regulatory compliance teams

Abstract: Each new generation of frontier model pushes the edge of what is possible. Inside a harness -- a loop, a shell, a filesystem, the ability to write and run its own code -- a model can take on increasingly difficult work. A growing body of work treats the coding-agent harness as infrastructure for knowledge work rather than a coding tool, and a new wave of benchmarks tests it there. Most of them are generative: build a financial model, a memo, a spreadsheet from a brief, judged by a rubric or by another model. Few test the expensive part of knowledge work, which is investigative. In audit, tax, due diligence, regulatory reporting, claims or any dispute over numbers, the answer already exists -- hidden in incomplete and inconsistent evidence spread across systems -- and has to be recovered and defended to the cent. We introduce kw-bench, a benchmark built from real engagements specialists described to us. Its first task family is a five-year VAT reconstruction across three generations of SAP -- legacy ECC, a partial S/4HANA migration and a consultant's redesign -- delivered as raw system extracts: page headers, local number formats, partial link tables, no documentation of how anything connects. Grading is deterministic, with no LLM judge. We ran four harnesses -- Claude Code, Codex, OpenCode and micro-cc -- on the same model, DeepSeek v4.1 Flash, for two hours. The two independent, model-agnostic harnesses delivered (OpenCode 0.542, micro-cc 0.528); the two lab harnesses, away from their home model, delivered nothing (0.00 each). What defeated them was inertia: keeping many weak signals alive across a long session. On a frontier model the task collapses. micro-cc and Claude Code on Claude Opus 5.5 scored 0.9986 and 1.000, in 22 and 25 minutes. A weeks-long engagement now costs minutes and a few dollars; below the frontier, the harness decides how far a model gets.

Abstract: Change is inevitable in software systems, where requirements must continuously adapt to evolving stakeholder needs, regulations, and project scopes. Managing such changes is a fundamental software engineering challenge, as manually identifying the impact of a change request, given high requirements volume, Natural Language (NL) complexity, and intricate interdependencies, is error-prone and time-consuming. Large Language Models (LLMs) offer promising solutions for requirements change impact analysis (CIA), yet their effectiveness when enhanced with advanced techniques remains underexplored. We present DRILS, an LLM-based agentic framework for requirements Change Impact Analysis, employing a multi-agent architecture to generate candidate impact sets, with an LLM-as-a-judge promoting diversity during selection. DRILS features two key components: RE-RAG, a carefully engineered RAG technique providing focused domain context, and a requirement dependency graph (RDG) as an additional knowledge source. Evaluation across three manually curated datasets demonstrates superior performance over state-of-the-art baselines, achieving an average recall of 87.1%, precision of 76.3%, and F1 score of 81% at a lower cost. Precision gains are particularly significant, 41.1%, 83.4%, and 149.8% over the closest baseline per dataset, as higher precision directly reduces false positives and manual validation effort. We also contribute three annotated CIA datasets to support future research.

Abstract: Engineering standards are challenging knowledge sources for large language models (LLMs) because they are hierarchical, multimodal, and densely connected through figures, tables, cross-references, and normative rules. Existing retrieval-augmented generation (RAG) methods often rely on unstructured text retrieval and struggle with questions requiring joint reasoning over textual provisions, visual specifications, and document structure. We propose GEMS-RAG (Graph-grounded Engineering Multimodal Standards Retrieval-Augmented Generation), a framework for grounded reasoning over engineering standards. GEMS-RAG combines hierarchical parsing, fine-grained visual entity extraction, multimodal knowledge graph construction, and graph-grounded hybrid retrieval. Instead of treating figures as monolithic images, it decomposes them into meaningful visual entities linked to textual provisions, engineering concepts, and regulatory rules. During inference, dense, sparse, and visual retrieval are combined with graph expansion to identify relevant evidence across text, figures, and document structure. We evaluate GEMS-RAG on the Manual on Uniform Traffic Control Devices (MUTCD), a large engineering standard containing egulations, technical drawings, and standardized traffic-control devices. Results show that graph-grounded multimodal retrieval improves retrieval effectiveness and question-answering performance over text-only, multimodal, and graph-based RAG baselines. Beyond MUTCD, GEMS-RAG offers a general framework for reasoning over engineering codes, technical manuals, regulatory documents, and scientific standards in safety-critical applications.

Abstract: Legal language understanding requires reasoning over semi-structured regulatory documents, yet standard pre-training objectives do not explicitly target this structural capability. Because current legal language models rely solely on masked language modeling, they treat statutory text as flat token sequences and ignore the underlying hierarchical constraints. To bridge this gap, we present a multi-task pre-training approach for German traffic law (StVO). Our approach combines three complementary objectives: Next Sentence Prediction (NSP), Knowledge Graph (KG) Guided Clustering (CLU), and Knowledge Graph (KG) Aware Paragraph Classification (CLS). This approach explicitly encodes structured legal knowledge directly into the model weights. This encoding enables zero-shot legal reasoning at inference time and removes the need for discrete graph traversal. We evaluate our approach across nine Transformer-based encoders and 324 experimental runs. On synthetic StVO question answering, our models demonstrate consistent relative improvements between 10.3% and 155.0% over baselines. Furthermore, relative to the established LexGLUE baselines (Chalkidis et al., 2022), our models improve over prior results on four of seven legal benchmarks. It also achieves consistent zero-shot transfer gains between 3.4% and 15.8% on Austrian driving regulations. These experimental results indicate that our training approach generalizes successfully beyond the initial training legal domain.

