dorsal/arxiv
View SchemaAdaptive Multi-Stage Patent Claim Generation with Unified Quality Assessment
| Authors | Chen-Wei Liang, Bin Guo, Zhen-Yuan Wei, Mu-Jiang-Shan Wang |
|---|---|
| Categories | |
| ArXiv ID | 2601.09120vv1 |
| URL | https://arxiv.org/abs/2601.09120 |
| License | http://creativecommons.org/licenses/by/4.0/ |
Abstract
Current patent claim generation systems face three fundamental limitations: poor cross-jurisdictional generalization, inadequate semantic relationship modeling between claims and prior art, and unreliable quality assessment. We introduce a novel three-stage framework that addresses these challenges through relationship-aware similarity analysis, domain-adaptive claim generation, and unified quality assessment. Our approach employs multi-head attention with eight specialized heads for explicit relationship modeling, integrates curriculum learning with dynamic LoRA adapter selection across five patent domains, and implements cross-attention mechanisms between evaluation aspects for comprehensive quality assessment. Extensive experiments on USPTO HUPD dataset, EPO patent collections, and Patent-CE benchmark demonstrate substantial improvements: 7.6-point ROUGE-L gain over GPT-4o, 8.3\% BERTScore enhancement over Llama-3.1-8B, and 0.847 correlation with human experts compared to 0.623 for separate evaluation models. Our method maintains 89.4\% cross-jurisdictional performance retention versus 76.2\% for baselines, establishing a comprehensive solution for automated patent prosecution workflows.
{
"annotation_id": "5e1e6309-ac2b-4055-a521-17546b45818f",
"date_created": "2026-02-17T05:53:20.064000Z",
"date_modified": "2026-02-17T05:53:20.064000Z",
"file_hash": "3efb1f36a33fe364babf73426dd8ccc76930348346e399cffd62559969ee75e1",
"private": false,
"record": {
"abstract": "Current patent claim generation systems face three fundamental limitations: poor cross-jurisdictional generalization, inadequate semantic relationship modeling between claims and prior art, and unreliable quality assessment. We introduce a novel three-stage framework that addresses these challenges through relationship-aware similarity analysis, domain-adaptive claim generation, and unified quality assessment. Our approach employs multi-head attention with eight specialized heads for explicit relationship modeling, integrates curriculum learning with dynamic LoRA adapter selection across five patent domains, and implements cross-attention mechanisms between evaluation aspects for comprehensive quality assessment. Extensive experiments on USPTO HUPD dataset, EPO patent collections, and Patent-CE benchmark demonstrate substantial improvements: 7.6-point ROUGE-L gain over GPT-4o, 8.3\\% BERTScore enhancement over Llama-3.1-8B, and 0.847 correlation with human experts compared to 0.623 for separate evaluation models. Our method maintains 89.4\\% cross-jurisdictional performance retention versus 76.2\\% for baselines, establishing a comprehensive solution for automated patent prosecution workflows.",
"arxiv_id": "2601.09120",
"authors": [
"Chen-Wei Liang",
"Bin Guo",
"Zhen-Yuan Wei",
"Mu-Jiang-Shan Wang"
],
"categories": [
"cs.CL",
"cs.AI"
],
"license": "http://creativecommons.org/licenses/by/4.0/",
"title": "Adaptive Multi-Stage Patent Claim Generation with Unified Quality Assessment",
"url": "https://arxiv.org/abs/2601.09120",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "c013d66b-7f97-4666-bac2-2e6d85a67670",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}