Pith. sign in

REVIEW 6 major objections 5 minor 37 references

PL-CA: A Parametric Legal Case Augmentation Framework

T0 review · 6 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that packing retrieved legal cases into an LLM's weights, rather than its prompt, lifts a 7B model's legal-article F1 to 24.01 against 15.43 for GPT-4o.

desk verdict A genuinely useful new legal benchmark wrapped around an efficiency claim that the paper never measures and that the per-query LoRA training procedure probably contradicts. read the letter →

arxiv 2509.06356 v1 pith:DBC2X43Q submitted 2025-09-08 cs.CL cs.AI

Ao Chang , Yubo Chen , Jun Zhao This is my paper · ORCID
classification cs.CLcs.AI
keywords parametricRAGLoRAinjectionlegaljudgmentpredictionstatutearticlegenerationmulti-taskbenchmarkChineseAIlong-contextdegradationdataaugmentation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper is trying to establish that retrieval-augmented generation for law does not have to work by stuffing retrieved case text into the LLM's prompt. Its PL-CA framework instead turns each legal case into a small LoRA weight update, so the model absorbs legal knowledge parametrically and keeps its context window short. If the claim holds, legal assistants could handle multi-step, mixed legal tasks with compact inputs and still beat both conventional RAG and much larger closed models on statute and judgment prediction. The paper also contributes Legal-CA, an expert-annotated multi-task benchmark covering criminal, administrative, and civil law, plus a 1,990-case parametric training corpus.

What carries the argument

The central mechanism is parametric RAG (P-RAG), adapted to legal cases and implemented through Low-Rank Adaptation (LoRA). Offline, each of 1,990 professional cases is rewritten by GPT-4o-mini into four variants of four components, multiplying each case into sixteen training pairs used to learn low-rank matrices $A_i$ and $B_i$; online, BM25 retrieves a similar case, the same parameterization is applied, and the trained adapter is merged into the frozen model. This turns retrieval output from prompt text into a weight update, which is what removes the retrieved document from the context and lets the model carry legal knowledge internally.

What would settle it

Measure wall-clock time and FLOPs for one online PL-CA query—retrieve, train one LoRA adapter on the retrieved case, merge, generate—and compare it with vanilla RAG on the same query; if PL-CA is not cheaper, the overhead-reduction claim fails regardless of accuracy. A second check is to see whether correctly generated statute articles actually appear in the top-1 retrieved case, or whether they come from the offline parametric corpus, which would reveal what the online retrieval path is contributing.

Watch

Extended reading notes

Core claim

PL-CA shows that a legal case can be decomposed into fact, dispute focus, reasoning, applicable statutes, and judgment; augmented into paraphrased variants; and encoded as a case-specific LoRA adapter that is merged into a frozen LLM. An offline stage parameterizes 1,990 professionally annotated cases, and an online stage retrieves one similar case from the Legal-KD corpus and parameterizes it on the spot before generation. On Legal-CA, Qwen1.5-7B-Chat with PL-CA reaches an LA-F1 of 24.01 on statute-article tasks versus 15.43 for GPT-4o-2024-11-20, and it outperforms vanilla RAG on most subtasks. This is taken as evidence that parametric knowledge injection is more effective than contextual injection for long legal documents, while also reducing the context-length overhead.

Load-bearing premise

The load-bearing premise is that the online stage's per-query process—training a LoRA adapter on the retrieved case and merging it into the model—costs no more than injecting the retrieved text as context; the paper reports no runtime or cost measurement, so if that premise fails the overhead-reduction claim collapses even if accuracy stays high.

