{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:W3REE26YVOE7PLRKK6IGV2XNA6","short_pith_number":"pith:W3REE26Y","schema_version":"1.0","canonical_sha256":"b6e2426bd8ab89f7ae2a57906aeaed07a9b197978e6912b99f5c66e78a4cc432","source":{"kind":"arxiv","id":"2505.23701","version":1},"attestation_state":"computed","paper":{"title":"Can LLMs Reason Abstractly Over Math Word Problems Without CoT? Disentangling Abstract Formulation From Arithmetic Computation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jackie Chi Kit Cheung, Leila Pishdad, Meng Cao, Yanshuai Cao, Ziling Cheng","submitted_at":"2025-05-29T17:37:57Z","abstract_excerpt":"Final-answer-based metrics are commonly used for evaluating large language models (LLMs) on math word problems, often taken as proxies for reasoning ability. However, such metrics conflate two distinct sub-skills: abstract formulation (capturing mathematical relationships using expressions) and arithmetic computation (executing the calculations). Through a disentangled evaluation on GSM8K and SVAMP, we find that the final-answer accuracy of Llama-3 and Qwen2.5 (1B-32B) without CoT is overwhelmingly bottlenecked by the arithmetic computation step and not by the abstract formulation step. Contra"},"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":"2505.23701","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-05-29T17:37:57Z","cross_cats_sorted":[],"title_canon_sha256":"cb6308e19d5989c4975508c9df5f32cef9a0c31dc8fdfa7eef289b0eac4b8712","abstract_canon_sha256":"589493eb6ab5c3a18b5ca1064d927ff9e94fa3c8f1da8b827ae967fe834101f7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:12:12.889576Z","signature_b64":"yHnBfCGXDePxsxHqWXXpm3Fz3W0cxCYRn7sdcVThENsnypWjP0M+1DbjaasqDXLuvYFTzSBPucz8SJI/as7sDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b6e2426bd8ab89f7ae2a57906aeaed07a9b197978e6912b99f5c66e78a4cc432","last_reissued_at":"2026-07-05T11:12:12.889099Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:12:12.889099Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Can LLMs Reason Abstractly Over Math Word Problems Without CoT? Disentangling Abstract Formulation From Arithmetic Computation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jackie Chi Kit Cheung, Leila Pishdad, Meng Cao, Yanshuai Cao, Ziling Cheng","submitted_at":"2025-05-29T17:37:57Z","abstract_excerpt":"Final-answer-based metrics are commonly used for evaluating large language models (LLMs) on math word problems, often taken as proxies for reasoning ability. However, such metrics conflate two distinct sub-skills: abstract formulation (capturing mathematical relationships using expressions) and arithmetic computation (executing the calculations). Through a disentangled evaluation on GSM8K and SVAMP, we find that the final-answer accuracy of Llama-3 and Qwen2.5 (1B-32B) without CoT is overwhelmingly bottlenecked by the arithmetic computation step and not by the abstract formulation step. Contra"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.23701","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/2505.23701/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":"2505.23701","created_at":"2026-07-05T11:12:12.889152+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.23701v1","created_at":"2026-07-05T11:12:12.889152+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.23701","created_at":"2026-07-05T11:12:12.889152+00:00"},{"alias_kind":"pith_short_12","alias_value":"W3REE26YVOE7","created_at":"2026-07-05T11:12:12.889152+00:00"},{"alias_kind":"pith_short_16","alias_value":"W3REE26YVOE7PLRK","created_at":"2026-07-05T11:12:12.889152+00:00"},{"alias_kind":"pith_short_8","alias_value":"W3REE26Y","created_at":"2026-07-05T11:12:12.889152+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.22630","citing_title":"Stochastic Chameleons: Irrelevant Context Hallucinations Reveal Class-Based (Mis)Generalization in LLMs","ref_index":13,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/W3REE26YVOE7PLRKK6IGV2XNA6","json":"https://pith.science/pith/W3REE26YVOE7PLRKK6IGV2XNA6.json","graph_json":"https://pith.science/api/pith-number/W3REE26YVOE7PLRKK6IGV2XNA6/graph.json","events_json":"https://pith.science/api/pith-number/W3REE26YVOE7PLRKK6IGV2XNA6/events.json","paper":"https://pith.science/paper/W3REE26Y"},"agent_actions":{"view_html":"https://pith.science/pith/W3REE26YVOE7PLRKK6IGV2XNA6","download_json":"https://pith.science/pith/W3REE26YVOE7PLRKK6IGV2XNA6.json","view_paper":"https://pith.science/paper/W3REE26Y","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.23701&json=true","fetch_graph":"https://pith.science/api/pith-number/W3REE26YVOE7PLRKK6IGV2XNA6/graph.json","fetch_events":"https://pith.science/api/pith-number/W3REE26YVOE7PLRKK6IGV2XNA6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/W3REE26YVOE7PLRKK6IGV2XNA6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/W3REE26YVOE7PLRKK6IGV2XNA6/action/storage_attestation","attest_author":"https://pith.science/pith/W3REE26YVOE7PLRKK6IGV2XNA6/action/author_attestation","sign_citation":"https://pith.science/pith/W3REE26YVOE7PLRKK6IGV2XNA6/action/citation_signature","submit_replication":"https://pith.science/pith/W3REE26YVOE7PLRKK6IGV2XNA6/action/replication_record"}},"created_at":"2026-07-05T11:12:12.889152+00:00","updated_at":"2026-07-05T11:12:12.889152+00:00"}