{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:WBLOU7HTZ6KICW5KBO7GYWMWQT","short_pith_number":"pith:WBLOU7HT","schema_version":"1.0","canonical_sha256":"b056ea7cf3cf94815baa0bbe6c599684e5914fe794129121abcaa6a9a269633a","source":{"kind":"arxiv","id":"2303.03846","version":2},"attestation_state":"computed","paper":{"title":"Larger language models do in-context learning differently","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Albert Webson, Da Huang, Denny Zhou, Dustin Tran, Hanxiao Liu, Jason Wei, Jerry Wei, Tengyu Ma, Xinyun Chen, Yifeng Lu, Yi Tay","submitted_at":"2023-03-07T12:24:17Z","abstract_excerpt":"We study how in-context learning (ICL) in language models is affected by semantic priors versus input-label mappings. We investigate two setups-ICL with flipped labels and ICL with semantically-unrelated labels-across various model families (GPT-3, InstructGPT, Codex, PaLM, and Flan-PaLM). First, experiments on ICL with flipped labels show that overriding semantic priors is an emergent ability of model scale. While small language models ignore flipped labels presented in-context and thus rely primarily on semantic priors from pretraining, large models can override semantic priors when presente"},"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":"2303.03846","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-03-07T12:24:17Z","cross_cats_sorted":[],"title_canon_sha256":"5db8ddacdbfb4f7530cbbda3651790aaf2c58d5b2458dfa644e0f8a1fd6da6b0","abstract_canon_sha256":"1ecd1040690f5dd6cc51a3e114a7708931bacb11428e12e6e698cad932430853"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:49:18.953185Z","signature_b64":"aCJnx73ndBqMgr0pBCiGX4kt1Q8s/w+uc8+Ady9pLq4LUb1j4c3/wFPoNY+vzu8CGKZsHKWvMYPyCB42iLlUCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b056ea7cf3cf94815baa0bbe6c599684e5914fe794129121abcaa6a9a269633a","last_reissued_at":"2026-07-05T05:49:18.952636Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:49:18.952636Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Larger language models do in-context learning differently","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Albert Webson, Da Huang, Denny Zhou, Dustin Tran, Hanxiao Liu, Jason Wei, Jerry Wei, Tengyu Ma, Xinyun Chen, Yifeng Lu, Yi Tay","submitted_at":"2023-03-07T12:24:17Z","abstract_excerpt":"We study how in-context learning (ICL) in language models is affected by semantic priors versus input-label mappings. We investigate two setups-ICL with flipped labels and ICL with semantically-unrelated labels-across various model families (GPT-3, InstructGPT, Codex, PaLM, and Flan-PaLM). First, experiments on ICL with flipped labels show that overriding semantic priors is an emergent ability of model scale. While small language models ignore flipped labels presented in-context and thus rely primarily on semantic priors from pretraining, large models can override semantic priors when presente"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2303.03846","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/2303.03846/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":"2303.03846","created_at":"2026-07-05T05:49:18.952701+00:00"},{"alias_kind":"arxiv_version","alias_value":"2303.03846v2","created_at":"2026-07-05T05:49:18.952701+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2303.03846","created_at":"2026-07-05T05:49:18.952701+00:00"},{"alias_kind":"pith_short_12","alias_value":"WBLOU7HTZ6KI","created_at":"2026-07-05T05:49:18.952701+00:00"},{"alias_kind":"pith_short_16","alias_value":"WBLOU7HTZ6KICW5K","created_at":"2026-07-05T05:49:18.952701+00:00"},{"alias_kind":"pith_short_8","alias_value":"WBLOU7HT","created_at":"2026-07-05T05:49:18.952701+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":25,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.13001","citing_title":"CFALR: Collaborative Filtering-Augmented Large Language Model for Personalized Fashion Outfit Recommendation","ref_index":50,"is_internal_anchor":false},{"citing_arxiv_id":"2606.08068","citing_title":"DICE: Entropy-Regularized Equilibrium Selection for Stable Multi-Agent LLM Coordination","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04320","citing_title":"OpenRFM: Dissecting Relational In-Context Learning","ref_index":52,"is_internal_anchor":false},{"citing_arxiv_id":"2606.03089","citing_title":"Constitutional On-Policy Safe Distillation","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02953","citing_title":"Linguistic Productivity in Large Language Models: Models Coerce, but do not Preempt","ref_index":263,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26350","citing_title":"When