{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:MBY6CFOV5EG5PY4LNXJR3CM7SI","short_pith_number":"pith:MBY6CFOV","schema_version":"1.0","canonical_sha256":"6071e115d5e90dd7e38b6dd31d899f92346ef867b08d9d4873c1e973f5fa9fc7","source":{"kind":"arxiv","id":"2103.13033","version":1},"attestation_state":"computed","paper":{"title":"Thinking Aloud: Dynamic Context Generation Improves Zero-Shot Reasoning Performance of GPT-2","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Christian Voigt, Gregor Betz, Kyle Richardson","submitted_at":"2021-03-24T07:33:25Z","abstract_excerpt":"Thinking aloud is an effective meta-cognitive strategy human reasoners apply to solve difficult problems. We suggest to improve the reasoning ability of pre-trained neural language models in a similar way, namely by expanding a task's context with problem elaborations that are dynamically generated by the language model itself. Our main result is that dynamic problem elaboration significantly improves the zero-shot performance of GPT-2 in a deductive reasoning and natural language inference task: While the model uses a syntactic heuristic for predicting an answer, it is capable (to some degree"},"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":"2103.13033","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2021-03-24T07:33:25Z","cross_cats_sorted":[],"title_canon_sha256":"57482c44c9298ece0c4e67acf7dc3e4fcb987a2ef79659406fc1bcb4398144c9","abstract_canon_sha256":"fc84250acd47d45fac418f1347ff2f1f81b8fc26683e4820e4921e86a4ed7bd2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:26:04.927391Z","signature_b64":"d9FGGkczlzaHluT82Olb9Fa9IuJRwEPi2iyneGbL1MLrg7h/vYXULXrbJt6GVlgUyShDC6V5MQ15dxg+9FhOAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6071e115d5e90dd7e38b6dd31d899f92346ef867b08d9d4873c1e973f5fa9fc7","last_reissued_at":"2026-07-05T02:26:04.927015Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:26:04.927015Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Thinking Aloud: Dynamic Context Generation Improves Zero-Shot Reasoning Performance of GPT-2","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Christian Voigt, Gregor Betz, Kyle Richardson","submitted_at":"2021-03-24T07:33:25Z","abstract_excerpt":"Thinking aloud is an effective meta-cognitive strategy human reasoners apply to solve difficult problems. We suggest to improve the reasoning ability of pre-trained neural language models in a similar way, namely by expanding a task's context with problem elaborations that are dynamically generated by the language model itself. Our main result is that dynamic problem elaboration significantly improves the zero-shot performance of GPT-2 in a deductive reasoning and natural language inference task: While the model uses a syntactic heuristic for predicting an answer, it is capable (to some degree"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2103.13033","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/2103.13033/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":"2103.13033","created_at":"2026-07-05T02:26:04.927078+00:00"},{"alias_kind":"arxiv_version","alias_value":"2103.13033v1","created_at":"2026-07-05T02:26:04.927078+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2103.13033","created_at":"2026-07-05T02:26:04.927078+00:00"},{"alias_kind":"pith_short_12","alias_value":"MBY6CFOV5EG5","created_at":"2026-07-05T02:26:04.927078+00:00"},{"alias_kind":"pith_short_16","alias_value":"MBY6CFOV5EG5PY4L","created_at":"2026-07-05T02:26:04.927078+00:00"},{"alias_kind":"pith_short_8","alias_value":"MBY6CFOV","created_at":"2026-07-05T02:26:04.927078+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2211.01910","citing_title":"Large Language Models Are Human-Level Prompt Engineers","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MBY6CFOV5EG5PY4LNXJR3CM7SI","json":"https://pith.science/pith/MBY6CFOV5EG5PY4LNXJR3CM7SI.json","graph_json":"https://pith.science/api/pith-number/MBY6CFOV5EG5PY4LNXJR3CM7SI/graph.json","events_json":"https://pith.science/api/pith-number/MBY6CFOV5EG5PY4LNXJR3CM7SI/events.json","paper":"https://pith.science/paper/MBY6CFOV"},"agent_actions":{"view_html":"https://pith.science/pith/MBY6CFOV5EG5PY4LNXJR3CM7SI","download_json":"https://pith.science/pith/MBY6CFOV5EG5PY4LNXJR3CM7SI.json","view_paper":"https://pith.science/paper/MBY6CFOV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2103.13033&json=true","fetch_graph":"https://pith.science/api/pith-number/MBY6CFOV5EG5PY4LNXJR3CM7SI/graph.json","fetch_events":"https://pith.science/api/pith-number/MBY6CFOV5EG5PY4LNXJR3CM7SI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MBY6CFOV5EG5PY4LNXJR3CM7SI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MBY6CFOV5EG5PY4LNXJR3CM7SI/action/storage_attestation","attest_author":"https://pith.science/pith/MBY6CFOV5EG5PY4LNXJR3CM7SI/action/author_attestation","sign_citation":"https://pith.science/pith/MBY6CFOV5EG5PY4LNXJR3CM7SI/action/citation_signature","submit_replication":"https://pith.science/pith/MBY6CFOV5EG5PY4LNXJR3CM7SI/action/replication_record"}},"created_at":"2026-07-05T02:26:04.927078+00:00","updated_at":"2026-07-05T02:26:04.927078+00:00"}