Dw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-20T02:40:55.667431Z"},"content_sha256":"b646a43690d787013b5a7a16e85a6b12b6b37978c28108e4f203674a7c6f2e49","schema_version":"1.0","event_id":"sha256:b646a43690d787013b5a7a16e85a6b12b6b37978c28108e4f203674a7c6f2e49"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/PFHXQCNERNF35JBXQJB2ZRTIYW/bundle.json","state_url":"https://pith.science/pith/PFHXQCNERNF35JBXQJB2ZRTIYW/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/PFHXQCNERNF35JBXQJB2ZRTIYW/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":"2026-06-20T02:40:55Z","links":{"resolver":"https://pith.science/pith/PFHXQCNERNF35JBXQJB2ZRTIYW","bundle":"https://pith.science/pith/PFHXQCNERNF35JBXQJB2ZRTIYW/bundle.json","state":"https://pith.science/pith/PFHXQCNERNF35JBXQJB2ZRTIYW/state.json","well_known_bundle":"https://pith.science/.well-known/pith/PFHXQCNERNF35JBXQJB2ZRTIYW/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:PFHXQCNERNF35JBXQJB2ZRTIYW","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":"068fd7890b1efb0e682140afe9a05f6bb4fc613fca527fed5e8aa52b2f53ffa5","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-27T11:00:26Z","title_canon_sha256":"90569763f9fc3d0900aa195b18baf0ba6d4ea661ffa3e241deb0685aa4b076a1"},"schema_version":"1.0","source":{"id":"2605.28304","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.28304","created_at":"2026-05-28T01:05:05Z"},{"alias_kind":"arxiv_version","alias_value":"2605.28304v1","created_at":"2026-05-28T01:05:05Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.28304","created_at":"2026-05-28T01:05:05Z"},{"alias_kind":"pith_short_12","alias_value":"PFHXQCNERNF3","created_at":"2026-05-28T01:05:05Z"},{"alias_kind":"pith_short_16","alias_value":"PFHXQCNERNF35JBX","created_at":"2026-05-28T01:05:05Z"},{"alias_kind":"pith_short_8","alias_value":"PFHXQCNE","created_at":"2026-05-28T01:05:05Z"}],"graph_snapshots":[{"event_id":"sha256:b646a43690d787013b5a7a16e85a6b12b6b37978c28108e4f203674a7c6f2e49","target":"graph","created_at":"2026-05-28T01:05:05Z","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/2605.28304/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Composing autoregressive models remains a core challenge in understanding how large language models can combine behaviors or skills learned across tasks. We introduce a new and principled composition strategy for autoregressive systems, inspired by composition methods developed for diffusion models. Under a factorized-conditionals assumption, we show that the resulting composition is projective: each component model preserves control over its own designated subspace of the output distribution avoiding interference between models. This property is further preserved under smooth reparameterizati","authors_text":"Aakash Kumar, Emanuele Natale, Maria Sofia Bucarelli","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-27T11:00:26Z","title":"Compositional Generalization in Autoregressive Models via Logit Composition"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.28304","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":"sha256:86db4b83ad8e9cb1548607da98b5380e531ca3057ff00398f9ab50a08f06672e","target":"record","created_at":"2026-05-28T01:05:05Z","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":"068fd7890b1efb0e682140afe9a05f6bb4fc613fca527fed5e8aa52b2f53ffa5","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-27T11:00:26Z","title_canon_sha256":"90569763f9fc3d0900aa195b18baf0ba6d4ea661ffa3e241deb0685aa4b076a1"},"schema_version":"1.0","source":{"id":"2605.28304","kind":"arxiv","version":1}},"canonical_sha256":"794f7809a48b4bbea4378243acc668c5b4e0618d174a0640d585b378e74fa23b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"794f7809a48b4bbea4378243acc668c5b4e0618d174a0640d585b378e74fa23b","first_computed_at":"2026-05-28T01:05:05.826279Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-28T01:05:05.826279Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"PcUsnmBLCJvtACY9FUHI9uzxTo5HOUQXRPRGYIAXlboVfMMMCBycBtuzcFJFavT4E+CXoAdOKbBqMV4HGf+fBA==","signature_status":"signed_v1","signed_at":"2026-05-28T01:05:05.827852Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.28304","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:86db4b83ad8e9cb1548607da98b5380e531ca3057ff00398f9ab50a08f06672e","sha256:b646a43690d787013b5a7a16e85a6b12b6b37978c28108e4f203674a7c6f2e49"],"state_sha256":"096db7960a65b0173b6ca1586d489c156e14622397d537963b5cff33b632189d"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"392i3Y1c6COOQAB/xd6eXBN1A98WI6YotBgA3RSoImXT5xdfYoLpKcTzU3D1mX93TZoU64i0iArlszPFxcUnBQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-20T02:40:55.669445Z","bundle_sha256":"47994731a7ee9c45bae0b5362b2b3e2da61f4bcec118745e0aee259bd1cde8fc"}}