{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:MGKSERVWT4HMPU65YPFJB2US2Y","short_pith_number":"pith:MGKSERVW","schema_version":"1.0","canonical_sha256":"61952246b69f0ec7d3ddc3ca90ea92d605c1a84a9821b335ddf8569db05c3ad2","source":{"kind":"arxiv","id":"2310.07096","version":1},"attestation_state":"computed","paper":{"title":"Sparse Universal Transformer","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Aaron Courville, Chuang Gan, Shawn Tan, Yikang Shen, Zhenfang Chen","submitted_at":"2023-10-11T00:38:57Z","abstract_excerpt":"The Universal Transformer (UT) is a variant of the Transformer that shares parameters across its layers. Empirical evidence shows that UTs have better compositional generalization than Vanilla Transformers (VTs) in formal language tasks. The parameter-sharing also affords it better parameter efficiency than VTs. Despite its many advantages, scaling UT parameters is much more compute and memory intensive than scaling up a VT. This paper proposes the Sparse Universal Transformer (SUT), which leverages Sparse Mixture of Experts (SMoE) and a new stick-breaking-based dynamic halting mechanism to re"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2310.07096","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-10-11T00:38:57Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"adfffbe13bc7d7906e3ee7a9df2564f65a097cbbf75bc5e5dc0ae8d39f6d4bdf","abstract_canon_sha256":"c298ace201086b001001849c3132fe38cc590ef16fa87d170e6e85aee9c4919f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:59:40.431608Z","signature_b64":"eP4mzQYxW4A5BjbXD9fqlAZx8TJO/idVysr63O5koEXDMDAAFkWI81vSg3ggsH6z9CrWo1RbMytttyZefdOJDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"61952246b69f0ec7d3ddc3ca90ea92d605c1a84a9821b335ddf8569db05c3ad2","last_reissued_at":"2026-07-05T06:59:40.431198Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:59:40.431198Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Sparse Universal Transformer","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Aaron Courville, Chuang Gan, Shawn Tan, Yikang Shen, Zhenfang Chen","submitted_at":"2023-10-11T00:38:57Z","abstract_excerpt":"The Universal Transformer (UT) is a variant of the Transformer that shares parameters across its layers. Empirical evidence shows that UTs have better compositional generalization than Vanilla Transformers (VTs) in formal language tasks. The parameter-sharing also affords it better parameter efficiency than VTs. Despite its many advantages, scaling UT parameters is much more compute and memory intensive than scaling up a VT. This paper proposes the Sparse Universal Transformer (SUT), which leverages Sparse Mixture of Experts (SMoE) and a new stick-breaking-based dynamic halting mechanism to re"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.07096","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/2310.07096/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2310.07096","created_at":"2026-07-05T06:59:40.431255+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.07096v1","created_at":"2026-07-05T06:59:40.431255+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.07096","created_at":"2026-07-05T06:59:40.431255+00:00"},{"alias_kind":"pith_short_12","alias_value":"MGKSERVWT4HM","created_at":"2026-07-05T06:59:40.431255+00:00"},{"alias_kind":"pith_short_16","alias_value":"MGKSERVWT4HMPU65","created_at":"2026-07-05T06:59:40.431255+00:00"},{"alias_kind":"pith_short_8","alias_value":"MGKSERVW","created_at":"2026-07-05T06:59:40.431255+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.07604","citing_title":"Contribution Weights: A Geometrical Analysis of Self-Attention Transformers","ref_index":122,"is_internal_anchor":false},{"citing_arxiv_id":"2502.05171","citing_title":"Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach","ref_index":155,"is_internal_anchor":false},{"citing_arxiv_id":"2604.11791","citing_title":"A Mechanistic Analysis of Looped Reasoning Language Models","ref_index":27,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MGKSERVWT4HMPU65YPFJB2US2Y","json":"https://pith.science/pith/MGKSERVWT4HMPU65YPFJB2US2Y.json","graph_json":"https://pith.science/api/pith-number/MGKSERVWT4HMPU65YPFJB2US2Y/graph.json","events_json":"https://pith.science/api/pith-number/MGKSERVWT4HMPU65YPFJB2US2Y/events.json","paper":"https://pith.science/paper/MGKSERVW"},"agent_actions":{"view_html":"https://pith.science/pith/MGKSERVWT4HMPU65YPFJB2US2Y","download_json":"https://pith.science/pith/MGKSERVWT4HMPU65YPFJB2US2Y.json","view_paper":"https://pith.science/paper/MGKSERVW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.07096&json=true","fetch_graph":"https://pith.science/api/pith-number/MGKSERVWT4HMPU65YPFJB2US2Y/graph.json","fetch_events":"https://pith.science/api/pith-number/MGKSERVWT4HMPU65YPFJB2US2Y/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MGKSERVWT4HMPU65YPFJB2US2Y/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MGKSERVWT4HMPU65YPFJB2US2Y/action/storage_attestation","attest_author":"https://pith.science/pith/MGKSERVWT4HMPU65YPFJB2US2Y/action/author_attestation","sign_citation":"https://pith.science/pith/MGKSERVWT4HMPU65YPFJB2US2Y/action/citation_signature","submit_replication":"https://pith.science/pith/MGKSERVWT4HMPU65YPFJB2US2Y/action/replication_record"}},"created_at":"2026-07-05T06:59:40.431255+00:00","updated_at":"2026-07-05T06:59:40.431255+00:00"}