{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:F5ZBIQXZXUAQF5FAYUOX6HFSDF","short_pith_number":"pith:F5ZBIQXZ","schema_version":"1.0","canonical_sha256":"2f721442f9bd0102f4a0c51d7f1cb2197fbeb106be65e093886e6b318e84ccfa","source":{"kind":"arxiv","id":"2502.08794","version":1},"attestation_state":"computed","paper":{"title":"Spectral Journey: How Transformers Predict the Shortest Path","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Andrew Cohen, Andrey Gromov, Kaiyu Yang, Yuandong Tian","submitted_at":"2025-02-12T21:17:30Z","abstract_excerpt":"Decoder-only transformers lead to a step-change in capability of large language models. However, opinions are mixed as to whether they are really planning or reasoning. A path to making progress in this direction is to study the model's behavior in a setting with carefully controlled data. Then interpret the learned representations and reverse-engineer the computation performed internally. We study decoder-only transformer language models trained from scratch to predict shortest paths on simple, connected and undirected graphs. In this setting, the representations and the dynamics learned by t"},"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":"2502.08794","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-12T21:17:30Z","cross_cats_sorted":[],"title_canon_sha256":"36100fc02d12f0c0d0d6a0fcf12c10fcf94d8f6ef965897b6c6a1e57b8d3381d","abstract_canon_sha256":"9093c2c461944be2be46f5e4f22bf326170b46fb4fcf2899c6440a12bbb2c4bf"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:13:29.951251Z","signature_b64":"u3r3bazULMfha7Rz3FA0DgzuttLMkBhBVET9ydgvgcN7ameWbfJyVu8WGWQDtB5WQgRlgn7KbUywXKijLCwuDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2f721442f9bd0102f4a0c51d7f1cb2197fbeb106be65e093886e6b318e84ccfa","last_reissued_at":"2026-07-05T10:13:29.950759Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:13:29.950759Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Spectral Journey: How Transformers Predict the Shortest Path","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Andrew Cohen, Andrey Gromov, Kaiyu Yang, Yuandong Tian","submitted_at":"2025-02-12T21:17:30Z","abstract_excerpt":"Decoder-only transformers lead to a step-change in capability of large language models. However, opinions are mixed as to whether they are really planning or reasoning. A path to making progress in this direction is to study the model's behavior in a setting with carefully controlled data. Then interpret the learned representations and reverse-engineer the computation performed internally. We study decoder-only transformer language models trained from scratch to predict shortest paths on simple, connected and undirected graphs. In this setting, the representations and the dynamics learned by t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.08794","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/2502.08794/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":"2502.08794","created_at":"2026-07-05T10:13:29.950818+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.08794v1","created_at":"2026-07-05T10:13:29.950818+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.08794","created_at":"2026-07-05T10:13:29.950818+00:00"},{"alias_kind":"pith_short_12","alias_value":"F5ZBIQXZXUAQ","created_at":"2026-07-05T10:13:29.950818+00:00"},{"alias_kind":"pith_short_16","alias_value":"F5ZBIQXZXUAQF5FA","created_at":"2026-07-05T10:13:29.950818+00:00"},{"alias_kind":"pith_short_8","alias_value":"F5ZBIQXZ","created_at":"2026-07-05T10:13:29.950818+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.15306","citing_title":"Generalization in LLM Problem Solving: The Case of the Shortest Path","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/F5ZBIQXZXUAQF5FAYUOX6HFSDF","json":"https://pith.science/pith/F5ZBIQXZXUAQF5FAYUOX6HFSDF.json","graph_json":"https://pith.science/api/pith-number/F5ZBIQXZXUAQF5FAYUOX6HFSDF/graph.json","events_json":"https://pith.science/api/pith-number/F5ZBIQXZXUAQF5FAYUOX6HFSDF/events.json","paper":"https://pith.science/paper/F5ZBIQXZ"},"agent_actions":{"view_html":"https://pith.science/pith/F5ZBIQXZXUAQF5FAYUOX6HFSDF","download_json":"https://pith.science/pith/F5ZBIQXZXUAQF5FAYUOX6HFSDF.json","view_paper":"https://pith.science/paper/F5ZBIQXZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.08794&json=true","fetch_graph":"https://pith.science/api/pith-number/F5ZBIQXZXUAQF5FAYUOX6HFSDF/graph.json","fetch_events":"https://pith.science/api/pith-number/F5ZBIQXZXUAQF5FAYUOX6HFSDF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/F5ZBIQXZXUAQF5FAYUOX6HFSDF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/F5ZBIQXZXUAQF5FAYUOX6HFSDF/action/storage_attestation","attest_author":"https://pith.science/pith/F5ZBIQXZXUAQF5FAYUOX6HFSDF/action/author_attestation","sign_citation":"https://pith.science/pith/F5ZBIQXZXUAQF5FAYUOX6HFSDF/action/citation_signature","submit_replication":"https://pith.science/pith/F5ZBIQXZXUAQF5FAYUOX6HFSDF/action/replication_record"}},"created_at":"2026-07-05T10:13:29.950818+00:00","updated_at":"2026-07-05T10:13:29.950818+00:00"}