CS Paper Radar

GT-FENG: General Theory of Factual Evaluation and Normative Grounding in Language Models

Jiayu Feng

cs.CL ·submitted 2026-11-13 15:11 UTC · v2
code/data

Abstract

Whether a language model's answer is correct often depends on context: the same value can be right under one jurisdiction, date, or legal definition and wrong under another. We introduce GT-FENG (General Theory of Factual Evaluation and Normative Grounding), the first general framework to unify contextual/temporal correctness and rule-based error attribution across law, policy, and other rule-governed domains. GT-FENG checks answers against the applicable rules and identifies documented rules that match incorrect answers. It shows that when multiple rules give the same value, the answer alone cannot tell which rule the model used. In Medicaid income-limit answers from two models, documenting the asked state's own records reduces the number of wrong answers that match only other states' values by 68-78%, and more than half of stable wrong answers then match two or more documented sources. GT-FENG provides a foundation for evaluating language models in rule-governed domains and for future benchmarks and evaluation toolkits. We release GT-FENG as an open-source Python package, gtfeng.

Comment
7 pages, 4 tables. Code: https://github.com/VonFlair/gtfeng

Closest papers

Nearest neighbours by meaning across the whole index.

Can We Trust the Judges? Validation of Factuality Evaluation Methods via Answer Perturbation

cs.CL ·2026-09-14 ·Sarra Gharsallah, Adele Robaldo, Mariia Tokareva, Giovanni Gatti Pinheiro, Ilyana Guendouz +3 more
published similarity 0.65
Abstract · 22 pages, 7 figures. Extended version of a paper accepted at EvalLLM 2025 (CORIA-TALN 2025)

Evaluating the factual correctness of large language models (LLMs) is vital for many applications. But are our evaluation tools themselves trustworthy? Despite the rise of factuality-based metrics, their sensitivity and reliability remain underexplored. This paper introduces a meta-evaluation framework that systematically tests these metrics using controlled corruptions of gold standard answers. Our method generates ranked outputs with known degrees of degradation to probe how metrics capture nuanced changes in truthfulness. Our experiments reveal that pipeline-based methods, such as the RAGAS's factual correctness metric, better track degradation than LLM-as-judge approaches. We also propose a new variant of the factual correctness metric that provides a competitive and cost-efficient.

0.26

Do Large Language Models Know Colombian Law? A Reliability Benchmark for the Colombian Legal System

cs.AI cs.CY ·2026-10-02 ·Rubén Manrique, Michelle Castellanos, Jorge Morales, Juan David Gutiérrez, Antonio Barreto Rozo +1 more
benchmark similarity 0.64
Abstract · 38 pages, 23 figures, 8 tables

Large language models (LLMs) are increasingly used to support legal practice, education, and research, yet their reliability in national legal systems outside the United States remains largely undocumented. We introduce an expert-validated benchmark for evaluating LLM reliability on the Colombian legal system. The benchmark comprises 1,042 items spanning ten areas of law and three question formats (closed multiple-choice, semi-open, and open-ended IRAC), built through a human-in-the-loop pipeline with multi-stage expert review. We evaluate 15 contemporary proprietary and open-weight models with format-appropriate metrics. Accuracy on closed questions ranges widely, from 0.905 (Gemini 3.1 Pro) to 0.577, but on free-text legal answers factual correctness never exceeds 0.45 (on a 0-1 scale) for any model. We find a dissociation between answer relevancy and correctness (Spearman rho = -0.46): models reliably sound responsive while frequently being wrong, a pattern of particular concern for non-expert users. Closed-question accuracy and free-text correctness are strongly rank-correlated (rho = 0.94), so cheap multiple-choice screening predicts model ranking but overstates absolute reliability. An independent rubric-based LLM judge and blind human expert scoring both reproduce the free-text ranking (rho >= 0.88). The judge further reveals that only about half of the norms models cite are correct; the rest are wrong or non-existent. Reliability varies systematically by legal area and follows an inverted-U across question complexity. Our results indicate that current LLMs require expert supervision for Colombian legal tasks, and that grounding answers in authoritative sources is a promising path to higher reliability. We release the benchmark construction pipeline to support reproducible evaluation.

