{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:KFEPYIT6W4RQGHQAJIF6FTPFF4","short_pith_number":"pith:KFEPYIT6","schema_version":"1.0","canonical_sha256":"5148fc227eb723031e004a0be2cde52f11e14fe6dd67f1c3a60f5ab3876b0f58","source":{"kind":"arxiv","id":"2412.00868","version":1},"attestation_state":"computed","paper":{"title":"Quantifying perturbation impacts for large language models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","math.ST","stat.ML","stat.TH"],"primary_cat":"cs.LG","authors_text":"Mihaela van der Schaar, Paulius Rauba, Qiyao Wei","submitted_at":"2024-12-01T16:13:09Z","abstract_excerpt":"We consider the problem of quantifying how an input perturbation impacts the outputs of large language models (LLMs), a fundamental task for model reliability and post-hoc interpretability. A key obstacle in this domain is disentangling the meaningful changes in model responses from the intrinsic stochasticity of LLM outputs. To overcome this, we introduce Distribution-Based Perturbation Analysis (DBPA), a framework that reformulates LLM perturbation analysis as a frequentist hypothesis testing problem. DBPA constructs empirical null and alternative output distributions within a low-dimensiona"},"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":"2412.00868","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-12-01T16:13:09Z","cross_cats_sorted":["cs.CL","math.ST","stat.ML","stat.TH"],"title_canon_sha256":"deb5105307d7648943162687b144fd9244421d833198e26260c82dcd52415d44","abstract_canon_sha256":"78d98864071bfbde72910b725800f47b6c909f00d004a27a87cb97273d33eec0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:42:49.591862Z","signature_b64":"DSSDIWlTV8D/hcq+Ff6U8nwm6pbsbU23xrTRYI/Ko5FReA09WbW2x8kKL3pFG6RGCWU+ypH2gmKq/5ouZgdgCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5148fc227eb723031e004a0be2cde52f11e14fe6dd67f1c3a60f5ab3876b0f58","last_reissued_at":"2026-07-05T09:42:49.591463Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:42:49.591463Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Quantifying perturbation impacts for large language models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","math.ST","stat.ML","stat.TH"],"primary_cat":"cs.LG","authors_text":"Mihaela van der Schaar, Paulius Rauba, Qiyao Wei","submitted_at":"2024-12-01T16:13:09Z","abstract_excerpt":"We consider the problem of quantifying how an input perturbation impacts the outputs of large language models (LLMs), a fundamental task for model reliability and post-hoc interpretability. A key obstacle in this domain is disentangling the meaningful changes in model responses from the intrinsic stochasticity of LLM outputs. To overcome this, we introduce Distribution-Based Perturbation Analysis (DBPA), a framework that reformulates LLM perturbation analysis as a frequentist hypothesis testing problem. DBPA constructs empirical null and alternative output distributions within a low-dimensiona"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.00868","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/2412.00868/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":"2412.00868","created_at":"2026-07-05T09:42:49.591515+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.00868v1","created_at":"2026-07-05T09:42:49.591515+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.00868","created_at":"2026-07-05T09:42:49.591515+00:00"},{"alias_kind":"pith_short_12","alias_value":"KFEPYIT6W4RQ","created_at":"2026-07-05T09:42:49.591515+00:00"},{"alias_kind":"pith_short_16","alias_value":"KFEPYIT6W4RQGHQA","created_at":"2026-07-05T09:42:49.591515+00:00"},{"alias_kind":"pith_short_8","alias_value":"KFEPYIT6","created_at":"2026-07-05T09:42:49.591515+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.09606","citing_title":"Evaluating Reliability Gaps in Large Language Model Safety via Repeated Prompt Sampling","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10516","citing_title":"Consistency as a Testable Property: Statistical Methods to Evaluate AI Agent Reliability","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01048","citing_title":"Compared to What? Baselines and Metrics for Counterfactual Prompting","ref_index":20,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KFEPYIT6W4RQGHQAJIF6FTPFF4","json":"https://pith.science/pith/KFEPYIT6W4RQGHQAJIF6FTPFF4.json","graph_json":"https://pith.science/api/pith-number/KFEPYIT6W4RQGHQAJIF6FTPFF4/graph.json","events_json":"https://pith.science/api/pith-number/KFEPYIT6W4RQGHQAJIF6FTPFF4/events.json","paper":"https://pith.science/paper/KFEPYIT6"},"agent_actions":{"view_html":"https://pith.science/pith/KFEPYIT6W4RQGHQAJIF6FTPFF4","download_json":"https://pith.science/pith/KFEPYIT6W4RQGHQAJIF6FTPFF4.json","view_paper":"https://pith.science/paper/KFEPYIT6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.00868&json=true","fetch_graph":"https://pith.science/api/pith-number/KFEPYIT6W4RQGHQAJIF6FTPFF4/graph.json","fetch_events":"https://pith.science/api/pith-number/KFEPYIT6W4RQGHQAJIF6FTPFF4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KFEPYIT6W4RQGHQAJIF6FTPFF4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KFEPYIT6W4RQGHQAJIF6FTPFF4/action/storage_attestation","attest_author":"https://pith.science/pith/KFEPYIT6W4RQGHQAJIF6FTPFF4/action/author_attestation","sign_citation":"https://pith.science/pith/KFEPYIT6W4RQGHQAJIF6FTPFF4/action/citation_signature","submit_replication":"https://pith.science/pith/KFEPYIT6W4RQGHQAJIF6FTPFF4/action/replication_record"}},"created_at":"2026-07-05T09:42:49.591515+00:00","updated_at":"2026-07-05T09:42:49.591515+00:00"}