{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:INMNL2LORS3WDEIA3424QIBX5H","short_pith_number":"pith:INMNL2LO","schema_version":"1.0","canonical_sha256":"4358d5e96e8cb7619100df35c82037e9d06b1875b26885d79f9dcc9273895236","source":{"kind":"arxiv","id":"2504.02733","version":1},"attestation_state":"computed","paper":{"title":"Enhancing LLM Robustness to Perturbed Instructions: An Empirical Study","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aryan Agrawal, Lisa Alazraki, Marek Rei, Shahin Honarvar","submitted_at":"2025-04-03T16:17:56Z","abstract_excerpt":"Large Language Models (LLMs) are highly vulnerable to input perturbations, as even a small prompt change may result in a substantially different output. Existing methods to enhance LLM robustness are primarily focused on perturbed data samples, whereas improving resiliency to perturbations of task-level instructions has remained relatively underexplored. In this work, we focus on character- and word-level edits of task-specific instructions, which substantially degrade downstream performance. We experiment with a variety of techniques to enhance the robustness of LLMs, including self-denoising"},"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":"2504.02733","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-03T16:17:56Z","cross_cats_sorted":[],"title_canon_sha256":"f6dd37e7591d3cf10411e4df716bf969594d8f7599acd85d91e842b3a1edac0c","abstract_canon_sha256":"2d418f6967f700db92b15db1f0377977d9fa23b6a29e0256155699b24269269b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:44:01.447952Z","signature_b64":"5yyUyqhYpWSLRjAqZGkX/+Z4DJd1OaZDGMiqfULBJ+tszvTVNOzL1lNKSvOzwj3AvH6cgtYswJSGo/096A3UAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4358d5e96e8cb7619100df35c82037e9d06b1875b26885d79f9dcc9273895236","last_reissued_at":"2026-07-05T10:44:01.447512Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:44:01.447512Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Enhancing LLM Robustness to Perturbed Instructions: An Empirical Study","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aryan Agrawal, Lisa Alazraki, Marek Rei, Shahin Honarvar","submitted_at":"2025-04-03T16:17:56Z","abstract_excerpt":"Large Language Models (LLMs) are highly vulnerable to input perturbations, as even a small prompt change may result in a substantially different output. Existing methods to enhance LLM robustness are primarily focused on perturbed data samples, whereas improving resiliency to perturbations of task-level instructions has remained relatively underexplored. In this work, we focus on character- and word-level edits of task-specific instructions, which substantially degrade downstream performance. We experiment with a variety of techniques to enhance the robustness of LLMs, including self-denoising"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.02733","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/2504.02733/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":"2504.02733","created_at":"2026-07-05T10:44:01.447569+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.02733v1","created_at":"2026-07-05T10:44:01.447569+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.02733","created_at":"2026-07-05T10:44:01.447569+00:00"},{"alias_kind":"pith_short_12","alias_value":"INMNL2LORS3W","created_at":"2026-07-05T10:44:01.447569+00:00"},{"alias_kind":"pith_short_16","alias_value":"INMNL2LORS3WDEIA","created_at":"2026-07-05T10:44:01.447569+00:00"},{"alias_kind":"pith_short_8","alias_value":"INMNL2LO","created_at":"2026-07-05T10:44:01.447569+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.17634","citing_title":"Prompt Perturbation for Reliable LLM Evaluation over Comparison Graphs","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.27209","citing_title":"Learning to Act under Noise: Enhancing Agent Robustness via Noisy Environments","ref_index":63,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07186","citing_title":"The Text Uncanny Valley: Non-Monotonic Performance Degradation in LLM Information Retrieval","ref_index":2,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/INMNL2LORS3WDEIA3424QIBX5H","json":"https://pith.science/pith/INMNL2LORS3WDEIA3424QIBX5H.json","graph_json":"https://pith.science/api/pith-number/INMNL2LORS3WDEIA3424QIBX5H/graph.json","events_json":"https://pith.science/api/pith-number/INMNL2LORS3WDEIA3424QIBX5H/events.json","paper":"https://pith.science/paper/INMNL2LO"},"agent_actions":{"view_html":"https://pith.science/pith/INMNL2LORS3WDEIA3424QIBX5H","download_json":"https://pith.science/pith/INMNL2LORS3WDEIA3424QIBX5H.json","view_paper":"https://pith.science/paper/INMNL2LO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.02733&json=true","fetch_graph":"https://pith.science/api/pith-number/INMNL2LORS3WDEIA3424QIBX5H/graph.json","fetch_events":"https://pith.science/api/pith-number/INMNL2LORS3WDEIA3424QIBX5H/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/INMNL2LORS3WDEIA3424QIBX5H/action/timestamp_anchor","attest_storage":"https://pith.science/pith/INMNL2LORS3WDEIA3424QIBX5H/action/storage_attestation","attest_author":"https://pith.science/pith/INMNL2LORS3WDEIA3424QIBX5H/action/author_attestation","sign_citation":"https://pith.science/pith/INMNL2LORS3WDEIA3424QIBX5H/action/citation_signature","submit_replication":"https://pith.science/pith/INMNL2LORS3WDEIA3424QIBX5H/action/replication_record"}},"created_at":"2026-07-05T10:44:01.447569+00:00","updated_at":"2026-07-05T10:44:01.447569+00:00"}