Securing patent protection across countries requires drafting claims separately for each national patent office, in the relevant language and according to its legal and stylistic conventions. Patent claims can therefore differ across languages and jurisdictions because of translation, target-office drafting practices, and prosecution-stage changes. We formulate cross-lingual patent drafting as generating plausible candidate claim sets for a specified target language, jurisdiction, and prosecution stage from a source claim set in another language. The task models jurisdiction- and stage-specific transformations beyond content-preserving translation. We construct a large-scale Japanese-English dataset by linking Japan Patent Office (JPO) and United States Patent and Trademark Office (USPTO) records through 884,179 Patent Cooperation Treaty (PCT)-derived patent families. Across six released configurations, these families yield 4,067,168 directional claim-set instances; four core configurations form the main benchmark. Commercial machine translation and zero- and few-shot prompting capture part of the task, while supervised adaptation provides the strongest overall baseline, showing that task-specific patterns can be learned from linked cross-jurisdictional claim sets.
Figures & tables
Figure 1: Cross-lingual patent drafting and its instantiation in our dataset, illustrated by the ja_pre → en_granted direction. Each task instance links source and target claim sets within the same PCT-derived patent family.
Dataset
Languages
Count
Unit
Task
Stages
Jurisdictions
Granularity
JaParaPat Nagata et al. (2024)
JA, EN
∼ 350M
sentence pairs
translation
pre
JPO, USPTO
sentence
NTCIR collections Fujii et al. (2007)
JA, EN
—
documents
retrieval, classif.
mixed
JPO, USPTO
document
HUPD Suzgun et al. (2023)
EN
4.52M
applications
multi-purpose
pre
USPTO
document
ClaimBrush Kawano et al. (2024)
JA
2.25M
rewriting pairs
refinement
CS
JPO
claim set
PatentEdits Lee et al. (2024)
EN
105K
draft–grant pairs
edit prediction
CS
USPTO
claim
Patent-CR Jiang et al. (2025b)
EN
22.6K
revision pairs
revision
CS
EPO
claim set
Table 1: Comparison with existing patent datasets. Counts use each dataset’s native unit and are therefore not directly comparable across rows. For ours, 884,179 families yield 4,067,168 directional linked claim-set instances across six released configurations. Pre = pre-grant; grant = granted; CS = cross-stage (pre to grant); a claim set is all claims in one publication. Bold identifies our dataset.
Figure 2: Publication stages and task structure. Same-stage tasks link documents at matched stages across jurisdictions; cross-stage tasks link a pre-grant source to the other jurisdiction’s granted claims.
Task direction
Train
Val.
Test
Instances
Families
ja_pre → en_pre
886,840
134,693
235,715
1,257,248
834,964
en_pre → ja_pre
886,840
134,693
235,715
1,257,248
834,964
ja_pre → en_granted
538,392
95,492
183,467
817,351
560,858
en_pre → ja_granted
529,975
51,829
55,043
636,847
475,649
ja_granted → en_granted
48,642
107
488
49,237
30,581
en_granted → ja_granted
48,642
107
488
49,237
30,581
Table 2: Released task directions and final dataset statistics. The first four directions form the core benchmark; the two granted-to-granted directions below the rule are supplementary, with only ja_granted → en_granted evaluated in Appendix A . Instance counts are directional, and family counts overlap across tasks; the total family count is global and therefore not the sum of rows. Cross-stage tasks retain only instances for which the source was publicly visible before the target grant.
Task direction
Instances
Src vocab.
Tgt vocab.
Src claims
Tgt claims
Mean T/S
Src chars
Tgt chars
ja_pre → en_pre
1.26M
1.15M
2.30M
25.82
25.77
1.20
2490
5321
en_pre → ja_pre
1.26M
2.30M
1.15M
25.77
25.82
1.41
5321
2490
ja_pre → en_granted
817.4K
886.9K
1.52M
25.80
17.01
0.93
2546
5248
en_pre → ja_granted
636.8K
1.47M
665.9K
24.79
15.39
0.88
5416
1985
ja_granted → en_granted
49.2K
108.3K
211.7K
15.62
18.97
1.97
1713
4560
en_granted → ja_granted
49.2K
211.7K
108.3K
18.97
15.62
1.35
4560
1713
Table 3: Extended statistics for the six released task directions. Vocabulary size uses MeCab (IPA dictionary) for Japanese and whitespace-after-punctuation splitting for English. Claims are means per claim set; Mean T/S is the mean per-instance target/source claim-count ratio. Character columns are median NFC-normalized Unicode code-point counts.