Editorial extensions

If this is right

  • Legal LLMs can operate with short, roughly fixed-length prompts, because retrieved knowledge is carried in LoRA weights rather than in the input.
  • On the Legal-CA benchmark, PL-CA beats vanilla RAG on most subtasks and surpasses GPT-4o on statute-article F1 by about 8.5 points, all from a 7B open-weight base model.
  • The approach scales better with model size: Qwen1.5-7B-Chat shows clear gains from parametric injection, while the 1.8B model benefits more from plain context RAG.
  • Online and offline parametric injection both help, and online injection contributes more at equal data scale because the retrieved case is semantically closer to the query.
  • The Legal-CA dataset itself provides a multi-task, expert-annotated testbed that single-task benchmarks cannot offer, making the measured gains visible in the first place.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the per-query cost of LoRA training plus merging turns out acceptable, the same weight-injection idea transfers to other long-document, attention-diluted domains such as medical records, patents, or contracts, with no architectural change.
  • Because BM25 finds the correct statute in the top 5 only 3.6% of the time, the current online pipeline is often parameterizing a neighboring case rather than the decisive authority; a better retriever would likely raise the ceiling of P-RAG.
  • The offline stage is in effect a 1,990-case fine-tuning corpus, so the comparison against vanilla RAG does not isolate retrieval value from extra training signal; an offline-only, retrieval-free fine-tuned baseline at equal data would separate these two contributions.
  • Storing one LoRA adapter per case turns the retrieval index into a large parameter store, raising practical questions of storage cost, adapter routing, and model update hygiene that PL-CA does not address.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

6 major / 5 minor

Summary. The paper proposes PL-CA, a parametric retrieval-augmented generation framework for Chinese legal tasks. It combines an offline stage (LoRA fine-tuning on an augmented corpus of 1,990 legal cases) with an online stage (per-query retrieval of a similar case and LoRA fine-tuning on that retrieved case before answering) to 'encode' legal knowledge into model parameters instead of placing it in the context. The authors also introduce Legal-CA, a small expert-annotated multi-task benchmark (590 test / 1,990 training instances), and claim that PL-CA reduces the overhead of long contexts while matching or exceeding conventional RAG, and even surpasses GPT-4o on several legal subtasks. The paper's central efficiency claim is never measured, the online stage has a serious answer-leakage problem, and the reported results lack error bars and significance testing.

Significance. If validated, the idea of parameterizing retrieved legal knowledge could be of interest to the legal NLP community, and the Legal-CA benchmark would be a resource for multi-task evaluation. The authors have made a genuine effort to construct an expert-annotated Chinese legal dataset with structured components (fact, reason, articles, judgment). However, the current manuscript does not support its headline claims: the overhead-reduction claim is not backed by any efficiency measurement and is, as described, likely false because the online stage performs gradient-based LoRA training per query. The performance improvements are confounded by the online stage training on the retrieved case's known judgment and articles. The benchmark is not publicly accessible (no appendix or URL), so its community value cannot currently be realized. The paper is not in a publishable state.

