{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BES2UD2M7KSCDFWLVBA2JUZWTH","short_pith_number":"pith:BES2UD2M","schema_version":"1.0","canonical_sha256":"0925aa0f4cfaa42196cba841a4d33699e7d2e20f1bf2a66909dc3d2989ab3fc0","source":{"kind":"arxiv","id":"2505.22630","version":2},"attestation_state":"computed","paper":{"title":"Stochastic Chameleons: Irrelevant Context Hallucinations Reveal Class-Based (Mis)Generalization in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jackie Chi Kit Cheung, Marc-Antoine Rondeau, Meng Cao, Ziling Cheng","submitted_at":"2025-05-28T17:47:52Z","abstract_excerpt":"The widespread success of large language models (LLMs) on NLP benchmarks has been accompanied by concerns that LLMs function primarily as stochastic parrots that reproduce texts similar to what they saw during pre-training, often erroneously. But what is the nature of their errors, and do these errors exhibit any regularities? In this work, we examine irrelevant context hallucinations, in which models integrate misleading contextual cues into their predictions. Through behavioral analysis, we show that these errors result from a structured yet flawed mechanism that we term class-based (mis)gen"},"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":"2505.22630","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-05-28T17:47:52Z","cross_cats_sorted":[],"title_canon_sha256":"417abb8e8b94346b9da16f820c36757f6a79e76a8e2a3f3935b948a4e579f72a","abstract_canon_sha256":"b0785670dc6d7d6b5906e5df22f8af0e8aa2300a3a55c856054e385fa3daa188"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:12:34.395125Z","signature_b64":"TNZRAGKM/uK60TFOL10MYategijm+ytfRIzSms2Q+qRG5R4lh1UjV4yms+zLzEkkuIgdPvo+HLvDuMxuarB/Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0925aa0f4cfaa42196cba841a4d33699e7d2e20f1bf2a66909dc3d2989ab3fc0","last_reissued_at":"2026-07-05T11:12:34.394620Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:12:34.394620Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Stochastic Chameleons: Irrelevant Context Hallucinations Reveal Class-Based (Mis)Generalization in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Jackie Chi Kit Cheung, Marc-Antoine Rondeau, Meng Cao, Ziling Cheng","submitted_at":"2025-05-28T17:47:52Z","abstract_excerpt":"The widespread success of large language models (LLMs) on NLP benchmarks has been accompanied by concerns that LLMs function primarily as stochastic parrots that reproduce texts similar to what they saw during pre-training, often erroneously. But what is the nature of their errors, and do these errors exhibit any regularities? In this work, we examine irrelevant context hallucinations, in which models integrate misleading contextual cues into their predictions. Through behavioral analysis, we show that these errors result from a structured yet flawed mechanism that we term class-based (mis)gen"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.22630","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/2505.22630/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":"2505.22630","created_at":"2026-07-05T11:12:34.394686+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.22630v2","created_at":"2026-07-05T11:12:34.394686+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.22630","created_at":"2026-07-05T11:12:34.394686+00:00"},{"alias_kind":"pith_short_12","alias_value":"BES2UD2M7KSC","created_at":"2026-07-05T11:12:34.394686+00:00"},{"alias_kind":"pith_short_16","alias_value":"BES2UD2M7KSCDFWL","created_at":"2026-07-05T11:12:34.394686+00:00"},{"alias_kind":"pith_short_8","alias_value":"BES2UD2M","created_at":"2026-07-05T11:12:34.394686+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.23701","citing_title":"Can LLMs Reason Abstractly Over Math Word Problems Without CoT? Disentangling Abstract Formulation From Arithmetic Computation","ref_index":4,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BES2UD2M7KSCDFWLVBA2JUZWTH","json":"https://pith.science/pith/BES2UD2M7KSCDFWLVBA2JUZWTH.json","graph_json":"https://pith.science/api/pith-number/BES2UD2M7KSCDFWLVBA2JUZWTH/graph.json","events_json":"https://pith.science/api/pith-number/BES2UD2M7KSCDFWLVBA2JUZWTH/events.json","paper":"https://pith.science/paper/BES2UD2M"},"agent_actions":{"view_html":"https://pith.science/pith/BES2UD2M7KSCDFWLVBA2JUZWTH","download_json":"https://pith.science/pith/BES2UD2M7KSCDFWLVBA2JUZWTH.json","view_paper":"https://pith.science/paper/BES2UD2M","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.22630&json=true","fetch_graph":"https://pith.science/api/pith-number/BES2UD2M7KSCDFWLVBA2JUZWTH/graph.json","fetch_events":"https://pith.science/api/pith-number/BES2UD2M7KSCDFWLVBA2JUZWTH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BES2UD2M7KSCDFWLVBA2JUZWTH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BES2UD2M7KSCDFWLVBA2JUZWTH/action/storage_attestation","attest_author":"https://pith.science/pith/BES2UD2M7KSCDFWLVBA2JUZWTH/action/author_attestation","sign_citation":"https://pith.science/pith/BES2UD2M7KSCDFWLVBA2JUZWTH/action/citation_signature","submit_replication":"https://pith.science/pith/BES2UD2M7KSCDFWLVBA2JUZWTH/action/replication_record"}},"created_at":"2026-07-05T11:12:34.394686+00:00","updated_at":"2026-07-05T11:12:34.394686+00:00"}