Correct Demonstrations Hurt: Rethinking the Role of Exemplars in In-Context Learning","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07555","citing_title":"Priors Persist Through Suppression: A Stroop Paradigm for Lexical Override","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.29548","citing_title":"Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21974","citing_title":"Format-Constraint Coupling in Knowledge Graph Construction from Statistical Tables","ref_index":44,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23267","citing_title":"Fine-tuning vs. In-context Learning in Large Language Models: A Formal Language Learning Perspective","ref_index":69,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18830","citing_title":"In-Context Learning Operates as Concept Subspace Learning","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2506.07298","citing_title":"Pre-trained Large Language Models Learn Hidden Markov Models In-context","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2509.11295","citing_title":"The Prompt Engineering Report Distilled: Quick Start Guide for Life Sciences","ref_index":133,"is_internal_anchor":false},{"citing_arxiv_id":"2509.11295","citing_title":"The Prompt Engineering Report Distilled: Quick Start Guide for Life Sciences","ref_index":134,"is_internal_anchor":false},{"citing_arxiv_id":"2509.24164","citing_title":"Localizing Task Recognition and Task Learning in In-Context Learning via Attention Head Analysis","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2305.18565","citing_title":"PaLI-X: On Scaling up a Multilingual Vision and Language Model","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2310.14566","citing_title":"HallusionBench: An Advanced Diagnostic Suite for Entangled Language Hallucination and Visual Illusion in Large Vision-Language Models","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2309.03409","citing_title":"Large Language Models as Optimizers","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08904","citing_title":"OPT-BENCH: Evaluating the Iterative Self-Optimization of LLM Agents in Large-Scale Search Spaces","ref_index":88,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08295","citing_title":"In-Context Fixation: When Demonstrated Labels Override Semantics in Few-Shot Classification","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03780","citing_title":"Task Vector Geometry Underlies Dual Modes of Task Inference in Transformers","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2604.23267","citing_title":"Fine-tuning vs. In-context Learning in Large Language Models: A Formal Language Learning Perspective","ref_index":69,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13275","citing_title":"Better and Worse with Scale: How Contextual Entrainment Diverges with Model Size","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2403.07974","citing_title":"LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code","ref_index":199,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15789","citing_title":"A Systematic Study of Training-Free Methods for Trustworthy Large Language Models","ref_index":52,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WBLOU7HTZ6KICW5KBO7GYWMWQT","json":"https://pith.science/pith/WBLOU7HTZ6KICW5KBO7GYWMWQT.json","graph_json":"https://pith.science/api/pith-number/WBLOU7HTZ6KICW5KBO7GYWMWQT/graph.json","events_json":"https://pith.science/api/pith-number/WBLOU7HTZ6KICW5KBO7GYWMWQT/events.json","paper":"https://pith.science/paper/WBLOU7HT"},"agent_actions":{"view_html":"https://pith.science/pith/WBLOU7HTZ6KICW5KBO7GYWMWQT","download_json":"https://pith.science/pith/WBLOU7HTZ6KICW5KBO7GYWMWQT.json","view_paper":"https://pith.science/paper/WBLOU7HT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2303.03846&json=true","fetch_graph":"https://pith.science/api/pith-number/WBLOU7HTZ6KICW5KBO7GYWMWQT/graph.json","fetch_events":"https://pith.science/api/pith-number/WBLOU7HTZ6KICW5KBO7GYWMWQT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WBLOU7HTZ6KICW5KBO7GYWMWQT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WBLOU7HTZ6KICW5KBO7GYWMWQT/action/storage_attestation","attest_author":"https://pith.science/pith/WBLOU7HTZ6KICW5KBO7GYWMWQT/action/author_attestation","sign_citation":"https://pith.science/pith/WBLOU7HTZ6KICW5KBO7GYWMWQT/action/citation_signature","submit_replication":"https://pith.science/pith/WBLOU7HTZ6KICW5KBO7GYWMWQT/action/replication_record"}},"created_at":"2026-07-05T05:49:18.952701+00:00","updated_at":"2026-07-05T05:49:18.952701+00:00"}