major comments (6)
  1. [Online P-RAG, Eq. (2)-(3)] The abstract claims that PL-CA 'reduces the overhead associated with excessively long contexts,' but no runtime, FLOPs, GPU memory, or monetary cost is reported anywhere. The online stage trains a LoRA adapter on the retrieved case using next-token prediction (Eq. (2)), which requires at least one forward and backward pass through the LLM for each test query. This is generally more expensive than a single forward pass over a long context. Without any efficiency measurement, the central overhead-reduction claim is unsubstantiated and, as described, very likely false.
  2. [Online P-RAG, Figure 1] The online parametric injection trains the LoRA adapter on the full retrieved case, including its `Reason`, `Articles`, and `Judgment` components. The test query is then answered after this injection, meaning the model has been explicitly supervised on the answer (judgment and applicable articles) of a retrieved similar case before making its prediction. This is answer leakage and confounds the improvement of PL-CA over vanilla RAG in Table 4: the gain could come from memorizing the retrieved answer rather than from encoding legal knowledge. No control experiment removes the judgment/articles fields from the online training signal.
  3. [Results, Tables 4-7] The experimental comparisons lack error bars, statistical significance tests, and multiple random seeds. The test set has only 590 instances, and many numerical differences are small (for example, Table 4: Charge 92.39 vs. 88.62; Probation 40.52 vs. 40.79). The claim of consistent and significant improvement over baselines is therefore not established by the reported numbers.
  4. [Metric, Eq. (4)] The newly introduced evaluation metric d(r,h) = 1 - 1/(1 + exp(-|r-h|/(|r|+|h|+eps))) is not validated against human judgments or existing metrics. The LDG semantic-similarity metric based on chinese-roberta-wwm-ext is also used without any correlation analysis. Since the paper's central quality claims depend on these metrics, an unvalidated metric is a load-bearing issue.
  5. [Table 3 and Online P-RAG] BM25's article Recall@5 is only 3.6% (Table 3), meaning that the top-5 retrieved legal articles almost never contain the relevant statutes. The online stage nonetheless 'extracts the corresponding legal statutes from the top-5 retrieved cases' and uses them for parametric injection. The paper does not analyze how such low retrieval quality affects the injected knowledge, and it does not report the impact of retrieval errors on the final downstream performance.
  6. [Experiment Setting, Table 4] The comparison with GPT-4o is not matched: GPT-4o is evaluated as a zero-shot API baseline, while PL-CA is fine-tuned on 1,990 training instances and also uses online per-query adaptation. The claim of 'surpassing GPT-4o' is therefore not an apples-to-apples comparison and overstates the result.
minor comments (5)
  1. [Abstract] The abstract states 'Our code and dataset are provided in the appendix,' but the submission contains no appendix and no code or dataset link. This must be corrected.
  2. [Table 1] The 'avg token length' for Legal-CA is listed as 415, which conflicts with the text claim that Legal-CA 'ranks among the top in terms of average token count.' The column definition or the number appears to be inconsistent.
  3. [Table 4] The entry for gpt-4o's Fine column appears as '43.6325.75' (a missing space), making the table difficult to read.
  4. [Preliminary] The paper introduces the acronym 'SAG' for statute article generation but later uses 'SAR' in related work and 'LA' in the results. The terminology should be unified and defined consistently.
  5. [Equation (4)] The variable d is used both for documents (e.g., d_i in the offline stage) and for the metric function d(r,h). Renaming one of these would avoid confusion.

Circularity Check

1 steps flagged · score 6.0 of 10

Online P-RAG fits a LoRA adapter on the retrieved case's judgment and articles, then 'predicts' those same fields for the test query, so the SAG/LJP results are partly forced by construction; the overhead claim is unmeasured but not circular.

  1. fitted input called prediction [Section 'Online P-RAG', 'Parametric Injection' paragraph; Equation (2); also 'Offline P-RAG' answer construction]
    "In this setup, the case fact is treated as the query, while the remaining components serve as corresponding answers. ... The retriever performs online retrieval over the online corpus, selecting the top-1 most relevant case and extracting the corresponding legal statutes from the top-5 retrieved cases. ... After applying data augmentation, the parameterized update is carried out according to Equation 2, and the resulting parameters are injected into the LLM's internal representation."

    Equation (2) optimizes the LoRA adapter ΔΘ_i by next-token prediction on x = [q' ⊕ a'], where a' contains the retrieved case's reasoning, applicable statutes, and judgment. At generation time the same adapted model Θ_final is used to answer the test query, and the reported legal-article and judgment outputs are evaluated on those exact fields. For a test case whose facts align with the top-1 retrieved case, the correct charge and articles are the very labels used as supervision in the per-query LoRA update. The 'prediction' is therefore the fitted answer of a retrieved near-duplicate case, not an independent parametric generalization.

full rationale

PL-CA's reliance on P-RAG is an external citation (Su et al. 2025a), not a self-citation chain, and the offline dataset construction is a self-contained supervised procedure. The main circularity is confined to the online stage: a LoRA adapter is trained per query on the retrieved case's complete legal answer—including its judgment and applicable articles—and then used to generate the query's answer. This makes the legal-article and judgment predictions partly an evaluation of test-time memorization of the retrieved label rather than of independent parametric reasoning. The overhead-reduction claim is unsupported by any runtime or cost measurement, but that is an evidence gap, not a circular derivation. Considering the construction-level leakage in the central performance claim, a score of 6 is appropriate.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The central claim rests on standard LLM training assumptions and several domain-specific choices. The paper postulates that LoRA adapters trained on individual cases constitute a usable parametric memory, that BM25 retrieval with very low recall is adequate, and that a custom metric reflects legal accuracy. No free parameters are fitted to a target result, but several hyperparameters are chosen ad hoc.

