{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:ZDIRMS4BXZBCCA6EJQLKOEWWBE","short_pith_number":"pith:ZDIRMS4B","schema_version":"1.0","canonical_sha256":"c8d1164b81be422103c44c16a712d60933aa9c9345fd37dbcdffc5f1a6208c88","source":{"kind":"arxiv","id":"2604.04295","version":3},"attestation_state":"computed","paper":{"title":"Adaptive Cost-Efficient Evaluation for Reliable Patent Claim Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Hybrid system routes uncertain patent claims to LLMs via entropy, hitting 94.95% F1 at 78% lower cost.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Longbing Cao, Qiongkai Xu, Yongmin Yoo","submitted_at":"2026-04-05T22:25:36Z","abstract_excerpt":"Automated patent claim validation demands low error tolerance. However, existing approaches face a rigidity-resource dilemma: lightweight encoders cannot track long-range legal dependencies, while exhaustive LLM verification incurs 4-5X higher overhead at million-claim scale. A naive confidence-based cascade cannot resolve this because binary validity scores fail to distinguish structurally distinct error types which require different reasoning depths. We propose a two-stage framework: Adaptive Cost-efficient Evaluation (ACE), which exploits the categorical structure of patent errors for uncer"},"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":"2604.04295","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-04-05T22:25:36Z","cross_cats_sorted":[],"title_canon_sha256":"7d9d2c36963318dad2f7e0f3b4e804b290261f3a7be7769caa59273a44c889ff","abstract_canon_sha256":"4e4dbcef878f88a1ca64609fbaa02b852f2349152dfd13fb31c431079b68920a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:39.456546Z","signature_b64":"2hCrbI7v8IlesXtF+cOxnnycOD403x9T4816j1effQKq4af2VhJIPQU4TAKOeILJwMnYUCtsiM9M5k72XQVHBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c8d1164b81be422103c44c16a712d60933aa9c9345fd37dbcdffc5f1a6208c88","last_reissued_at":"2026-05-28T01:04:39.455898Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:39.455898Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Adaptive Cost-Efficient Evaluation for Reliable Patent Claim Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"Hybrid system routes uncertain patent claims to LLMs via entropy, hitting 94.95% F1 at 78% lower cost.","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Longbing Cao, Qiongkai Xu, Yongmin Yoo","submitted_at":"2026-04-05T22:25:36Z","abstract_excerpt":"Automated patent claim validation demands low error tolerance. However, existing approaches face a rigidity-resource dilemma: lightweight encoders cannot track long-range legal dependencies, while exhaustive LLM verification incurs 4-5X higher overhead at million-claim scale. A naive confidence-based cascade cannot resolve this because binary validity scores fail to distinguish structurally distinct error types which require different reasoning depths. We propose a two-stage framework: Adaptive Cost-efficient Evaluation (ACE), which exploits the categorical structure of patent errors for uncer"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On our constructed benchmark, ACE achieves the best F1 among the evaluated methods at 94.95% while reducing operational costs by 78% compared to standalone LLM deployments. Crucially, the entropy-based routing threshold transfers directly to authentic USPTO §112(b) rejections without re-calibration.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"Predictive entropy from the lightweight encoder reliably flags claims that require LLM-level review for long-range legal dependencies, and the CoPT protocol enables the LLM to resolve those dependencies effectively without additional fine-tuning or calibration.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"ACE achieves 94.95% F1 on patent claim validation by routing high-entropy cases to an LLM with CoPT reasoning, cutting costs 78% versus full LLM use, with the threshold transferring to real USPTO rejections.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Hybrid system routes uncertain patent claims to LLMs via entropy, hitting 94.95% F1 at 78% lower cost.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"e272aac39bef5cc717e87d63b0c8ee736cb6a5e184424b2386507f061eaed3e4"},"source":{"id":"2604.04295","kind":"arxiv","version":3},"verdict":{"id":"537cf0b7-8c6e-4212-b76e-67819fbf28f0","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T21:59:02.580771Z","strongest_claim":"On our constructed benchmark, ACE achieves the best F1 among the evaluated methods at 94.95% while reducing operational costs by 78% compared to standalone LLM deployments. Crucially, the entropy-based routing threshold transfers directly to authentic USPTO §112(b) rejections without re-calibration.","one_line_summary":"ACE achieves 94.95% F1 on patent claim validation by routing high-entropy cases to an LLM with CoPT reasoning, cutting costs 78% versus full LLM use, with the threshold transferring to real USPTO rejections.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"Predictive entropy from the lightweight encoder reliably flags claims that require LLM-level review for long-range legal dependencies, and the CoPT protocol enables the LLM to resolve those dependencies effectively without additional fine-tuning or calibration.","pith_extraction_headline":"Hybrid system routes uncertain patent claims to LLMs via entropy, hitting 94.95% F1 at 78% lower cost."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.04295/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":"2604.04295","created_at":"2026-05-28T01:04:39.455975+00:00"},{"alias_kind":"arxiv_version","alias_value":"2604.04295v3","created_at":"2026-05-28T01:04:39.455975+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.04295","created_at":"2026-05-28T01:04:39.455975+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZDIRMS4BXZBC","created_at":"2026-05-28T01:04:39.455975+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZDIRMS4BXZBCCA6E","created_at":"2026-05-28T01:04:39.455975+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZDIRMS4B","created_at":"2026-05-28T01:04:39.455975+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/ZDIRMS4BXZBCCA6EJQLKOEWWBE","json":"https://pith.science/pith/ZDIRMS4BXZBCCA6EJQLKOEWWBE.json","graph_json":"https://pith.science/api/pith-number/ZDIRMS4BXZBCCA6EJQLKOEWWBE/graph.json","events_json":"https://pith.science/api/pith-number/ZDIRMS4BXZBCCA6EJQLKOEWWBE/events.json","paper":"https://pith.science/paper/ZDIRMS4B"},"agent_actions":{"view_html":"https://pith.science/pith/ZDIRMS4BXZBCCA6EJQLKOEWWBE","download_json":"https://pith.science/pith/ZDIRMS4BXZBCCA6EJQLKOEWWBE.json","view_paper":"https://pith.science/paper/ZDIRMS4B","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2604.04295&json=true","fetch_graph":"https://pith.science/api/pith-number/ZDIRMS4BXZBCCA6EJQLKOEWWBE/graph.json","fetch_events":"https://pith.science/api/pith-number/ZDIRMS4BXZBCCA6EJQLKOEWWBE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZDIRMS4BXZBCCA6EJQLKOEWWBE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZDIRMS4BXZBCCA6EJQLKOEWWBE/action/storage_attestation","attest_author":"https://pith.science/pith/ZDIRMS4BXZBCCA6EJQLKOEWWBE/action/author_attestation","sign_citation":"https://pith.science/pith/ZDIRMS4BXZBCCA6EJQLKOEWWBE/action/citation_signature","submit_replication":"https://pith.science/pith/ZDIRMS4BXZBCCA6EJQLKOEWWBE/action/replication_record"}},"created_at":"2026-05-28T01:04:39.455975+00:00","updated_at":"2026-05-28T01:04:39.455975+00:00"}