{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:4GTS5QFZCUC7MXFLPBAFJ7M6ZI","short_pith_number":"pith:4GTS5QFZ","schema_version":"1.0","canonical_sha256":"e1a72ec0b91505f65cab784054fd9eca15cc76957929688a57c489a585f4a4ef","source":{"kind":"arxiv","id":"2406.06467","version":3},"attestation_state":"computed","paper":{"title":"How Far Can Transformers Reason? The Globality Barrier and Inductive Scratchpad","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Aryo Lotfi, Colin Sandon, Emmanuel Abbe, Omid Saremi, Samy Bengio","submitted_at":"2024-06-10T17:05:12Z","abstract_excerpt":"Can Transformers predict new syllogisms by composing established ones? More generally, what type of targets can be learned by such models from scratch? Recent works show that Transformers can be Turing-complete in terms of expressivity, but this does not address the learnability objective. This paper puts forward the notion of 'globality degree' of a target distribution to capture when weak learning is efficiently achievable by regular Transformers. This measure shows a contrast with the expressivity results of Transformers captured by $TC^0/TC^1$ classes (further studied here), since the glob"},"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":"2406.06467","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-06-10T17:05:12Z","cross_cats_sorted":["cs.AI","stat.ML"],"title_canon_sha256":"3893d4f38289ac91896004d463842a3e2a72b0073b93b07e998cc38ad9e8f229","abstract_canon_sha256":"26b7b4d8461e7185898c540355619e75a60fb7290b074a7a783de9e216b816cc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:29:49.468940Z","signature_b64":"xp4+Koe8wuYdl6l7lOjcR2DuSu7Dg5FyKDSyWq8xWuzy/C+Afer3DBT9GF4LrI32yM/tpRrt3CkbCZ8yz2I7BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e1a72ec0b91505f65cab784054fd9eca15cc76957929688a57c489a585f4a4ef","last_reissued_at":"2026-07-05T09:29:49.468492Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:29:49.468492Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Far Can Transformers Reason? The Globality Barrier and Inductive Scratchpad","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","stat.ML"],"primary_cat":"cs.LG","authors_text":"Aryo Lotfi, Colin Sandon, Emmanuel Abbe, Omid Saremi, Samy Bengio","submitted_at":"2024-06-10T17:05:12Z","abstract_excerpt":"Can Transformers predict new syllogisms by composing established ones? More generally, what type of targets can be learned by such models from scratch? Recent works show that Transformers can be Turing-complete in terms of expressivity, but this does not address the learnability objective. This paper puts forward the notion of 'globality degree' of a target distribution to capture when weak learning is efficiently achievable by regular Transformers. This measure shows a contrast with the expressivity results of Transformers captured by $TC^0/TC^1$ classes (further studied here), since the glob"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.06467","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/2406.06467/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":"2406.06467","created_at":"2026-07-05T09:29:49.468551+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.06467v3","created_at":"2026-07-05T09:29:49.468551+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.06467","created_at":"2026-07-05T09:29:49.468551+00:00"},{"alias_kind":"pith_short_12","alias_value":"4GTS5QFZCUC7","created_at":"2026-07-05T09:29:49.468551+00:00"},{"alias_kind":"pith_short_16","alias_value":"4GTS5QFZCUC7MXFL","created_at":"2026-07-05T09:29:49.468551+00:00"},{"alias_kind":"pith_short_8","alias_value":"4GTS5QFZ","created_at":"2026-07-05T09:29:49.468551+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.25166","citing_title":"Training Transformers as a Universal Computer","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4GTS5QFZCUC7MXFLPBAFJ7M6ZI","json":"https://pith.science/pith/4GTS5QFZCUC7MXFLPBAFJ7M6ZI.json","graph_json":"https://pith.science/api/pith-number/4GTS5QFZCUC7MXFLPBAFJ7M6ZI/graph.json","events_json":"https://pith.science/api/pith-number/4GTS5QFZCUC7MXFLPBAFJ7M6ZI/events.json","paper":"https://pith.science/paper/4GTS5QFZ"},"agent_actions":{"view_html":"https://pith.science/pith/4GTS5QFZCUC7MXFLPBAFJ7M6ZI","download_json":"https://pith.science/pith/4GTS5QFZCUC7MXFLPBAFJ7M6ZI.json","view_paper":"https://pith.science/paper/4GTS5QFZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.06467&json=true","fetch_graph":"https://pith.science/api/pith-number/4GTS5QFZCUC7MXFLPBAFJ7M6ZI/graph.json","fetch_events":"https://pith.science/api/pith-number/4GTS5QFZCUC7MXFLPBAFJ7M6ZI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4GTS5QFZCUC7MXFLPBAFJ7M6ZI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4GTS5QFZCUC7MXFLPBAFJ7M6ZI/action/storage_attestation","attest_author":"https://pith.science/pith/4GTS5QFZCUC7MXFLPBAFJ7M6ZI/action/author_attestation","sign_citation":"https://pith.science/pith/4GTS5QFZCUC7MXFLPBAFJ7M6ZI/action/citation_signature","submit_replication":"https://pith.science/pith/4GTS5QFZCUC7MXFLPBAFJ7M6ZI/action/replication_record"}},"created_at":"2026-07-05T09:29:49.468551+00:00","updated_at":"2026-07-05T09:29:49.468551+00:00"}