{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ZOGCZMTD6K3FAIJMBVN2CBYKXG","short_pith_number":"pith:ZOGCZMTD","schema_version":"1.0","canonical_sha256":"cb8c2cb263f2b650212c0d5ba1070ab98fd5b9c9347fdd95f3d912a170a3740a","source":{"kind":"arxiv","id":"2608.07261","version":1},"attestation_state":"computed","paper":{"title":"Why Knowing Both Hops Is Not Enough: Understanding Two-Hop Generalization in Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Heng Wang, Herun Wan, Minnan Luo, Yilin Wang, Zili Zhang","submitted_at":"2026-08-07T14:17:34Z","abstract_excerpt":"Large language models (LLMs) can solve complex multi-hop problems yet exhibit puzzling failures on simple two-hop queries: although a model may correctly store each individual hop, it often fails to combine them. To understand the internal mechanisms of this phenomenon, we train transformers from scratch in a controlled symbolic environment. Our experiments reveal a pattern in two-hop generalization: models generalize reliably when the second hop follows the training distribution, but always fail when it deviates.\n  Through mechanistic analysis, we provide a complete explanation for these dist"},"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":"2608.07261","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-08-07T14:17:34Z","cross_cats_sorted":[],"title_canon_sha256":"f0f10ac7da3ad4c31566a900bd77cb4c6fb4b70a3d702a2ea9a3fe44b846dbeb","abstract_canon_sha256":"70c3f8bf6a3b356ab65008d1d9f4229f091f6a8958f5b9d8c84a32b02c47016f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-10T01:14:47.559591Z","signature_b64":"oMOFotCXa/RM3cDdPedEOvN2jRkjb6/mOgde0iHIbwybj7ZKm9/i9A2f0tMHSspWdbT8QATddIBGSOldcYVRBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cb8c2cb263f2b650212c0d5ba1070ab98fd5b9c9347fdd95f3d912a170a3740a","last_reissued_at":"2026-08-10T01:14:47.557243Z","signature_status":"signed_v1","first_computed_at":"2026-08-10T01:14:47.557243Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Why Knowing Both Hops Is Not Enough: Understanding Two-Hop Generalization in Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Heng Wang, Herun Wan, Minnan Luo, Yilin Wang, Zili Zhang","submitted_at":"2026-08-07T14:17:34Z","abstract_excerpt":"Large language models (LLMs) can solve complex multi-hop problems yet exhibit puzzling failures on simple two-hop queries: although a model may correctly store each individual hop, it often fails to combine them. To understand the internal mechanisms of this phenomenon, we train transformers from scratch in a controlled symbolic environment. Our experiments reveal a pattern in two-hop generalization: models generalize reliably when the second hop follows the training distribution, but always fail when it deviates.\n  Through mechanistic analysis, we provide a complete explanation for these dist"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.07261","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/2608.07261/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":"2608.07261","created_at":"2026-08-10T01:14:47.558221+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.07261v1","created_at":"2026-08-10T01:14:47.558221+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.07261","created_at":"2026-08-10T01:14:47.558221+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZOGCZMTD6K3F","created_at":"2026-08-10T01:14:47.558221+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZOGCZMTD6K3FAIJM","created_at":"2026-08-10T01:14:47.558221+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZOGCZMTD","created_at":"2026-08-10T01:14:47.558221+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZOGCZMTD6K3FAIJMBVN2CBYKXG","json":"https://pith.science/pith/ZOGCZMTD6K3FAIJMBVN2CBYKXG.json","graph_json":"https://pith.science/api/pith-number/ZOGCZMTD6K3FAIJMBVN2CBYKXG/graph.json","events_json":"https://pith.science/api/pith-number/ZOGCZMTD6K3FAIJMBVN2CBYKXG/events.json","paper":"https://pith.science/paper/ZOGCZMTD"},"agent_actions":{"view_html":"https://pith.science/pith/ZOGCZMTD6K3FAIJMBVN2CBYKXG","download_json":"https://pith.science/pith/ZOGCZMTD6K3FAIJMBVN2CBYKXG.json","view_paper":"https://pith.science/paper/ZOGCZMTD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.07261&json=true","fetch_graph":"https://pith.science/api/pith-number/ZOGCZMTD6K3FAIJMBVN2CBYKXG/graph.json","fetch_events":"https://pith.science/api/pith-number/ZOGCZMTD6K3FAIJMBVN2CBYKXG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZOGCZMTD6K3FAIJMBVN2CBYKXG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZOGCZMTD6K3FAIJMBVN2CBYKXG/action/storage_attestation","attest_author":"https://pith.science/pith/ZOGCZMTD6K3FAIJMBVN2CBYKXG/action/author_attestation","sign_citation":"https://pith.science/pith/ZOGCZMTD6K3FAIJMBVN2CBYKXG/action/citation_signature","submit_replication":"https://pith.science/pith/ZOGCZMTD6K3FAIJMBVN2CBYKXG/action/replication_record"}},"created_at":"2026-08-10T01:14:47.558221+00:00","updated_at":"2026-08-10T01:14:47.558221+00:00"}