JPO (English translation)
USPTO (original English)
Pre-grant
JP-2004501492-A (2004-01-15), Claim 1 A fuel cell comprising at least one cell having an electrolyte layer (2), a pair of gas-diffusion electrode layers (3, 4) sandwiching the electrolyte layer, and a pair of flow-distribution plates (5) defining passages (10, 11) for fuel gas and oxidant gas that contact the respective gas-diffusion electrode layers, wherein the electrolyte layer (2) has a frame (21) and an electrolyte (22) held within the frame, and the two flow-distribution plates and the frame are joined or bonded around their entire outer peripheries so as to make the interior airtight.
US-2004091765-A1 (2004-05-13), Claim 1 A fuel cell assembly comprising a plurality of cells each including an electrolyte layer, a pair of diffusion electrode layers interposing said electrolyte layer between them, and a pair of flow distribution plates for defining passages for fuel and oxidant fluids that contact said diffusion electrode layers, wherein: said fuel cells are arranged on a common plane.
Granted
JP-4989802-B2 (2012-08-01), Claim 1 A fuel cell comprising at least one cell having an electrolyte layer (2), a pair of gas-diffusion electrode layers (3, 4) sandwiching the electrolyte layer, and a pair of flow-distribution plates (5) defining passages (10, 11) for fuel gas and oxidant gas that contact the respective gas-diffusion electrode layers, wherein the electrolyte layer (2) has a frame (21) and an electrolyte (22) held within the frame, and the two flow-distribution plates and the frame are, around their entire outer peripheries, anodically bonded by an electrode film formed on the outer peripheral portions of the two flow-distribution plates and a glass film formed on the outer peripheral portion of the frame, with a heater for anodic bonding embedded beneath the electrode film , so as to make the interior airtight.
US-6991868-B2 (2006-01-31), Claim 1 A fuel cell assembly comprising a plurality of cells each including an electrolyte layer, a pair of diffusion electrode layers interposing said electrolyte layer between them, and a pair of flow distribution plates for defining passages for fuel and oxidant fluids that contact said diffusion electrode layers, wherein: said fuel cells are arranged on a common plane, wherein each of said flow distribution plates comprises a plurality of recesses formed on a side thereof facing said electrolyte layer so as to form individual cells, said recesses being communicated to one another according to a prescribed pattern by said communication passages.
Table 4: English-language claim comparison for one invention across jurisdictions and prosecution stages. The JPO claims are translated into English for presentation; the original Japanese appears in Appendix E . All four documents derive from PCT publication WO2001/095404 and were prosecuted independently at the JPO and USPTO. Bold text highlights limitations added during examination.
System
JA-P → EN-P
EN-P → JA-P
JA-P → EN-G
EN-P → JA-G
BLEU ↑
chrF ↑
SARI ↑
BLEU ↑
chrF ↑
SARI ↑
BLEU ↑
chrF ↑
SARI ↑
BLEU ↑
chrF ↑
SARI ↑
Google Translate
24.81
58.97
48.00
18.60
27.35
41.48
21.72
54.77
46.81
24.39
32.48
39.48
Qwen zero-shot
31.08
61.71
49.30
18.30
27.29
59.66
28.01
57.61
48.65
26.35
32.45
57.48
Qwen few-shot
31.69
61.84
49.49
17.87
27.13
61.19
27.51
57.37
48.53
29.58
34.68
58.82
Qwen simple-MT
30.87
61.55
49.01
18.43
27.49
59.86
26.94
57.19
47.72
26.35
32.93
56.98
GPT-5.4 zero-shot
31.24
62.42
50.72
17.58
26.93
58.03
28.18
58.57
50.40
24.74
30.88
58.88
Table 5: Unified benchmark results. The upper panel reports overlap and edit metrics for the four core tasks; the lower panel reports mean absolute claim-count differences from the reference ( ∣Δr∣ , lower is better) and source ( ∣Δs∣ ) on the cross-stage tasks. P and G denote pre-grant and granted stages; dashes indicate pipelines not evaluated on same-stage tasks. The Reference row records the references’ inherent ∣Δs∣ . Bold marks the best system value.
Comparison
Ev.
QLoRA
GPT
MT
Tie
Agr.
QLoRA–MT
H
20 (80%)
–
5 (20%)
0
96%
L
19 (76%)
–
6 (24%)
0
QLoRA–GPT
H
19 (76%)
4 (16%)
–
2 (8%)
64%
L
13 (52%)
8 (32%)
–
4 (16%)
GPT–MT
H
–
19 (76%)
6 (24%)
0
88%
L
–
16 (64%)
8 (32%)
1 (4%)
Table 6: Blind pairwise preferences. H and L denote the human and LLM evaluators; GPT denotes GPT-5.4 zero-shot and MT denotes Qwen simple-MT. Percentages use the 25 cases in each comparison, and Agr. is per-comparison human–LLM agreement over the three labels (first candidate, second candidate, tie).
Appendix figures & tables3 assets
Supplementary material from the paper’s appendix.
