{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2021:UX4SGRUFQXVK32SVDKF4ESWMY5","short_pith_number":"pith:UX4SGRUF","canonical_record":{"source":{"id":"2106.03760","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-06-07T16:25:27Z","cross_cats_sorted":["math.OC","stat.ML"],"title_canon_sha256":"c43f1dff7516e59a2951f2384cb296312a7e91c33a53abbf9e8d704b6bed4fc4","abstract_canon_sha256":"0d9ae6d7882f8b9b0360977a3904a207bed42fb5a9d85e9c9098b310a7b4bf16"},"schema_version":"1.0"},"canonical_sha256":"a5f923468585eaadea551a8bc24accc75e0370420d8e414f3b82b967c6af01c5","source":{"kind":"arxiv","id":"2106.03760","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2106.03760","created_at":"2026-07-05T03:44:54Z"},{"alias_kind":"arxiv_version","alias_value":"2106.03760v3","created_at":"2026-07-05T03:44:54Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.03760","created_at":"2026-07-05T03:44:54Z"},{"alias_kind":"pith_short_12","alias_value":"UX4SGRUFQXVK","created_at":"2026-07-05T03:44:54Z"},{"alias_kind":"pith_short_16","alias_value":"UX4SGRUFQXVK32SV","created_at":"2026-07-05T03:44:54Z"},{"alias_kind":"pith_short_8","alias_value":"UX4SGRUF","created_at":"2026-07-05T03:44:54Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2021:UX4SGRUFQXVK32SVDKF4ESWMY5","target":"record","payload":{"canonical_record":{"source":{"id":"2106.03760","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-06-07T16:25:27Z","cross_cats_sorted":["math.OC","stat.ML"],"title_canon_sha256":"c43f1dff7516e59a2951f2384cb296312a7e91c33a53abbf9e8d704b6bed4fc4","abstract_canon_sha256":"0d9ae6d7882f8b9b0360977a3904a207bed42fb5a9d85e9c9098b310a7b4bf16"},"schema_version":"1.0"},"canonical_sha256":"a5f923468585eaadea551a8bc24accc75e0370420d8e414f3b82b967c6af01c5","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:44:54.033495Z","signature_b64":"Z6XUNNTJ4xaLfOdQnTfqUDBQx/ymLYnenWBlggOUBB47aGOokWbkqoGCP9oOguAeVG0a17qtjQQlESpivs0vBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a5f923468585eaadea551a8bc24accc75e0370420d8e414f3b82b967c6af01c5","last_reissued_at":"2026-07-05T03:44:54.033040Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:44:54.033040Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2106.03760","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T03:44:54Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"3vHjY+DPC9YozZBXr9XsPbJrw+e8m1s1qVwYQ6Wb1zW52493tbQTjk4OQRakni6TkCJzJ0iAusq5QL51CkpnCg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-12T10:21:38.601233Z"},"content_sha256":"becf8b4572e90ee8c0833c7c480d90f254dd537b447654e2acb7c85b0d9edd15","schema_version":"1.0","event_id":"sha256:becf8b4572e90ee8c0833c7c480d90f254dd537b447654e2acb7c85b0d9edd15"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2021:UX4SGRUFQXVK32SVDKF4ESWMY5","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"DSelect-k: Differentiable Selection in the Mixture of Experts with Applications to Multi-Task Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["math.OC","stat.ML"],"primary_cat":"cs.LG","authors_text":"Aakanksha Chowdhery, Ed H. Chi, Hussein Hazimeh, Lichan Hong, Maheswaran Sathiamoorthy, Rahul Mazumder, Yihua Chen, Zhe Zhao","submitted_at":"2021-06-07T16:25:27Z","abstract_excerpt":"The Mixture-of-Experts (MoE) architecture is showing promising results in improving parameter sharing in multi-task learning (MTL) and in scaling high-capacity neural networks. State-of-the-art MoE models use a trainable sparse gate to select a subset of the experts for each input example. While conceptually appealing, existing sparse gates, such as Top-k, are not smooth. The lack of smoothness can lead to convergence and statistical performance issues when training with gradient-based methods. In this paper, we develop DSelect-k: a continuously differentiable and sparse gate for MoE, based on"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.03760","kind":"arxiv","version":3},"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/2106.03760/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":"2026-07-05T03:44:54Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"NgbVDm6DCwvPsbF9TuxxT6Hxp/zF5uXHGZ6R0ToR1M/EDBLCw41omeFqS5JOdAUQCuz5z1p5AkbHrC+YJu8uAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-12T10:21:38.601764Z"},"content_sha256":"4943e61df2f7d582854358a11bf0008651a47105daddbbabbee5f8d49f3b0161","schema_version":"1.0","event_id":"sha256:4943e61df2f7d582854358a11bf0008651a47105daddbbabbee5f8d49f3b0161"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/UX4SGRUFQXVK32SVDKF4ESWMY5/bundle.json","state_url":"https://pith.science/pith/UX4SGRUFQXVK32SVDKF4ESWMY5/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/UX4SGRUFQXVK32SVDKF4ESWMY5/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":"2026-08-12T10:21:38Z","links":{"resolver":"https://pith.science/pith/UX4SGRUFQXVK32SVDKF4ESWMY5","bundle":"https://pith.science/pith/UX4SGRUFQXVK32SVDKF4ESWMY5/bundle.json","state":"https://pith.science/pith/UX4SGRUFQXVK32SVDKF4ESWMY5/state.json","well_known_bundle":"https://pith.science/.well-known/pith/UX4SGRUFQXVK32SVDKF4ESWMY5/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2021:UX4SGRUFQXVK32SVDKF4ESWMY5","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":"0d9ae6d7882f8b9b0360977a3904a207bed42fb5a9d85e9c9098b310a7b4bf16","cross_cats_sorted":["math.OC","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-06-07T16:25:27Z","title_canon_sha256":"c43f1dff7516e59a2951f2384cb296312a7e91c33a53abbf9e8d704b6bed4fc4"},"schema_version":"1.0","source":{"id":"2106.03760","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2106.03760","created_at":"2026-07-05T03:44:54Z"},{"alias_kind":"arxiv_version","alias_value":"2106.03760v3","created_at":"2026-07-05T03:44:54Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.03760","created_at":"2026-07-05T03:44:54Z"},{"alias_kind":"pith_short_12","alias_value":"UX4SGRUFQXVK","created_at":"2026-07-05T03:44:54Z"},{"alias_kind":"pith_short_16","alias_value":"UX4SGRUFQXVK32SV","created_at":"2026-07-05T03:44:54Z"},{"alias_kind":"pith_short_8","alias_value":"UX4SGRUF","created_at":"2026-07-05T03:44:54Z"}],"graph_snapshots":[{"event_id":"sha256:4943e61df2f7d582854358a11bf0008651a47105daddbbabbee5f8d49f3b0161","target":"graph","created_at":"2026-07-05T03:44:54Z","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/2106.03760/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The Mixture-of-Experts (MoE) architecture is showing promising results in improving parameter sharing in multi-task learning (MTL) and in scaling high-capacity neural networks. State-of-the-art MoE models use a trainable sparse gate to select a subset of the experts for each input example. While conceptually appealing, existing sparse gates, such as Top-k, are not smooth. The lack of smoothness can lead to convergence and statistical performance issues when training with gradient-based methods. In this paper, we develop DSelect-k: a continuously differentiable and sparse gate for MoE, based on","authors_text":"Aakanksha Chowdhery, Ed H. Chi, Hussein Hazimeh, Lichan Hong, Maheswaran Sathiamoorthy, Rahul Mazumder, Yihua Chen, Zhe Zhao","cross_cats":["math.OC","stat.ML"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-06-07T16:25:27Z","title":"DSelect-k: Differentiable Selection in the Mixture of Experts with Applications to Multi-Task Learning"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.03760","kind":"arxiv","version":3},"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:becf8b4572e90ee8c0833c7c480d90f254dd537b447654e2acb7c85b0d9edd15","target":"record","created_at":"2026-07-05T03:44:54Z","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":"0d9ae6d7882f8b9b0360977a3904a207bed42fb5a9d85e9c9098b310a7b4bf16","cross_cats_sorted":["math.OC","stat.ML"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2021-06-07T16:25:27Z","title_canon_sha256":"c43f1dff7516e59a2951f2384cb296312a7e91c33a53abbf9e8d704b6bed4fc4"},"schema_version":"1.0","source":{"id":"2106.03760","kind":"arxiv","version":3}},"canonical_sha256":"a5f923468585eaadea551a8bc24accc75e0370420d8e414f3b82b967c6af01c5","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"a5f923468585eaadea551a8bc24accc75e0370420d8e414f3b82b967c6af01c5","first_computed_at":"2026-07-05T03:44:54.033040Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T03:44:54.033040Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"Z6XUNNTJ4xaLfOdQnTfqUDBQx/ymLYnenWBlggOUBB47aGOokWbkqoGCP9oOguAeVG0a17qtjQQlESpivs0vBg==","signature_status":"signed_v1","signed_at":"2026-07-05T03:44:54.033495Z","signed_message":"canonical_sha256_bytes"},"source_id":"2106.03760","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:becf8b4572e90ee8c0833c7c480d90f254dd537b447654e2acb7c85b0d9edd15","sha256:4943e61df2f7d582854358a11bf0008651a47105daddbbabbee5f8d49f3b0161"],"state_sha256":"907c2448620bacb965d5826a860ed0fc4ab73af89041c1b35ffc17acb5808c55"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ZF0SQS3442VHg70d1n1fRXWeW4epANKN5A5XZ7PqEhldQ0FalNcRr8yJ/+joVm+weeDSw3JNARI/RXCwBnY1DA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-12T10:21:38.605730Z","bundle_sha256":"23476dec3082f47e756594a8cce55502640ed508bd9b88c4909af2f77aa6f000"}}