{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:RGJSUULCIEBJQ7LCAXFRTI5QXC","short_pith_number":"pith:RGJSUULC","schema_version":"1.0","canonical_sha256":"89932a51624102987d6205cb19a3b0b8bbe9546143df81d7b139c59b26f2553e","source":{"kind":"arxiv","id":"2502.19981","version":1},"attestation_state":"computed","paper":{"title":"The Lookahead Limitation: Why Multi-Operand Addition is Hard for LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Josef van Genabith, Simon Ostermann, Tanja Baeumel","submitted_at":"2025-02-27T11:03:27Z","abstract_excerpt":"Autoregressive large language models (LLMs) exhibit impressive performance across various tasks but struggle with simple arithmetic, such as addition of two or more operands. We show that this struggle arises from LLMs' use of a simple one-digit lookahead heuristic, which works fairly well (but not perfect) for two-operand addition but fails in multi-operand cases, where the carry-over logic is more complex. Our probing experiments and digit-wise accuracy evaluation show that LLMs fail precisely where a one-digit lookahead is insufficient to account for cascading carries. We analyze the impact"},"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":"2502.19981","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-02-27T11:03:27Z","cross_cats_sorted":[],"title_canon_sha256":"b7daeafb9ed32a9fb009e3873469df1bf125aeb10612514fd444b374971e9ab3","abstract_canon_sha256":"3cbd8b175010f63ee8c81d9413aaeed696ef1a92ad149f757c30dd7accf45c04"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:20:56.298701Z","signature_b64":"VpUIN8Ku9SRlSdU9H8zPxXXzfXSu9IB9mJEOYJ54sHM2gBr5xVB+hsW4dKPrSB/4dDHImOC7GuWa2aTzPU8LAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"89932a51624102987d6205cb19a3b0b8bbe9546143df81d7b139c59b26f2553e","last_reissued_at":"2026-07-05T10:20:56.298111Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:20:56.298111Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Lookahead Limitation: Why Multi-Operand Addition is Hard for LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Josef van Genabith, Simon Ostermann, Tanja Baeumel","submitted_at":"2025-02-27T11:03:27Z","abstract_excerpt":"Autoregressive large language models (LLMs) exhibit impressive performance across various tasks but struggle with simple arithmetic, such as addition of two or more operands. We show that this struggle arises from LLMs' use of a simple one-digit lookahead heuristic, which works fairly well (but not perfect) for two-operand addition but fails in multi-operand cases, where the carry-over logic is more complex. Our probing experiments and digit-wise accuracy evaluation show that LLMs fail precisely where a one-digit lookahead is insufficient to account for cascading carries. We analyze the impact"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.19981","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/2502.19981/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":"2502.19981","created_at":"2026-07-05T10:20:56.298179+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.19981v1","created_at":"2026-07-05T10:20:56.298179+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.19981","created_at":"2026-07-05T10:20:56.298179+00:00"},{"alias_kind":"pith_short_12","alias_value":"RGJSUULCIEBJ","created_at":"2026-07-05T10:20:56.298179+00:00"},{"alias_kind":"pith_short_16","alias_value":"RGJSUULCIEBJQ7LC","created_at":"2026-07-05T10:20:56.298179+00:00"},{"alias_kind":"pith_short_8","alias_value":"RGJSUULC","created_at":"2026-07-05T10:20:56.298179+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.03645","citing_title":"The Shape of Addition: Geometric Structures of Arithmetic in Large Language Models","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2510.06824","citing_title":"Efficient numeracy in language models through single-token number embeddings","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2503.09567","citing_title":"Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Models","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05851","citing_title":"Hypothesis generation and updating in large language models","ref_index":40,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RGJSUULCIEBJQ7LCAXFRTI5QXC","json":"https://pith.science/pith/RGJSUULCIEBJQ7LCAXFRTI5QXC.json","graph_json":"https://pith.science/api/pith-number/RGJSUULCIEBJQ7LCAXFRTI5QXC/graph.json","events_json":"https://pith.science/api/pith-number/RGJSUULCIEBJQ7LCAXFRTI5QXC/events.json","paper":"https://pith.science/paper/RGJSUULC"},"agent_actions":{"view_html":"https://pith.science/pith/RGJSUULCIEBJQ7LCAXFRTI5QXC","download_json":"https://pith.science/pith/RGJSUULCIEBJQ7LCAXFRTI5QXC.json","view_paper":"https://pith.science/paper/RGJSUULC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.19981&json=true","fetch_graph":"https://pith.science/api/pith-number/RGJSUULCIEBJQ7LCAXFRTI5QXC/graph.json","fetch_events":"https://pith.science/api/pith-number/RGJSUULCIEBJQ7LCAXFRTI5QXC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RGJSUULCIEBJQ7LCAXFRTI5QXC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RGJSUULCIEBJQ7LCAXFRTI5QXC/action/storage_attestation","attest_author":"https://pith.science/pith/RGJSUULCIEBJQ7LCAXFRTI5QXC/action/author_attestation","sign_citation":"https://pith.science/pith/RGJSUULCIEBJQ7LCAXFRTI5QXC/action/citation_signature","submit_replication":"https://pith.science/pith/RGJSUULCIEBJQ7LCAXFRTI5QXC/action/replication_record"}},"created_at":"2026-07-05T10:20:56.298179+00:00","updated_at":"2026-07-05T10:20:56.298179+00:00"}