Legal case retrieval (LCR) defines relevance beyond text similarity, judgments may describe near-identical events yet be decided under different charges. Existing synthesis pipelines mine training pairs from text similarity and shared charges, which separates positives from negatives but leaves the difficulty of each pair implicit. We present Deka Retrieval, an LCR benchmark on Thai Supreme Court decisions with an LLM relevance cases mining pipeline for a civil-law jurisdiction. We decompose each judgment into case components (Charge, Fact, Reasoning, Decision) and split easy/hard level on exist binary signals and introduce an anomaly variant a conflict judgment that shares the anchor’s charge and facts but reaching a different reasoning and decision. Retriever performance follows query length: Fact > Lawyer expert > Lay people. Fine-tuning gains most on the weakest type, lifting Lay R@10 by 2.6x (7.6 -> 19.5) versus 1.5x on Lawyer and 1.3x on Fact. Training on the shortest query also transfers to the longer ones, but not the reverse. Adding anomaly cases as positives raises the chance of retrieving them at test time and degrades performance relative to training without them. Our dataset and models are available at https://hf.co/collections/VISAI-AI/deka-retrieval.

Free-form legal essay evaluation in NLP treats expert inter-rater stability as a single ceiling number: a correlation score between credentialed examiners is reported and used as the target that an LLM judge should approach. Whether that ceiling is uniform across items has rarely been tested on the expert side. We test it on Thai bar exam essays. Three Bar Council examiners and three anchor LLM judges (Claude 4.6 Opus, Gemini 3.1 Pro, GPT-5.4 Pro) independently graded the same 15 answers from the same four inputs: question, official rubric, gold answer, and student answer. On 10 of the 15 answers, both panels reach 𝛼 = 0.75. On the remaining 5, drawn from one question where the rubric does not prescribe how to grade a correct answer that omits a decisive citation, the human panel collapses to 𝛼 = -0.34 while the LLM panel holds at 𝛼 = 0.68. One examiner’s stricter reading of this underspecified rubric drives the gap. The cluster pattern replicates across seven of eight LLMs that took the exam in Phase 1 and were reused as judges, regardless of their own exam scores: the same model behaves differently in the two roles. The expert stability ceiling is not flat: free-form essay rubrics combine a deterministic correctness element (grounded by the gold answer) with an interpretive quality element (judged without one), and expert disagreement lives entirely on the interpretive element. The bottleneck of free-form essay evaluation is the rubric’s interpretive coverage, not LLM capability.

Retrieval-augmented generation depends on both the retrieval method and the units over which retrieval operates. Yet retrieval-unit choices remain underexplored for Thai legal documents, where language and document structure challenge common chunking assumptions. We introduce a 2,000-question Thai Royal Gazette dataset with single-hop and multi-hop tasks. The dataset annotates gold evidence as document character spans, which permits unit-independent evaluation. We compare 19 configurations across chunk sizes, boundary strategies, contextual chunks, atomic facts, whole documents, and retrieval methods. We evaluate them with span recall and precision, an Evidence-Fit judge, and pairwise judgments of answers generated from retrieved evidence. Contextual 200-word chunks with hybrid retrieval provide the strongest practical balance of evidence coverage, answer quality, and query cost. Across configurations, downstream scores correlate strongly with span recall, while span precision shows little association. These results favor retrieval designs that first recover complete evidence, then control cost and localization.

When a government reviews its laws, the public speaks in everyday words and rarely names a provision directly, so turning complaints into amendment targets is still expert work done by hand. We describe a pipeline deployed inside Thailand’s 2026 Regulatory Guillotine that maps 199,771 social media posts and 63 proposals from business associations to the related specific provisions. Our pipeline first removes irrelevant posts. A legal language model then labels each post with a written rationale, and the posts’ embeddings are clustered into 309 issues. The model grounds each issue in specific provisions using a Thai legal knowledge graph, which it must read before it cites, leaving 799 verified sections in 80 laws. The same graph then shows what an amendment would touch, namely the other sections, definitions, and case law that depend on each targeted section, a check that drafters currently do by hand. We report the system, what the Thai public asked to change, and the blast radius of the sections they targeted.

Multimodal retrieval systems are expected to operate in a semantic space, agnostic to the language or cultural origin of the query. In practice, however, retrieval outcomes systematically reflect perspectival biases: deviations shaped by linguistic prevalence and cultural associations. We introduce the Cross-Cultural, Cross-Modal, Cross-lingual Multimodal (3XCM) benchmark to isolate these effects. Results from our studies indicate that, for image-to-text retrieval, models tend to favor entries from prevalent languages over those that are semantically faithful. For text-to-image retrieval, we observe a consistent "tugging effect” in the joint embedding space between semantic alignment and language-conditioned cultural association. When semantic representations are insufficiently resolved, particularly in low-resource languages, similarity is increasingly governed by culturally familiar visual patterns, leading to systematic association bias in retrieval. Our findings suggest that achieving equitable multimodal retrieval necessitates targeted strategies that explicitly decouple language from culture, rather than relying solely on broader data exposure. This work highlights the need to treat linguistic and cultural biases as distinct, measurable challenges in multimodal representation learning.