free parameters (6)
  • LoRA rank r = 2
    Set by hand in the Hyperparameters section; the paper does not study how this affects the method.
  • LoRA alpha = 32
    Set by hand; no sensitivity analysis.
  • Learning rate = 1e-5
    Set by hand for LoRA training.
  • Training epochs = 1
    Set by hand.
  • Number of GPT-4o-mini rewrites per component = 3
    Each case component is rewritten three times, expanding the data by a factor of 4; this choice affects the size and diversity of the training set.
  • Online retrieval top-k = top-1 case, top-5 statutes
    The retriever selects the top-1 case and top-5 legal articles for online injection; no ablation on k is reported.
assumptions (5)
  • domain assumption Standard next-token prediction objective with LoRA training internalizes factual knowledge.
    The paper relies on this to claim that trained LoRA adapters represent the legal knowledge of each case.
  • domain assumption BM25 retrieval with Recall@5 of 3.6% (Table 3) is sufficient for the online stage.
    The paper adopts BM25 despite very low recall of relevant legal articles, and does not analyze how this affects the parametric injection.
  • domain assumption The test set, published after 2025-01-01 and drawn from Wenshu Court, does not overlap with the offline PKULaw corpus or online Legal-KD (2018-2021).
    No formal deduplication or overlap analysis is reported; the paper only states the date filter.
  • domain assumption LLMs perform better with parametric knowledge than with contextual knowledge (Yu and Ananiadou 2024).
    This citation is used as motivation; the paper does not test this claim independently.
  • domain assumption The custom metric in Equation 4 accurately measures numeric prediction quality.
    The relative-difference smooth function is introduced without validation against human judgments or standard metrics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PL-CA: A Parametric Legal Case Augmentation Framework." pith.science (2026). https://pith.science/paper/DBC2X43Q

@misc{pith2026250906356,
  author       = {Pith},
  title        = {Pith review of: PL-CA: A Parametric Legal Case Augmentation Framework},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DBC2X43Q}},
  note         = {Machine review of arXiv:2509.06356}
}
read the original abstract

Conventional RAG is considered one of the most effective methods for addressing model knowledge insufficiency and hallucination, particularly in the judicial domain that requires high levels of knowledge rigor, logical consistency, and content integrity. However, the conventional RAG method only injects retrieved documents directly into the model's context, which severely constrains models due to their limited context windows and introduces additional computational overhead through excessively long contexts, thereby disrupting models' attention and degrading performance on downstream tasks. Moreover, many existing benchmarks lack expert annotation and focus solely on individual downstream tasks while real-world legal scenarios consist of multiple mixed legal tasks, indicating conventional benchmarks' inadequacy for reflecting models' true capabilities. To address these limitations, we propose PL-CA, which introduces a parametric RAG (P-RAG) framework to perform data augmentation on corpus knowledge and encode this legal knowledge into parametric vectors, and then integrates this parametric knowledge into the LLM's feed-forward networks (FFN) via LoRA, thereby alleviating models' context pressure. Additionally, we also construct a multi-task legal dataset comprising more than 2000 training and test instances, which are all expert-annotated and manually verified. We conduct our experiments on our dataset, and the experimental results demonstrate that our method reduces the overhead associated with excessively long contexts while maintaining competitive performance on downstream tasks compared to conventional RAG. Our code and dataset are provided in the appendix.

Figures

Figures reproduced from arXiv: 2509.06356 by the authors.

