{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:XMXG7MSMWAFKE6ID4CG6FMO7RL","short_pith_number":"pith:XMXG7MSM","schema_version":"1.0","canonical_sha256":"bb2e6fb24cb00aa27903e08de2b1df8af048e3575a93a4ae4ad84cdd942cfedd","source":{"kind":"arxiv","id":"2410.08996","version":1},"attestation_state":"computed","paper":{"title":"Hypothesis-only Biases in Large Language Model-Elicited Natural Language Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Adam Poliak, Grace Proebsting","submitted_at":"2024-10-11T17:09:22Z","abstract_excerpt":"We test whether replacing crowdsource workers with LLMs to write Natural Language Inference (NLI) hypotheses similarly results in annotation artifacts. We recreate a portion of the Stanford NLI corpus using GPT-4, Llama-2 and Mistral 7b, and train hypothesis-only classifiers to determine whether LLM-elicited hypotheses contain annotation artifacts. On our LLM-elicited NLI datasets, BERT-based hypothesis-only classifiers achieve between 86-96% accuracy, indicating these datasets contain hypothesis-only artifacts. We also find frequent \"give-aways\" in LLM-generated hypotheses, e.g. the phrase \"s"},"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":"2410.08996","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-10-11T17:09:22Z","cross_cats_sorted":[],"title_canon_sha256":"b40f50a44b098c7692e13b50843c3c8af3d5e345e580a6b47afea75deb8d1a07","abstract_canon_sha256":"bea006f78108582c0ecc0df58fe84bbe190b0f3068c27374cfe0019037f4dbd5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:19:20.453184Z","signature_b64":"FLpnOL1RKVzLeMFAI4OlvmKYcUYS8QnDGLMVvYB6Nag3bXjwZj+PdY/O9pFz31oIutfB518Swghkd15Gj4LxAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bb2e6fb24cb00aa27903e08de2b1df8af048e3575a93a4ae4ad84cdd942cfedd","last_reissued_at":"2026-07-05T09:19:20.452404Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:19:20.452404Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hypothesis-only Biases in Large Language Model-Elicited Natural Language Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Adam Poliak, Grace Proebsting","submitted_at":"2024-10-11T17:09:22Z","abstract_excerpt":"We test whether replacing crowdsource workers with LLMs to write Natural Language Inference (NLI) hypotheses similarly results in annotation artifacts. We recreate a portion of the Stanford NLI corpus using GPT-4, Llama-2 and Mistral 7b, and train hypothesis-only classifiers to determine whether LLM-elicited hypotheses contain annotation artifacts. On our LLM-elicited NLI datasets, BERT-based hypothesis-only classifiers achieve between 86-96% accuracy, indicating these datasets contain hypothesis-only artifacts. We also find frequent \"give-aways\" in LLM-generated hypotheses, e.g. the phrase \"s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.08996","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/2410.08996/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":"2410.08996","created_at":"2026-07-05T09:19:20.452464+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.08996v1","created_at":"2026-07-05T09:19:20.452464+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.08996","created_at":"2026-07-05T09:19:20.452464+00:00"},{"alias_kind":"pith_short_12","alias_value":"XMXG7MSMWAFK","created_at":"2026-07-05T09:19:20.452464+00:00"},{"alias_kind":"pith_short_16","alias_value":"XMXG7MSMWAFKE6ID","created_at":"2026-07-05T09:19:20.452464+00:00"},{"alias_kind":"pith_short_8","alias_value":"XMXG7MSM","created_at":"2026-07-05T09:19:20.452464+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2508.19922","citing_title":"HEAL: A Hypothesis-Based Preference-Aware Analysis Framework","ref_index":18,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XMXG7MSMWAFKE6ID4CG6FMO7RL","json":"https://pith.science/pith/XMXG7MSMWAFKE6ID4CG6FMO7RL.json","graph_json":"https://pith.science/api/pith-number/XMXG7MSMWAFKE6ID4CG6FMO7RL/graph.json","events_json":"https://pith.science/api/pith-number/XMXG7MSMWAFKE6ID4CG6FMO7RL/events.json","paper":"https://pith.science/paper/XMXG7MSM"},"agent_actions":{"view_html":"https://pith.science/pith/XMXG7MSMWAFKE6ID4CG6FMO7RL","download_json":"https://pith.science/pith/XMXG7MSMWAFKE6ID4CG6FMO7RL.json","view_paper":"https://pith.science/paper/XMXG7MSM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.08996&json=true","fetch_graph":"https://pith.science/api/pith-number/XMXG7MSMWAFKE6ID4CG6FMO7RL/graph.json","fetch_events":"https://pith.science/api/pith-number/XMXG7MSMWAFKE6ID4CG6FMO7RL/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XMXG7MSMWAFKE6ID4CG6FMO7RL/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XMXG7MSMWAFKE6ID4CG6FMO7RL/action/storage_attestation","attest_author":"https://pith.science/pith/XMXG7MSMWAFKE6ID4CG6FMO7RL/action/author_attestation","sign_citation":"https://pith.science/pith/XMXG7MSMWAFKE6ID4CG6FMO7RL/action/citation_signature","submit_replication":"https://pith.science/pith/XMXG7MSMWAFKE6ID4CG6FMO7RL/action/replication_record"}},"created_at":"2026-07-05T09:19:20.452464+00:00","updated_at":"2026-07-05T09:19:20.452464+00:00"}