Jurisdictionally bound domains, such as law, often lack standardized, machine-readable data formats, requiring foundational infrastructure before downstream applications can succeed. We present ThLexGraph, the first unified temporal knowledge graph for Thai legal data, integrating 3,840 laws (6,273 versions) with 87,394 Supreme Court decisions, updated daily. The graph encodes hierarchy, temporal versioning, cross-references, and sequential order, all extracted from unstructured official sources where no structured representation previously existed. A five-setting comparison on NitiBench-Tax isolates data infrastructure as the sole variable: graph-structured retrieval achieves Citation F1 of 0.812 versus 0.666 for practitioner-standard web search and 0.685 for flat vector retrieval, while searching a corpus 53x larger. Trace analysis of 820 agent-issued queries reveals that hierarchy traversal and cross-reference following, capabilities absent from generic retrieval, are exercised in 50% and 16% of questions, respectively. Our system demonstrates that structured modeling of hierarchy, temporal versioning, cross-references, and sequential order can overcome structural limitations of legal data published without standardized formats.

Legal practitioners in Thailand must navigate fragmented government websites to research over 3,800 active laws and 87,000 Supreme Court decisions, with no unified tool for cross-referencing, version tracking, or structural navigation. We present FourCorners, a deployed platform that addresses five practitioner pain points through three modules built on a temporal legal knowledge graph covering 552K nodes and 6.3M edges: (1) an AI legal assistant that performs grounded generation via structured graph retrieval, streaming verified source content inline with responses; (2) an interactive law reader that translates graph structure into navigation and comparison features; and (3) a court decision explorer with version-aware citations produced by temporal entity resolution across 87,394 rulings. The system discovers implicit cross-corpus relationships through co-citation analysis of Supreme Court decisions. Interviews with 20 legal professionals reveal that inline source verification fundamentally changes how practitioners interact with AI-generated legal content, and that cross-corpus enrichment surfaces legal relationships that existing tools leave invisible.

The Retrieval-Augmented Generation (RAG) systems’ performance on Thai legal question answering is still limited, especially for questions requiring extensive, complex legal reasoning. To address these limitations, we introduce a resource-efficient approach that aligns Large Language Models (LLMs) for improved citation accuracy and response quality using Group-Relative Policy Optimization (GRPO). Our proposed method leverages BGE-M3 embeddings as a cost-efficient semantic-similarity reward, significantly reducing computational expenses up to 2.5x compared to an LLM-based reward model. Experiments on the NitiBench benchmark demonstrate substantial improvements: GRPO achieves up to 90% citation-F1 gains relative to the base model and a 31% increase in joint quality metrics over instruction tuning. Crucially, our approach provides a practical and effective solution for enhancing legal LLMs in resource-constrained environments.

This paper details our system for the JUST-NLP 2025 Shared Task on English-to-Hindi Legal Machine Translation. We propose a novel two-stage, data-centric approach. First, we annotate the training data by translation difficulty and create easy and hard subsets.We perform SFT on the easier subset to establish a robust “cold start”. Then, we apply RLVR exclusively on the harder subset, using machine translation metrics as reward signals. This strategy allowed our system to significantly outperform strong baselines, demonstrating the capability of our systems for machine translation tasks. Source code and model weights are available at https://github.com/ppaolong/FourCorners-JustNLP-MT-Shared-Task

We describe an inexpensive system that ranked first in the JUST-NLP 2025 L-SUMM task, summarizing very long Indian court judgments (up to 857k characters) using a single 80GB GPU and a total budget of about $50. Our pipeline first filters out length–summary outliers, then applies two-stage LoRA SFT on Qwen3-4B-Instruct-2507 to learn style and extend context, and finally runs RLVR tuned to BLEU, ROUGE-2, and ROUGE-L, with BLEU upweighted. We showed that two-stage SFT is better than a single-stage run, and RLVR gives the largest gains, reaching 32.71 internal vs. 16.15 base and 29.91 on the test leaderboard. In ablation on prompting, we find that a simple, naive prompt converges faster but saturates earlier, while the curated legal-structured prompt keeps improving with longer training and yields higher final scores, and the finetuned model remains fairly robust to unseen prompts. Our code are fully open-sourced, available for reproducibility.

Large language models (LLMs) show promise in legal question answering (QA), yet Thai legal QA systems face challenges due to limited data and complex legal structures. We introduce NitiBench, a novel benchmark featuring two datasets: (1) NitiBench-CCL, covering Thai financial laws, and (2) NitiBench-Tax, containing Thailand’s official tax rulings. Our benchmark also consists of specialized evaluation metrics suited for Thai legal QA. We evaluate retrieval-augmented generation (RAG) and long-context LLM (LCLM) approaches across three key dimensions: (1) the benefits of domain-specific techniques like hierarchy-aware chunking and cross-referencing, (2) comparative performance of RAG components, e.g., retrievers and LLMs, and (3) the potential of long-context LLMs to replace traditional RAG systems. Our results reveal that domain-specific components slightly improve over naive methods. At the same time, existing retrieval models still struggle with complex legal queries, and long-context LLMs have limitations in consistent legal reasoning. Our study highlights current limitations in Thai legal NLP and lays a foundation for future research in this emerging domain.