{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:XERAXLQLYTAYCRLH2NXQ3DRNTY","short_pith_number":"pith:XERAXLQL","schema_version":"1.0","canonical_sha256":"b9220bae0bc4c1814567d36f0d8e2d9e13ec6809706a700284d71955a7d86b13","source":{"kind":"arxiv","id":"2412.05117","version":2},"attestation_state":"computed","paper":{"title":"Transformers Can Navigate Mazes With Multi-Step Prediction","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Adina Williams, Mark Ibrahim, Mike Rabbat, Niklas Nolte, Ouail Kitouni","submitted_at":"2024-12-06T15:19:10Z","abstract_excerpt":"Despite their remarkable success in language modeling, transformers trained to predict the next token in a sequence struggle with long-term planning. This limitation is particularly evident in tasks requiring foresight to plan multiple steps ahead such as maze navigation. The standard next single token prediction objective, however, offers no explicit mechanism to predict multiple steps ahead - or revisit the path taken so far. Consequently, in this work we study whether explicitly predicting multiple steps ahead (and backwards) can improve transformers' maze navigation. We train parameter-mat"},"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":"2412.05117","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-06T15:19:10Z","cross_cats_sorted":[],"title_canon_sha256":"a76db19bf63111fc3c8f7154e585567e0519b69ca9c65d8b5654eaf6a93d755f","abstract_canon_sha256":"44e26385503395e760fb54debf948f3160abc66674e09a7bd7783ee9fa3d3d9e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:50:54.829245Z","signature_b64":"8Rky+i3j6B+PL97fzrgvvscH0QPrfLNPGFR+CIjneBkbzCD9AX579u4CgC/M6Y0zLVFa2SM4KB9MB3XxHT7JCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b9220bae0bc4c1814567d36f0d8e2d9e13ec6809706a700284d71955a7d86b13","last_reissued_at":"2026-07-05T09:50:54.828747Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:50:54.828747Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Transformers Can Navigate Mazes With Multi-Step Prediction","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Adina Williams, Mark Ibrahim, Mike Rabbat, Niklas Nolte, Ouail Kitouni","submitted_at":"2024-12-06T15:19:10Z","abstract_excerpt":"Despite their remarkable success in language modeling, transformers trained to predict the next token in a sequence struggle with long-term planning. This limitation is particularly evident in tasks requiring foresight to plan multiple steps ahead such as maze navigation. The standard next single token prediction objective, however, offers no explicit mechanism to predict multiple steps ahead - or revisit the path taken so far. Consequently, in this work we study whether explicitly predicting multiple steps ahead (and backwards) can improve transformers' maze navigation. We train parameter-mat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.05117","kind":"arxiv","version":2},"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/2412.05117/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":"2412.05117","created_at":"2026-07-05T09:50:54.828803+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.05117v2","created_at":"2026-07-05T09:50:54.828803+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.05117","created_at":"2026-07-05T09:50:54.828803+00:00"},{"alias_kind":"pith_short_12","alias_value":"XERAXLQLYTAY","created_at":"2026-07-05T09:50:54.828803+00:00"},{"alias_kind":"pith_short_16","alias_value":"XERAXLQLYTAYCRLH","created_at":"2026-07-05T09:50:54.828803+00:00"},{"alias_kind":"pith_short_8","alias_value":"XERAXLQL","created_at":"2026-07-05T09:50:54.828803+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/XERAXLQLYTAYCRLH2NXQ3DRNTY","json":"https://pith.science/pith/XERAXLQLYTAYCRLH2NXQ3DRNTY.json","graph_json":"https://pith.science/api/pith-number/XERAXLQLYTAYCRLH2NXQ3DRNTY/graph.json","events_json":"https://pith.science/api/pith-number/XERAXLQLYTAYCRLH2NXQ3DRNTY/events.json","paper":"https://pith.science/paper/XERAXLQL"},"agent_actions":{"view_html":"https://pith.science/pith/XERAXLQLYTAYCRLH2NXQ3DRNTY","download_json":"https://pith.science/pith/XERAXLQLYTAYCRLH2NXQ3DRNTY.json","view_paper":"https://pith.science/paper/XERAXLQL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.05117&json=true","fetch_graph":"https://pith.science/api/pith-number/XERAXLQLYTAYCRLH2NXQ3DRNTY/graph.json","fetch_events":"https://pith.science/api/pith-number/XERAXLQLYTAYCRLH2NXQ3DRNTY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XERAXLQLYTAYCRLH2NXQ3DRNTY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XERAXLQLYTAYCRLH2NXQ3DRNTY/action/storage_attestation","attest_author":"https://pith.science/pith/XERAXLQLYTAYCRLH2NXQ3DRNTY/action/author_attestation","sign_citation":"https://pith.science/pith/XERAXLQLYTAYCRLH2NXQ3DRNTY/action/citation_signature","submit_replication":"https://pith.science/pith/XERAXLQLYTAYCRLH2NXQ3DRNTY/action/replication_record"}},"created_at":"2026-07-05T09:50:54.828803+00:00","updated_at":"2026-07-05T09:50:54.828803+00:00"}