Figure 1
Figure 1. The detailed structure of Test Cases and Parametric [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Comparison between traditional RAG and P-RAG. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Overview of the Case Augmentation and LLM Injection Pipeline. Our method involves five document sections: [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: The performance of plain RAG and structure RAG. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 7 canonical work pages

  1. [1]

    , " * write output.state after.block = add.period write newline

    ENTRY address archivePrefix author booktitle chapter edition editor eid eprint howpublished institution isbn journal key month note number organization pages publisher school series title type volume year label extra.label sort.label short.list INTEGERS output.state before.all mid.sentence after.sentence after.block FUNCTION init.state.consts #0 'before.a...

  2. [2]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION word.in bbl.in capitalize " " * FUNCT...

  3. [3]

    C.; Eren, M

    Barron, R. C.; Eren, M. E.; Serafimova, O. M.; Matuszek, C.; and Alexandrov, B. S. 2025. Bridging Legal Knowledge and AI: Retrieval-Augmented Generation with Vector Stores, Knowledge Graphs, and Hierarchical Non-negative Matrix Factorization. arXiv:2502.20364

    arXiv 2025

  4. [4]

    M.; and Søgaard, A

    Chalkidis, I.; Garneau, N.; Goanta, C.; Katz, D. M.; and Søgaard, A. 2023. LeXFiles and LegalLAMA: Facilitating English Multinational Legal Language Model Development. arXiv:2305.07507

    arXiv 2023

  5. [5]

    Chang, A.; Zhou, T.; Chen, Y.; Qiu, D.; Liu, S.; Liu, K.; and Zhao, J. 2025. ASP2LJ : An Adversarial Self-Play Laywer Augmented Legal Judgment Framework. arXiv:2506.18768

    arXiv 2025

  6. [6]

    Chen, G.; Fan, L.; Gong, Z.; Xie, N.; Li, Z.; Liu, Z.; Li, C.; Qu, Q.; Alinejad-Rokny, H.; Ni, S.; and Yang, M. 2025 a . AgentCourt: Simulating Court with Adversarial Evolvable Lawyer Agents. arXiv:2408.08089

    arXiv 2025

  7. [7]

    Chen, Z.; Ren, P.; Sun, F.; Wang, X.; Li, Y.; Zhao, S.; and Yang, T. 2025 b . SLARD : A C hinese Superior Legal Article Retrieval Dataset. In Rambow, O.; Wanner, L.; Apidianaki, M.; Al-Khalifa, H.; Eugenio, B. D.; and Schockaert, S., eds., Proceedings of the 31st International Conference on Computational Linguistics, 740--754. Abu Dhabi, UAE: Association ...

    work page 2025

  8. [8]

    Cui, J.; Ning, M.; Li, Z.; Chen, B.; Yan, Y.; Li, H.; Ling, B.; Tian, Y.; and Yuan, L. 2024. Chatlaw: A Multi-Agent Collaborative Legal Assistant with Knowledge Graph Enhanced Mixture-of-Experts Large Language Model. arXiv:2306.16092

    arXiv 2024

Show all 37 references
  1. [9]

    Cui, Y.; Che, W.; Liu, T.; Qin, B.; Yang, Z.; Wang, S.; and Hu, G. 2019. Pre-Training with Whole Word Masking for Chinese BERT. arXiv preprint arXiv:1906.08101

    2019 arXiv

  2. [10]

    Fan, W.; Zheng, T.; Hu, Y.; Deng, Z.; Wang, W.; Xu, B.; Li, C.; Li, H.; Shen, W.; and Song, Y. 2025. Legal Rule Induction: Towards Generalizable Principle Discovery from Analogous Judicial Precedents. arXiv:2505.14104

    2025 arXiv

  3. [11]

    Fei, Z.; Shen, X.; Zhu, D.; Zhou, F.; Han, Z.; Zhang, S.; Chen, K.; Shen, Z.; and Ge, J. 2023. LawBench: Benchmarking Legal Knowledge of Large Language Models. arXiv:2309.16289

    2023 arXiv

  4. [12]

