{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:V7IYH7PDEBWHDLCGHH3LFFLPT2","short_pith_number":"pith:V7IYH7PD","schema_version":"1.0","canonical_sha256":"afd183fde3206c71ac4639f6b2956f9e83df2cae0fc3b61265fb997640d00cbe","source":{"kind":"arxiv","id":"2503.07604","version":3},"attestation_state":"computed","paper":{"title":"Implicit Reasoning in Transformers is Reasoning through Shortcuts","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Deqing Yang, Jian Xie, Siyu Yuan, Tianhe Lin","submitted_at":"2025-03-10T17:58:31Z","abstract_excerpt":"Test-time compute is emerging as a new paradigm for enhancing language models' complex multi-step reasoning capabilities, as demonstrated by the success of OpenAI's o1 and o3, as well as DeepSeek's R1. Compared to explicit reasoning in test-time compute, implicit reasoning is more inference-efficient, requiring fewer generated tokens. However, why does the advanced reasoning capability fail to emerge in the implicit reasoning style? In this work, we train GPT-2 from scratch on a curated multi-step mathematical reasoning dataset and conduct analytical experiments to investigate how language mod"},"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":"2503.07604","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-03-10T17:58:31Z","cross_cats_sorted":[],"title_canon_sha256":"912a40ccac1c24ccc03e89bf51f2570caf7fa2daaf2001745e7cab11cbdf27c1","abstract_canon_sha256":"626993830e878d1e36827dc6f4542e84ca1cacff275b36cbe75af9a37f3c55dd"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:14:05.069522Z","signature_b64":"j9GvHXG+cL4NYvJ+7UdmLj++hC3kf1gDyqv9/cVWjzUYQiWExZFSCRrd+qL+HinWkEmeat955g7fe81IjNC4Dw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"afd183fde3206c71ac4639f6b2956f9e83df2cae0fc3b61265fb997640d00cbe","last_reissued_at":"2026-07-05T11:14:05.069032Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:14:05.069032Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Implicit Reasoning in Transformers is Reasoning through Shortcuts","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Deqing Yang, Jian Xie, Siyu Yuan, Tianhe Lin","submitted_at":"2025-03-10T17:58:31Z","abstract_excerpt":"Test-time compute is emerging as a new paradigm for enhancing language models' complex multi-step reasoning capabilities, as demonstrated by the success of OpenAI's o1 and o3, as well as DeepSeek's R1. Compared to explicit reasoning in test-time compute, implicit reasoning is more inference-efficient, requiring fewer generated tokens. However, why does the advanced reasoning capability fail to emerge in the implicit reasoning style? In this work, we train GPT-2 from scratch on a curated multi-step mathematical reasoning dataset and conduct analytical experiments to investigate how language mod"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2503.07604","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/2503.07604/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":"2503.07604","created_at":"2026-07-05T11:14:05.069091+00:00"},{"alias_kind":"arxiv_version","alias_value":"2503.07604v3","created_at":"2026-07-05T11:14:05.069091+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2503.07604","created_at":"2026-07-05T11:14:05.069091+00:00"},{"alias_kind":"pith_short_12","alias_value":"V7IYH7PDEBWH","created_at":"2026-07-05T11:14:05.069091+00:00"},{"alias_kind":"pith_short_16","alias_value":"V7IYH7PDEBWHDLCG","created_at":"2026-07-05T11:14:05.069091+00:00"},{"alias_kind":"pith_short_8","alias_value":"V7IYH7PD","created_at":"2026-07-05T11:14:05.069091+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.03058","citing_title":"Neuron-Anchored Rule Extraction for Large Language Models via Contrastive Hierarchical Ablation","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2504.01990","citing_title":"Advances and Challenges in Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systems","ref_index":197,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03058","citing_title":"Neuron-Anchored Rule Extraction for Large Language Models via Contrastive Hierarchical Ablation","ref_index":28,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/V7IYH7PDEBWHDLCGHH3LFFLPT2","json":"https://pith.science/pith/V7IYH7PDEBWHDLCGHH3LFFLPT2.json","graph_json":"https://pith.science/api/pith-number/V7IYH7PDEBWHDLCGHH3LFFLPT2/graph.json","events_json":"https://pith.science/api/pith-number/V7IYH7PDEBWHDLCGHH3LFFLPT2/events.json","paper":"https://pith.science/paper/V7IYH7PD"},"agent_actions":{"view_html":"https://pith.science/pith/V7IYH7PDEBWHDLCGHH3LFFLPT2","download_json":"https://pith.science/pith/V7IYH7PDEBWHDLCGHH3LFFLPT2.json","view_paper":"https://pith.science/paper/V7IYH7PD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2503.07604&json=true","fetch_graph":"https://pith.science/api/pith-number/V7IYH7PDEBWHDLCGHH3LFFLPT2/graph.json","fetch_events":"https://pith.science/api/pith-number/V7IYH7PDEBWHDLCGHH3LFFLPT2/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/V7IYH7PDEBWHDLCGHH3LFFLPT2/action/timestamp_anchor","attest_storage":"https://pith.science/pith/V7IYH7PDEBWHDLCGHH3LFFLPT2/action/storage_attestation","attest_author":"https://pith.science/pith/V7IYH7PDEBWHDLCGHH3LFFLPT2/action/author_attestation","sign_citation":"https://pith.science/pith/V7IYH7PDEBWHDLCGHH3LFFLPT2/action/citation_signature","submit_replication":"https://pith.science/pith/V7IYH7PDEBWHDLCGHH3LFFLPT2/action/replication_record"}},"created_at":"2026-07-05T11:14:05.069091+00:00","updated_at":"2026-07-05T11:14:05.069091+00:00"}