0.11

Do Small Models Use the Law You Give Them? Measuring Context Use on a Bilingual Bangladesh Legal Benchmark

cs.CL ·2026-08-31 ·Moniruzzaman Mahadi, Abrar Mohammed Tanzim Alam, Sayma Siddika Monalisa, Mir Mohammad Asif Abdullah, Swakkhar Shatabda +1 more
benchmarkcode/data similarity 0.63
Abstract · Legal Data Benchmark for Bangladesh

Fine-tuning can improve legal question-answering accuracy without improving how models use law supplied in context. We study this distinction in bilingual Bangladeshi legal QA, where observed errors can arise from answer scoring, retrieval, or failure to use relevant law. We construct a hierarchy-preserving statutory corpus, 2,165 reviewed bilingual fine-tuning examples, and a 150-item supplied-law control. We evaluate six instruction-tuned models: Llama-3.2-1B, Llama-3.2-3B, Qwen3.5-0.8B, Qwen3.5-2B, Qwen3.5-4B, and Gemma-4-E2B, with three LoRA seeds per model. To separate effects, we combine constrained option-letter scoring, cyclic option rotation, and controlled removal of the governing provision. On 398 Bar Council outputs, an exact-line parser attributes an accuracy gain of 50.0\% to the Qwen3.5-2B seed-42 adapter, whereas option scoring yields only $3.0\%$. For Gemma-4-E2B, the two scoring methods favor different systems. When the governing provision is guaranteed to be present, five of six reference models improve by $14.7\%-19.3\%$ under the four-order criterion. Removing that provision reduces accuracy by $8.0\%-15.3\%$ for models and by $13.8\%-14.9\%$ points for their adapters. However, difference-in differences estimates show no increase in reliance on the governing provision after fine-tuning. Results show that legal adaptation claims require separating scorer, retriever, and model effects. Our Code and data are available at https://anonymous.4open.science/r/bangladesh-legal-qa-11E3

0.35

GRACE: Grounded Adversarial Reasoning over Canadian Law

cs.CL cs.AI ·2026-09-20 ·Jiakang Xu, Wantong Huo, Udom Silparcha, Jonathan H. Chan
similarity 0.62
Abstract

Large language models have shown strong performance across a range of legal tasks, but existing benchmarks rarely evaluate the ability to take and defend a legal position, reason under incomplete information, or synthesize multiple statutory provisions. This gap is particularly pronounced for Canadian law, which remains underrepresented in legal NLP. We introduce GRACE (Grounded Reasoning Adversarial Canadian LEgal examples), a dataset of 1,915 question-reasoning-answer instances grounded in Canadian federal legislation. GRACE covers three reasoning modes: adversarial advocacy, uncertainty, and applied reasoning. We develop a pipeline that partitions raw statutory text, generates scenario-based questions and reasoning, and filters examples through model-free citation verification and LLM-based quality auditing. As a proof of concept, we fine-tune CLeAR-4B (Canadian Legal Adversarial Reasoning), a lightweight model for grounded legal reasoning, and evaluate it against the unmodified Qwen3-4B base model in open- and closed-book settings. CLeAR-4B substantially improves agreement with teacher outputs and statutory citation behavior when the relevant act text is provided, while its grounding degrades sharply when the statute is withheld. These results suggest that GRACE can support the development of lightweight legal models that reason more effectively from supplied statutory text.

0.24

When Local Definitions Are Missing: Stable Errors in Language Models

cs.CL cs.IR ·2026-11-11 ·Zengqing Wu, Chuan Xiao
similarity 0.62
Abstract

Contracts, technical manuals, and online communities often give familiar words special meanings and define them in one place. A language model that reads a passage without that definition may use the usual meaning and give an answer that sounds right but is wrong. We study this with multiple-choice questions whose correct answers depend on a local definition. The questions come from a synthetic vocabulary, commercial contracts, Reddit community glossaries, and the documentation of a decentralized exchange. We ask three language models each question with and without the definition. Without using the answer key, we then test whether their repeated answers and stated confidence reveal their own errors. Separately, we give a small open model the answer key and test whether it can predict which questions larger models get wrong. We report three findings. (1) Familiar words that a source redefines cause the most errors. Models often repeat the same wrong answer, and neither repeated answers nor stated confidence reliably flag it. (2) The source definition fixes most errors on synthetic questions, but a definition drawn from the model's own prior knowledge does not. In real contracts, the source definition fixes only about half of the errors. (3) Given an answer key, a small model can rank questions by how likely larger models are to get them wrong. These results call for tests that check answers against the full source, including its local definitions.