    Gao, C.; Xiao, C.; Liu, Z.; Chen, H.; Liu, Z.; and Sun, M. 2024. Enhancing Legal Case Retrieval via Scaling High-quality Synthetic Query-Candidate Pairs. arXiv:2410.06581

    2024 arXiv

  5. [13]

    He, Z.; Cao, P.; Wang, C.; Jin, Z.; Chen, Y.; Xu, J.; Li, H.; Jiang, X.; Liu, K.; and Zhao, J. 2024. AgentsCourt: Building Judicial Decision-Making Agents with Court Debate Simulation and Legal Knowledge Augmentation. arXiv:2403.02959

    2024 arXiv

  6. [14]

    Kim, M.; Jung, H.; and Koo, M.-W. 2024. SELF - EXPERTISE : Knowledge-based Instruction Dataset Augmentation for a Legal Expert Language Model. In Duh, K.; Gomez, H.; and Bethard, S., eds., Findings of the Association for Computational Linguistics: NAACL 2024, 1098--1112. Mexic...

    2024

  7. [15]

    Li, H.; Ai, Q.; Chen, J.; Dong, Q.; Wu, Y.; Liu, Y.; Chen, C.; and Tian, Q. 2023 a . SAILER: Structure-aware Pre-trained Language Model for Legal Case Retrieval. arXiv:2304.11370

    2023 arXiv

  8. [16]

    Li, H.; Ai, Q.; Dong, Q.; and Liu, Y. 2024 a . Lexilaw: A Scalable Legal Language Model for Comprehensive Legal Understanding

    2024

  9. [17]

    Li, H.; Ai, Q.; Han, X.; Chen, J.; Dong, Q.; Liu, Y.; Chen, C.; and Tian, Q. 2024 b . DELTA: Pre-train a Discriminative Encoder for Legal Case Retrieval via Structural Word Alignment. arXiv:2403.18435

    2024 arXiv

  10. [18]

    Li, H.; Chen, J.; Yang, J.; Ai, Q.; Jia, W.; Liu, Y.; Lin, K.; Wu, Y.; Yuan, G.; Hu, Y.; Wang, W.; Liu, Y.; and Huang, M. 2024 c . LegalAgentBench: Evaluating LLM Agents in Legal Domain. arXiv:2412.17259

    2024 arXiv

  11. [19]

    Li, H.; Chen, Y.; Ai, Q.; Wu, Y.; Zhang, R.; and Liu, Y. 2024 d . LexEval: A Comprehensive Chinese Legal Benchmark for Evaluating Large Language Models. arXiv:2409.20288

    2024 arXiv

  12. [20]

    Li, H.; Shao, Y.; Wu, Y.; Ai, Q.; Ma, Y.; and Liu, Y. 2023 b . LeCaRDv2: A Large-Scale Chinese Legal Case Retrieval Dataset. arXiv:2310.17609

    2023 arXiv

  13. [21]

    Li, H.; Ye, J.; Hu, Y.; Chen, J.; Ai, Q.; Wu, Y.; Chen, J.; Chen, Y.; Luo, C.; Zhou, Q.; and Liu, Y. 2025. CaseGen: A Benchmark for Multi-Stage Legal Case Documents Generation. arXiv:2502.17943

    2025 arXiv

  14. [22]

    Li, J.; Yuan, Y.; and Zhang, Z. 2024. Enhancing LLM Factual Accuracy with RAG to Counter Hallucinations: A Case Study on Domain-Specific Queries in Private Knowledge-Bases. arXiv:2403.10446

    2024 arXiv

  15. [23]

    D.; Yue, X.; and Chen, W

    Li, T.; Zhang, G.; Do, Q. D.; Yue, X.; and Chen, W. 2024 e . Long-context LLMs Struggle with Long In-context Learning. arXiv:2404.02060

    2024 arXiv

  16. [24]

