{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:4ZAR7TLWGHFL6ZGF23ABLLBEFH","short_pith_number":"pith:4ZAR7TLW","schema_version":"1.0","canonical_sha256":"e6411fcd7631cabf64c5d6c015ac2429d8570cf6f75c881c0bef9f38f19849f7","source":{"kind":"arxiv","id":"2407.15720","version":2},"attestation_state":"computed","paper":{"title":"Do Large Language Models Have Compositional Ability? An Investigation into Limitations and Scalability","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Yingyu Liang, Zhenmei Shi, Zhuoyan Xu","submitted_at":"2024-07-22T15:22:34Z","abstract_excerpt":"Large language models (LLMs) have emerged as powerful tools for many AI problems and exhibit remarkable in-context learning (ICL) capabilities. Compositional ability, solving unseen complex tasks that combine two or more simple tasks, is an essential reasoning ability for Artificial General Intelligence. Despite the tremendous success of LLMs, how they approach composite tasks, especially those not encountered during the pretraining phase, remains an open and largely underexplored question. In this study, we delve into the ICL capabilities of LLMs on composite tasks, with only simple tasks as "},"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":"2407.15720","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-07-22T15:22:34Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"b86c9f4ecffc3d6b092749f0b89e7a993b1e7d7fae0438dc70b4880370637300","abstract_canon_sha256":"f38568ce1595cb98079bcaa591b4a1fea0f3af0d071924119f5bdb4035bc8241"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:54:16.015636Z","signature_b64":"yenQ0H0NMXtXQQIYO7n4P5blkj2PLs5nu6tTVAzjFN7y+BqFV+TFeOOaV+Y5nwIY1HPYi3JKeEe+4f0e2RpqDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e6411fcd7631cabf64c5d6c015ac2429d8570cf6f75c881c0bef9f38f19849f7","last_reissued_at":"2026-07-05T08:54:16.015164Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:54:16.015164Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Do Large Language Models Have Compositional Ability? An Investigation into Limitations and Scalability","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Yingyu Liang, Zhenmei Shi, Zhuoyan Xu","submitted_at":"2024-07-22T15:22:34Z","abstract_excerpt":"Large language models (LLMs) have emerged as powerful tools for many AI problems and exhibit remarkable in-context learning (ICL) capabilities. Compositional ability, solving unseen complex tasks that combine two or more simple tasks, is an essential reasoning ability for Artificial General Intelligence. Despite the tremendous success of LLMs, how they approach composite tasks, especially those not encountered during the pretraining phase, remains an open and largely underexplored question. In this study, we delve into the ICL capabilities of LLMs on composite tasks, with only simple tasks as "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2407.15720","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/2407.15720/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":"2407.15720","created_at":"2026-07-05T08:54:16.015220+00:00"},{"alias_kind":"arxiv_version","alias_value":"2407.15720v2","created_at":"2026-07-05T08:54:16.015220+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2407.15720","created_at":"2026-07-05T08:54:16.015220+00:00"},{"alias_kind":"pith_short_12","alias_value":"4ZAR7TLWGHFL","created_at":"2026-07-05T08:54:16.015220+00:00"},{"alias_kind":"pith_short_16","alias_value":"4ZAR7TLWGHFL6ZGF","created_at":"2026-07-05T08:54:16.015220+00:00"},{"alias_kind":"pith_short_8","alias_value":"4ZAR7TLW","created_at":"2026-07-05T08:54:16.015220+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.09338","citing_title":"Multi-Hop Knowledge Composition is Bound by Pretraining Exposure","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2511.02627","citing_title":"DecompSR: A dataset for decomposed analyses of compositional multihop spatial reasoning","ref_index":23,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4ZAR7TLWGHFL6ZGF23ABLLBEFH","json":"https://pith.science/pith/4ZAR7TLWGHFL6ZGF23ABLLBEFH.json","graph_json":"https://pith.science/api/pith-number/4ZAR7TLWGHFL6ZGF23ABLLBEFH/graph.json","events_json":"https://pith.science/api/pith-number/4ZAR7TLWGHFL6ZGF23ABLLBEFH/events.json","paper":"https://pith.science/paper/4ZAR7TLW"},"agent_actions":{"view_html":"https://pith.science/pith/4ZAR7TLWGHFL6ZGF23ABLLBEFH","download_json":"https://pith.science/pith/4ZAR7TLWGHFL6ZGF23ABLLBEFH.json","view_paper":"https://pith.science/paper/4ZAR7TLW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2407.15720&json=true","fetch_graph":"https://pith.science/api/pith-number/4ZAR7TLWGHFL6ZGF23ABLLBEFH/graph.json","fetch_events":"https://pith.science/api/pith-number/4ZAR7TLWGHFL6ZGF23ABLLBEFH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4ZAR7TLWGHFL6ZGF23ABLLBEFH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4ZAR7TLWGHFL6ZGF23ABLLBEFH/action/storage_attestation","attest_author":"https://pith.science/pith/4ZAR7TLWGHFL6ZGF23ABLLBEFH/action/author_attestation","sign_citation":"https://pith.science/pith/4ZAR7TLWGHFL6ZGF23ABLLBEFH/action/citation_signature","submit_replication":"https://pith.science/pith/4ZAR7TLWGHFL6ZGF23ABLLBEFH/action/replication_record"}},"created_at":"2026-07-05T08:54:16.015220+00:00","updated_at":"2026-07-05T08:54:16.015220+00:00"}