Appendix
JA-P → EN-P
JA-G → EN-G
System
BLEU ↑
chrF ↑
SARI ↑
BLEU ↑
chrF ↑
SARI ↑
Google Translate
24.81
58.97
48.00
30.52
56.16
51.44
Qwen zero-shot
31.08
61.71
49.30
35.59
58.63
54.93
Qwen few-shot
31.69
61.84
49.49
35.48
58.54
55.00
Qwen simple-MT
30.87
61.55
49.01
35.40
58.57
54.25
GPT-5.4 zero-shot
31.24
62.42
50.72
36.04
59.09
56.23
Appendix
Table 7: Supplementary comparison of Japanese-to-English pre-grant and granted-stage transfer. P and G denote pre-grant and granted stages; SARI scoring follows Section 5 . Bold marks the best system within each task and metric.
Table 9: Original Japanese JPO claims for the translations in Table 4 ; bold marks the limitation added in the granted claim.
(a) Successful consolidation: EN-P → JA-G
Source 20 claims Claim 8
The producing method according to claim 1 , wherein the protective film shows a main peak of a ratio (tan δ ) between a storage modulus and a loss modulus in dynamic viscoelasticity measurement at a temperature of 80 ∘ C or higher.
Table 10: Qualitative comparison of Google Translate and QLoRA against the reference. Panel (a) shows reference-like consolidation; panel (b), source-supported narrowing by incorporating source Claim 2’s ranges into independent Claim 1; and panel (c), over-consolidation in which QLoRA reduces four method branches to one and omits the glycerol-export and comorbidity branches. Examples are illustrative, not assessments of legal validity; claim numbering differs because outputs are complete claim sets. P and G denote pre-grant and granted stages.
Drafting patent claims is time-intensive, costly, and requires professional skill. Therefore, researchers have investigated large language models (LLMs) to assist inventors in writing claims. However, existing work has largely relied on datasets from the United States Patent and Trademark Office (USPTO). To enlarge research scope regarding various jurisdictions, drafting conventions, and legal standards, we introduce EPD, a European patent dataset. EPD presents rich textual data and structured metadata to support multiple patent-related tasks, including claim generation. This dataset enriches the field in three critical aspects: (1) Jurisdictional diversity: Patents from different offices vary in legal and drafting conventions. EPD fills a critical gap by providing a benchmark of European patents to enable more comprehensive evaluation. (2) Quality improvement: EPD offers high-quality granted patents with finalized and legally approved texts, whereas others consist of patent applications that are unexamined or provisional. Experiments show that LLMs fine-tuned on EPD significantly outperform those trained on previous datasets and even GPT-4o in claim quality and cross-domain generalization. (3) Real-world simulation: We propose a difficult subset of EPD to better reflect real-world challenges. Results reveal that all tested LLMs perform substantially worse on challenging samples, which highlights the need for future research.
We present PatMT2 (Patent Machine Translation, v2), a two-stage LoRA pipeline that adapts the Hy-MT2-7B dedicated translation model to four-language (Chinese/English/Japanese/Korean) patent translation. On six translation directions, the Stage-2 checkpoint reaches a macro-average BLEU of 43.39 (BF16) to 43.57 (AWQ INT4), improving over the Hy-MT2-7B base by up to +8.8 BLEU and also exceeding Qwen3-32B and several external systems on the same average. We then measure the cost of adaptation: probing 38 languages absent from the fine-tuning data, the model still produces correct, fluent output on 34, while four low-resource languages collapse into distinct, reproducible failure modes. We treat these as diagnostic cases rather than retention rates, and report their mechanisms without attributing them to a single cause. Paired bootstrap tests confirm the translation gains over all baselines (p < 0.001).
Xiaowei Su, Xusen Cheng
Renmin University of China, Beijing, China · Beijing Bayuegua Technology Co., Ltd., Beijing, China
We present PatTransBench, a 12,000-item Chinese-English/Japanese/Korean patent machine translation (PatMT) benchmark with a four-dimensional evaluation protocol. The benchmark comprises 6,000 short-text (sentence/claim) and 6,000 long-text (paragraph) pairs, evenly split across three language pairs and both translation directions, built via IPC-section-stratified sampling and dualthreshold quality gating from upstream patent-family parallel corpora. The evaluation protocol spans literal n-gram overlap (BLEU, ROUGE-L, chrF), cross-lingual semantic textual similarity (CLSTS), 38-probe-language consistency testing, and serving performance under concurrency. A case study on PatMT2-7B and its fp8/AWQ4/AutoRound-4bit variants reveals direction-dependent quantization effects masked by aggregate scores and non-monotonic language-failure patterns across quantization schemes that are invisible to single-metric evaluation. We plan to release the benchmark, protocol, and case-study results upon publication to support reproducible PatMT evaluation.
Xiaowei Su, Xusen Cheng
Renmin University of China, Beijing, China · Beijing Bayuegua Technology Co., Ltd., Beijing, China