Abstract: Spin in clinical trials includes reporting practices that distort the presentation of results. This is particularly critical in medicine, where spin is present in more than 50% of randomized controlled trials that fail to reach statistical significance. The comparison of primary and reported outcomes is crucial for detecting several types of spin, including outcome switching. We used 300 pairs of outcomes labeled with semantic similarity to develop a system for automatic detection of outcome switching. We evaluated baseline text similarity models and open-source LLMs using generated similarity scores and the Youden index to determine the classification threshold. The proposed approach involves prompt engineering, classification based on token probabilities, and majority voting for the final decision. The results on the test set of 2,496 examples with an F1 score of 0.78 and an accuracy of 0.90 outperform baseline text similarity models but trail behind fine-tuned versions of BERT. We used LLMs to generate natural language explanations for the classified instances and manually assessed their quality.

Abstract: As artificial intelligence is increasingly deployed, algorithmic unfairness has raised growing concerns and intensified demands for transparent fairness auditing. In practice, the tolerable degree of algorithmic unfairness depends on the specific legal, ethical, or application context. Given a prespecified tolerance threshold, an important statistical question is how to determine whether a group disparity exceeds the allowable tolerance across different auditing objectives. To address this problem, we develop a unified tolerance-based fairness auditing framework for two complementary auditing objectives: violation certification, which prioritizes control of false violation declarations, and sensitivity screening, which prioritizes reducing missed violations. For the first objective, we develop a constrained empirical likelihood test for formal settings that uses least-favorable-point calibration and can be combined with false flagging rate control for simultaneous subgroup auditing. For the second objective, we develop split empirical likelihood and adjusted split empirical likelihood tests using an adaptive boundary-proxy principle for early-warning settings. Numerical experiments show the distinct error-control--sensitivity trade-offs of these procedures. A COMPAS analysis illustrates the framework in predictive fairness auditing.

Abstract: The EU AI Act (Art. 27) requires deployers of high-risk AI systems to conduct Fundamental Rights Impact Assessments (FRIAs) before deployment, yet the evidence needed for credible assessments is fragmented across incompatible incident repositories, risk vocabularies, and legal texts. We present a reusable Semantic Web-based framework that consolidates this evidence for two high-risk public sector categories: employment and worker management (Annex III(4)) and access to essential public services (Annex III(5)(a)). A curated 150-record corpus is annotated along four axes using keyword, LLM, and hybrid methods and serialised as a SPARQL-queryable knowledge graph of 1,351 RDF triples. Five FRIA demonstration scenarios surface 103 records (68.7% coverage). Evaluation against a 69-record gold standard reveals that LLM-assisted classification of the employment domain achieves only $κ= 0.045$, a cautionary result for automated fairness-related evidence retrieval in this domain. All artefacts are released openly to support adoption by regulators, national authorities, and SMEs.

Abstract: The disambiguation of semantically similar statutory text across jurisdictions is a retrieval problem that existing methods do not solve. This inter-context conflict can steer generative models toward confidently produced answers grounded in topically relevant but jurisdictionally incorrect sources. Tobacco and nicotine regulations vary by US jurisdiction, often sharing similar language, thus, robust reasoning requires identifying which jurisdiction's law governs a given product, not merely retrieving relevant text. Emerging products (e.g., pouches) exploit ambiguous definitions to evade regulation. State-of-the-art (SOTA) document retrieval-augmented generation (RAG) methods struggle to address this inter-context conflict, and thus struggle to connect image attributes (e.g., rich attribute captions) to the set of similar legislation texts. We introduce NicoPRISM (Nicotine Product and Regulation Image-and-Text Surveillance Multimodal), comprising 161,563 images, attribute captions, a knowledge base of product, health, and legislative documents spanning 13 US jurisdictions, and 1,495 validated question-answer pairs across two tasks: policy compliance QA and product knowledge QA. We also propose PRISM-RAG, a multimodal hypergraph RAG framework built over images, captions, and entities without any LLM calls at index time, grounding every query in a product image and routes retrieval through a jurisdiction-aware context assembly mechanism guaranteeing that statutory text from the queried jurisdiction reaches the language model by construction. PRISM-RAG retrieves passages from the correct jurisdiction in 93.9% of policy compliance queries, a 48.6 percentage point advantage over standard RAG (p<0.001), using zero LLM calls at index time and one at query time, and is competitive with or outperforms SOTA RAG frameworks across keyword, semantic, jurisdiction-, and compliance-accuracy metrics.

Abstract: Open-weight large language models are rapidly entering hiring pipelines, yet their discriminatory failure modes -- and the regulatory exposure these create under the EU AI Act high-risk classification (Annex III) and U.S. EEOC adverse-impact analysis -- remain poorly understood. We present the first systematic, multi-model audit of open-weight LLMs that treats job-posting language as the primary experimental variable, evaluating six models (Llama 3.2, Mistral, Gemma 3, Qwen 3, Phi 3, DeepSeek-R1) across four controlled experiments that jointly probe recruiter-simulation and job-seeker-simulation tasks. We find that (1) agentic posting language depresses recruiter recommendation scores for female candidates (r_rb = 0.309, p_Bonf = 7x10^-5; model-fixed-effects r_rb = 0.448), while communal language partially reverses the penalty; and (2) coded-exclusion language suppresses non-White recruiter scores at large effect sizes (r_rb = 0.646-0.758) and, on the job-seeker side, selectively deters non-White personas from expressing interest -- operationalizing a chilling-effect mechanism at scale. A label-ablation experiment isolates the explicit demographic persona label as the primary causal driver, and Word Embedding Association Tests corroborate these findings at the representational level (d = 1.01-1.45 under Caliskan et al.'s multi-word gender attribute lists). We translate these results into a concrete pre-deployment audit protocol -- posting-vocabulary scoring, persona-conditioned LLM probing, and adverse-impact flagging against the four-fifths threshold -- that operationalizes the documentation and risk-management obligations Annex III imposes on high-risk AI in recruitment.