    N.; Patnaik, S.; Raghav, R.; and Kar, R

    Mullick, A.; Nandy, A.; Kapadnis, M. N.; Patnaik, S.; Raghav, R.; and Kar, R. 2022. An evaluation framework for legal document summarization. arXiv preprint arXiv:2205.08478

    2022 arXiv

  17. [25]

    Pipitone, N.; and Alami, G. H. 2024. LegalBench-RAG: A Benchmark for Retrieval-Augmented Generation in the Legal Domain. arXiv:2408.10343

    2024 arXiv

  18. [26]

    Qian, H.; Liu, Z.; Zhang, P.; Mao, K.; Zhou, Y.; Chen, X.; and Dou, Z. 2024. Are Long-LLMs A Necessity For Long-Context Tasks? arXiv:2405.15318

    2024 arXiv

  19. [27]

    Qin, W.; Cao, Z.; Yu, W.; Si, Z.; Chen, S.; and Xu, J. 2024. Explicitly Integrating Judgment Prediction with Legal Document Retrieval: A Law-Guided Generative Approach. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Ret...

    2024

  20. [28]

    Su, W.; Hu, Y.; Xie, A.; Ai, Q.; Bing, Q.; Zheng, N.; Liu, Y.; Shen, W.; and Liu, Y. 2024. STARD : A C hinese Statute Retrieval Dataset Derived from Real-life Queries by Non-professionals. In Al-Onaizan, Y.; Bansal, M.; and Chen, Y.-N., eds., Findings of the Association for Co...

    2024

  21. [29]

    Su, W.; Tang, Y.; Ai, Q.; Yan, J.; Wang, C.; Wang, H.; Ye, Z.; Zhou, Y.; and Liu, Y. 2025 a . Parametric Retrieval Augmented Generation. arXiv:2501.15915

    2025 arXiv

  22. [30]

    Su, W.; Yue, B.; Ai, Q.; Hu, Y.; Li, J.; Wang, C.; Zhang, K.; Wu, Y.; and Liu, Y. 2025 b . JuDGE: Benchmarking Judgment Document Generation for Chinese Legal System. arXiv:2503.14258

    2025 arXiv

  23. [31]

    Sun, J.; Dai, C.; Luo, Z.; Chang, Y.; and Li, Y. 2024. LawLuo: A Multi-Agent Collaborative Framework for Multi-Round Chinese Legal Consultation. arXiv:2407.16252

    2024 arXiv

  24. [32]

    Wu, Y.; Zhou, S.; Liu, Y.; Lu, W.; Liu, X.; Zhang, Y.; Sun, C.; Wu, F.; and Kuang, K. 2023. Precedent-Enhanced Legal Judgment Prediction with LLM and Domain-Model Collaboration. arXiv:2310.09241

    2023 arXiv

  25. [33]

    Xiao, C.; Hu, X.; Liu, Z.; Tu, C.; and Sun, M. 2021. Lawformer: A Pre-trained Language Model for Chinese Legal Long Documents. arXiv:2105.03887

    2021 arXiv

  26. [34]

    Xiao, C.; Zhong, H.; Guo, Z.; Tu, C.; Liu, Z.; Sun, M.; Feng, Y.; Han, X.; Hu, Z.; Wang, H.; and Xu, J. 2018. CAIL2018: A Large-Scale Legal Dataset for Judgment Prediction. arXiv:1807.02478

    2018 arXiv

  27. [35]

    Yu, Z.; and Ananiadou, S. 2024. Neuron-Level Knowledge Attribution in Large Language Models. arXiv:2312.12141

    2024 arXiv

  28. [36]

    Yue, S.; Huang, T.; Jia, Z.; Wang, S.; Liu, S.; Song, Y.; Huang, X.; and Wei, Z. 2025. Multi-Agent Simulator Drives Language Models for Legal Intensive Interaction. arXiv:2502.06882

    2025 arXiv

  29. [37]

    Zhang, K.; Yu, W.; Dai, S.; and Xu, J. 2025. CitaLaw: Enhancing LLM with Citations in Legal Domain. arXiv:2412.14556

    2025 arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.