0.20

Verdict Without the Rule: Diagnosing and Auditing Regulatory Rule Sensitivity in LLM Compliance Systems

cs.AI cs.CLcs.CY ·2026-10-08 ·Saisab Sadhu, Aadit Sengupta, Vinay kumar Sankarapu, Pratinav Seth
similarity 0.61
Abstract

Large language model compliance systems are deployed on the assumption that a verdict depends on the regulatory rule it is given. We test this directly across five models and 20 regulatory and platform-policy domains: delete, swap, or negate the governing rule while holding the case fixed, and check whether the verdict changes (OCS) or the model's internal representation of compliance shifts at all (ICS-delta). Neither moves much: models' verdicts are often invariant to substantial perturbations of the supplied rule, and the guard model, evaluated here under a custom-rule adaptation of its native taxonomy, is the least rule-sensitive and least accurate of the five, barely above chance (51%, versus 90-92% for general-purpose models). This reflects easy cases more than blanket neglect: on cases where deleting the rule changes a previously correct model prediction, models do track it closely. Neither better prompting nor direct intervention on the model's internal representations closes this gap. Accuracy alone does not establish that a compliance verdict is grounded in the supplied rule.

0.13

The Limit Is in the Input, Not the Model: State-Dependent Command Safety in Natural-Language Interfaces to Industrial Controllers

cs.CR cs.SE ·2026-11-18 ·Hemen Taukoor, Zhaoxia Wang, Venky Shankararaman
published similarity 0.60
Abstract · 8 pages. Accepted for publication in the 2026 IEEE International Conference on Data Mining Workshops (ICDMW), 6th Machine Learning for Cyber…

Language models judging commands to industrial controllers may not receive machine state, the equipment's current condition. The same command can require opposite safety decisions in different states, so an incorrect answer may reflect missing information rather than poor reasoning. We propose an answerability audit: does the model receive the information needed to decide? The audit records the model's inputs and pairs the same command with two states requiring opposite answers. The paired score counts a pair only when both answers are correct. We add rules, then state values, to the model's input. In an author-built test of 15 pairs from three systems with four language models, none reliably distinguished the paired cases without state, even with rules. With state, scores ranged from 0 to 13 correct pairs out of 15; three models improved. With text alone, one model always allowed commands and another always denied them. Their overall (raw) accuracies differed by 32 percentage points; both scored 50% balanced accuracy, which weights safe and unsafe cases equally, and no correct pairs. A rule evaluator matched all 15 pairs by applying the same rules that supplied the expected answers; this is not an independent safety test. The audit complements systems that check actions against safety rules: it checks whether the model receives the information needed for those decisions. This helps evaluators separate missing-input problems from model errors. These small-sample scores do not establish general model reliability.

0.44

UtilityBench: Measuring Language Model Utility When Errors Have Different Costs

cs.LG cs.AI ·2026-10-19 ·Qifeng Lei, Zhiyong Yang, Zitai Wang, Shilong Bao, Linchao Zhang +2 more
benchmark similarity 0.60
Abstract

As language models are increasingly deployed in real-world scenarios with important outcomes, the impact of their mistakes often extends far beyond what correctness can capture. In domains like medicine, a single error can lead to serious consequences and put patient safety at risk. However, mainstream evaluation protocols still treat all errors as equal, overlooking the spectrum of real-world costs. To address this, we propose a utility-based evaluation paradigm that explicitly incorporates the varying consequences of errors. Guided by this principle, we develop UtilityBench, a large-scale benchmark of 15,000 instances with verifiable answers, quantifying error severity using Best-Worst Scaling to elicit meaningful, data-driven cost estimates. These quantified utilities shape when a model should answer or abstain, guiding it to make decisions that align with the true importance of each scenario. As another merit, we introduce UtilityAUC, a metric that measures the area under the utility curve determined by a model's response ordering, thereby enabling utility preference alignment beyond error frequency. Our theoretical analysis demonstrates that UtilityAUC can distinguish between models where standard accuracy and AURC fall short. Empirically, evaluating 20 state-of-the-art models, we find that greater accuracy or more advanced architectures do not reliably translate to higher UtilityAUC, highlighting a gap between problem-solving ability and responsible decision-making in language model evaluation.

0.27