{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:DETCQ2VL4XUORD7WOVMQE4IAAB","short_pith_number":"pith:DETCQ2VL","schema_version":"1.0","canonical_sha256":"1926286aabe5e8e88ff6755902710000521e4f7e475a8985cb809efdb5b6b5b2","source":{"kind":"arxiv","id":"2605.21318","version":1},"attestation_state":"computed","paper":{"title":"TextReg: Mitigating Prompt Distributional Overfitting via Regularized Text-Space Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"B. Aditya Prakash, Haibo Jin, Haohan Wang, Lucheng Fu, Ye Yu, Yiqiao Jin, Yiyang Wang","submitted_at":"2026-05-20T15:47:26Z","abstract_excerpt":"Large language models (LLMs) are highly sensitive to the prompts used to specify task objectives and behavioral constraints. Many recent prompt optimization methods iteratively rewrite prompts using LLM-generated feedback, but the resulting prompts often become longer, accumulate narrow sample-specific rules, and generalize poorly beyond the training distribution. We study this failure mode as prompt distributional overfitting and argue that it reflects a lack of representation control in discrete text-space optimization. We formalize this view through representational inefficiency, a dual-fac"},"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":"2605.21318","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-20T15:47:26Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"be4860dcd92a8f1ad63e41791bd7bc8f7f8f33f0ec868ad216be646f1c653677","abstract_canon_sha256":"c374bba1091653ea07d3a423ea9c8e61d4458ebcd6cbbfb7ad9b1392e0195844"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-21T02:05:28.967440Z","signature_b64":"A+nDNTd0bqSHaykJeU745dg/Bt/gzOBNL0To/aF6hqpDt15F8ajWqd2l6pAe5OFkIv1uNwDje47EwGU59c2zBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1926286aabe5e8e88ff6755902710000521e4f7e475a8985cb809efdb5b6b5b2","last_reissued_at":"2026-05-21T02:05:28.966606Z","signature_status":"signed_v1","first_computed_at":"2026-05-21T02:05:28.966606Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TextReg: Mitigating Prompt Distributional Overfitting via Regularized Text-Space Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"B. Aditya Prakash, Haibo Jin, Haohan Wang, Lucheng Fu, Ye Yu, Yiqiao Jin, Yiyang Wang","submitted_at":"2026-05-20T15:47:26Z","abstract_excerpt":"Large language models (LLMs) are highly sensitive to the prompts used to specify task objectives and behavioral constraints. Many recent prompt optimization methods iteratively rewrite prompts using LLM-generated feedback, but the resulting prompts often become longer, accumulate narrow sample-specific rules, and generalize poorly beyond the training distribution. We study this failure mode as prompt distributional overfitting and argue that it reflects a lack of representation control in discrete text-space optimization. We formalize this view through representational inefficiency, a dual-fac"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.21318","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/2605.21318/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":"2605.21318","created_at":"2026-05-21T02:05:28.966750+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.21318v1","created_at":"2026-05-21T02:05:28.966750+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.21318","created_at":"2026-05-21T02:05:28.966750+00:00"},{"alias_kind":"pith_short_12","alias_value":"DETCQ2VL4XUO","created_at":"2026-05-21T02:05:28.966750+00:00"},{"alias_kind":"pith_short_16","alias_value":"DETCQ2VL4XUORD7W","created_at":"2026-05-21T02:05:28.966750+00:00"},{"alias_kind":"pith_short_8","alias_value":"DETCQ2VL","created_at":"2026-05-21T02:05:28.966750+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DETCQ2VL4XUORD7WOVMQE4IAAB","json":"https://pith.science/pith/DETCQ2VL4XUORD7WOVMQE4IAAB.json","graph_json":"https://pith.science/api/pith-number/DETCQ2VL4XUORD7WOVMQE4IAAB/graph.json","events_json":"https://pith.science/api/pith-number/DETCQ2VL4XUORD7WOVMQE4IAAB/events.json","paper":"https://pith.science/paper/DETCQ2VL"},"agent_actions":{"view_html":"https://pith.science/pith/DETCQ2VL4XUORD7WOVMQE4IAAB","download_json":"https://pith.science/pith/DETCQ2VL4XUORD7WOVMQE4IAAB.json","view_paper":"https://pith.science/paper/DETCQ2VL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.21318&json=true","fetch_graph":"https://pith.science/api/pith-number/DETCQ2VL4XUORD7WOVMQE4IAAB/graph.json","fetch_events":"https://pith.science/api/pith-number/DETCQ2VL4XUORD7WOVMQE4IAAB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DETCQ2VL4XUORD7WOVMQE4IAAB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DETCQ2VL4XUORD7WOVMQE4IAAB/action/storage_attestation","attest_author":"https://pith.science/pith/DETCQ2VL4XUORD7WOVMQE4IAAB/action/author_attestation","sign_citation":"https://pith.science/pith/DETCQ2VL4XUORD7WOVMQE4IAAB/action/citation_signature","submit_replication":"https://pith.science/pith/DETCQ2VL4XUORD7WOVMQE4IAAB/action/replication_record"}},"created_at":"2026-05-21T02:05:28.966750+00:00","updated_at":"2026-05-21T02:05:28.966750+00:00"}