{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:W7D3JMFUX5SNEQGYLGSRFFZHJ5","short_pith_number":"pith:W7D3JMFU","canonical_record":{"source":{"id":"2306.03745","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-06-06T15:04:31Z","cross_cats_sorted":[],"title_canon_sha256":"048043caaa1893c7dd57a34e86a1e6eff88ea97aeb5a239bdc811764c11d3272","abstract_canon_sha256":"0f90d2cd8f24adf5b5eeda316e9154764d3efcea56ec59f0219de210086d340c"},"schema_version":"1.0"},"canonical_sha256":"b7c7b4b0b4bf64d240d859a51297274f64fab788368c36632b62b43d1ab1030c","source":{"kind":"arxiv","id":"2306.03745","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2306.03745","created_at":"2026-07-05T08:18:31Z"},{"alias_kind":"arxiv_version","alias_value":"2306.03745v2","created_at":"2026-07-05T08:18:31Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.03745","created_at":"2026-07-05T08:18:31Z"},{"alias_kind":"pith_short_12","alias_value":"W7D3JMFUX5SN","created_at":"2026-07-05T08:18:31Z"},{"alias_kind":"pith_short_16","alias_value":"W7D3JMFUX5SNEQGY","created_at":"2026-07-05T08:18:31Z"},{"alias_kind":"pith_short_8","alias_value":"W7D3JMFU","created_at":"2026-07-05T08:18:31Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:W7D3JMFUX5SNEQGYLGSRFFZHJ5","target":"record","payload":{"canonical_record":{"source":{"id":"2306.03745","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-06-06T15:04:31Z","cross_cats_sorted":[],"title_canon_sha256":"048043caaa1893c7dd57a34e86a1e6eff88ea97aeb5a239bdc811764c11d3272","abstract_canon_sha256":"0f90d2cd8f24adf5b5eeda316e9154764d3efcea56ec59f0219de210086d340c"},"schema_version":"1.0"},"canonical_sha256":"b7c7b4b0b4bf64d240d859a51297274f64fab788368c36632b62b43d1ab1030c","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:18:31.970774Z","signature_b64":"72GfQGbzbEptCz+IphGDbz/BSfpWjIDATVhy+BidfGyH6qyYKFRRzjCUIvjnNtESPsnGAK9fRe7wMdxmWBM4CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b7c7b4b0b4bf64d240d859a51297274f64fab788368c36632b62b43d1ab1030c","last_reissued_at":"2026-07-05T08:18:31.970359Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:18:31.970359Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2306.03745","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":"2026-07-05T08:18:31Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"GVo9x5GAeN5+aZw55w5sA0vroC9oRwFWaCsBaRAW7sMLjbMQLv7ZFpnh2ctPien/2QNSZKbvGOoGBZlMB5C6Ag==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-07T17:15:51.088968Z"},"content_sha256":"9b87f59977775b3517e03f14d556bc104bf5b60c15742155f05421d6cf130d79","schema_version":"1.0","event_id":"sha256:9b87f59977775b3517e03f14d556bc104bf5b60c15742155f05421d6cf130d79"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:W7D3JMFUX5SNEQGYLGSRFFZHJ5","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Soft Merging of Experts with Adaptive Routing","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Colin Raffel, Haokun Liu, Mohammed Muqeeth","submitted_at":"2023-06-06T15:04:31Z","abstract_excerpt":"Sparsely activated neural networks with conditional computation learn to route their inputs through different \"expert\" subnetworks, providing a form of modularity that densely activated models lack. Despite their possible benefits, models with learned routing often underperform their parameter-matched densely activated counterparts as well as models that use non-learned heuristic routing strategies. In this paper, we hypothesize that these shortcomings stem from the gradient estimation techniques used to train sparsely activated models that use non-differentiable discrete routing decisions. To"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.03745","kind":"arxiv","version":2},"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/2306.03745/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":"2026-07-05T08:18:31Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"KdbGBAcERc5Z2zXUThfzuaU1cVYhKWTe6OkUbMons4PRTeCLP2Cmk6t1/1HjExeSu+AcnfbQawBmTxdZ4bWTDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-07T17:15:51.089968Z"},"content_sha256":"de4c372fbba3c8f789fb87b9cf056f2218424997a5cc519eb07150728877fc2b","schema_version":"1.0","event_id":"sha256:de4c372fbba3c8f789fb87b9cf056f2218424997a5cc519eb07150728877fc2b"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/W7D3JMFUX5SNEQGYLGSRFFZHJ5/bundle.json","state_url":"https://pith.science/pith/W7D3JMFUX5SNEQGYLGSRFFZHJ5/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/W7D3JMFUX5SNEQGYLGSRFFZHJ5/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":"2026-08-07T17:15:51Z","links":{"resolver":"https://pith.science/pith/W7D3JMFUX5SNEQGYLGSRFFZHJ5","bundle":"https://pith.science/pith/W7D3JMFUX5SNEQGYLGSRFFZHJ5/bundle.json","state":"https://pith.science/pith/W7D3JMFUX5SNEQGYLGSRFFZHJ5/state.json","well_known_bundle":"https://pith.science/.well-known/pith/W7D3JMFUX5SNEQGYLGSRFFZHJ5/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:W7D3JMFUX5SNEQGYLGSRFFZHJ5","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":"0f90d2cd8f24adf5b5eeda316e9154764d3efcea56ec59f0219de210086d340c","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-06-06T15:04:31Z","title_canon_sha256":"048043caaa1893c7dd57a34e86a1e6eff88ea97aeb5a239bdc811764c11d3272"},"schema_version":"1.0","source":{"id":"2306.03745","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2306.03745","created_at":"2026-07-05T08:18:31Z"},{"alias_kind":"arxiv_version","alias_value":"2306.03745v2","created_at":"2026-07-05T08:18:31Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.03745","created_at":"2026-07-05T08:18:31Z"},{"alias_kind":"pith_short_12","alias_value":"W7D3JMFUX5SN","created_at":"2026-07-05T08:18:31Z"},{"alias_kind":"pith_short_16","alias_value":"W7D3JMFUX5SNEQGY","created_at":"2026-07-05T08:18:31Z"},{"alias_kind":"pith_short_8","alias_value":"W7D3JMFU","created_at":"2026-07-05T08:18:31Z"}],"graph_snapshots":[{"event_id":"sha256:de4c372fbba3c8f789fb87b9cf056f2218424997a5cc519eb07150728877fc2b","target":"graph","created_at":"2026-07-05T08:18:31Z","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/2306.03745/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Sparsely activated neural networks with conditional computation learn to route their inputs through different \"expert\" subnetworks, providing a form of modularity that densely activated models lack. Despite their possible benefits, models with learned routing often underperform their parameter-matched densely activated counterparts as well as models that use non-learned heuristic routing strategies. In this paper, we hypothesize that these shortcomings stem from the gradient estimation techniques used to train sparsely activated models that use non-differentiable discrete routing decisions. To","authors_text":"Colin Raffel, Haokun Liu, Mohammed Muqeeth","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-06-06T15:04:31Z","title":"Soft Merging of Experts with Adaptive Routing"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.03745","kind":"arxiv","version":2},"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":"sha256:9b87f59977775b3517e03f14d556bc104bf5b60c15742155f05421d6cf130d79","target":"record","created_at":"2026-07-05T08:18:31Z","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":"0f90d2cd8f24adf5b5eeda316e9154764d3efcea56ec59f0219de210086d340c","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2023-06-06T15:04:31Z","title_canon_sha256":"048043caaa1893c7dd57a34e86a1e6eff88ea97aeb5a239bdc811764c11d3272"},"schema_version":"1.0","source":{"id":"2306.03745","kind":"arxiv","version":2}},"canonical_sha256":"b7c7b4b0b4bf64d240d859a51297274f64fab788368c36632b62b43d1ab1030c","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"b7c7b4b0b4bf64d240d859a51297274f64fab788368c36632b62b43d1ab1030c","first_computed_at":"2026-07-05T08:18:31.970359Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:18:31.970359Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"72GfQGbzbEptCz+IphGDbz/BSfpWjIDATVhy+BidfGyH6qyYKFRRzjCUIvjnNtESPsnGAK9fRe7wMdxmWBM4CQ==","signature_status":"signed_v1","signed_at":"2026-07-05T08:18:31.970774Z","signed_message":"canonical_sha256_bytes"},"source_id":"2306.03745","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:9b87f59977775b3517e03f14d556bc104bf5b60c15742155f05421d6cf130d79","sha256:de4c372fbba3c8f789fb87b9cf056f2218424997a5cc519eb07150728877fc2b"],"state_sha256":"6a13a593d8fb222238d39f35e8f46a67f7900259b9af16f6c8900b754cec6109"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"7HCZfNiv2X0vbFABGZqAO8wkBA+sh+rhS072DfK555LwtLLo8DbtWm5Y4Vthwts6xEjfSHNmvXknXthLeyD9CA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-07T17:15:51.098457Z","bundle_sha256":"e405b82665ffd76d6b17f4829fdb566faaf1c33fb1bdc0dc849560f50b327910"}}