LLM Evaluation

Large language models (LLMs) have demonstrated remarkable capabilities in reasoning and code generation, raising the prospect that they could assist in developing and optimizing the very infrastructure that powers them. However, existing benchmarks mainly focus on isolated kernels, predefined operators, or pre-specified optimization targets, and therefore fail to evaluate the ability of LLMs to perform open-ended, long-horizon LLM infrastructure engineering. To address this gap, we present $Φ$-Bench, a benchmark for systematically evaluating LLMs on engineering the LLM infrastructure stack. Derived from optimization problems studied in frontier research and grounded in real-world code repositories, $Φ$-Bench provides broad coverage of the LLM infrastructure stack and spans tasks of varying complexity, ranging from localized kernel-level function completion to long-horizon implementation and end-to-end system optimization. Extensive experiments on frontier LLMs reveal their current capabilities and limitations in engineering complex LLM infrastructure, offering insights into the challenges that remain on the path toward autonomous optimization of future AI infrastructure.

Large language models (LLMs) increasingly act as research agents, yet their ability to track shifts in research attention is difficult to evaluate because reviews and research ideas lack uniquely verifiable outcomes. We introduce Research Attention Prediction (RAP), a rolling benchmark covering 278 AI/ML fields and 1,390 episodes. At each cut-off, an LLM agent searches a temporally restricted arXiv corpus and predicts the next six months' paper shares across eight frozen research directions. Search generally helps, but all four diagnostic models perform worse than an exact-count exponentially weighted moving average (EWMA) baseline in compositional accuracy. We identify two linked bottlenecks. Under cumulative-history access, State carry-forward outperforms direct Forecast for all four diagnostic models; frozen-evidence replay links a shared component of this reversal to Forecast-oriented policies retrieving a smaller share of recent evidence. Even with exact historical activity, future-specific updating remains limited, with only GPT-5.5 plus reopened Search slightly surpassing EWMA. Fine-tuning on realised outcomes improves Qwen3-4B's forecast Spearman correlation by 0.105 on held-out fields at later origins, with gains also on change-rich episodes.

We introduce MetroLLM-Bench, a 955-case benchmark for testing language models as the policy layer of a transit kiosk. It covers six real metro systems, ranging from 37 to 414 stations, and eleven categories that include routing, fare calculation, disruptions, accessibility, and adversarial input. In each case, the model must call structured tools and submit a machine-renderable terminal state containing an outcome, a per-ticket fare quote when applicable, and a kiosk action. Fourteen deterministic scoring components form Tier 1; eight semantic-quality components form Tier 2, six of which use a language-model judge. We report Tier 1 and the combined score of both tiers. A stratified 75/25 split reserves 717 cases for training-data generation and 238 for held-out evaluation. We evaluate twenty-six models from six vendors, of which twenty-three are ranked. On the held-out partition, a 4B Qwen 3.5 student trained through parameter-efficient fine-tuning (PEFT) exceeds both GPT-5.6 tiers on Tier 1 (91.3 against 90.6 and 90.0) and matches GPT-5.4 full at maximum reasoning effort (91.4), with a 2.6 GB Q4_K_M footprint. Larger 9B and 27B students provide no further Tier 1 improvement over the 4B student at this training scale. Across the four Qwen sizes, the PEFT gain over the corresponding base model decreases from +7.03 points at 2B (three training seeds) to -0.91 at 27B; every seed shows the same direction at every size. A deterministic rule-based baseline reaches 84.6 on Tier 1, with the remaining language-model advantage concentrated in policy adaptation, compound scenarios, accessibility, and temporal reasoning. Muse Glimmer 30B leads the composite ranking, and serving configuration alone moves the Qwen 3.5-to-3.8 comparison by 2.7 Tier 1 points. The benchmark, harness, reproduction guide, and fine-tuned students are released at https://github.com/continker/metrollm-bench.

Large Language Models (LLMs) have shown promise for translating Natural Language (NL) planning descriptions into PDDL problem instances. However, standard evaluation criteria such as syntactic validity or planner success can substantially overestimate faithfulness to the described task: a generated problem may be parseable and solvable while misrepresenting the intended initial state, goal, object structure, or optimization target. This paper studies an end-to-end NL-to-PDDL pipeline that combines LLM generation, checks in terms of PDDL parsing, planning and validation, a domain-conformance checker, an LLM critic, and iterative repair. Fine-grained repair feedback is constructed from the domain description, the generated problem, the natural language problem description, and operational diagnostics. Reference-based comparisons against curated benchmark PDDL problem descriptions are used for post-hoc benchmark analysis, and these offline checks include renaming-invariant structural matching and semantic equivalence, where domain support is available. Across Planetarium, AutoPlanBench, and curated PDDL2.1 problems, results show that operational success and benchmark-reference reconstruction can diverge substantially. Results also show that structured repair can be useful, and that PDDL2.1 remains challenging for reference reconstruction, even when operational success improves.

Language models fine-tuned on customer behavior can predict outcomes and generate explanations, but these readouts are often treated as interchangeable. Holding model checkpoint and prompt content fixed, we compare probabilities obtained by scoring answer tokens with predictions generated after a written rationale. Across 13 model-domain cells covering four retail tasks in three markets, including two using fully public data and checkpoints, the scored readout ranks outcomes more accurately in 12 of 13 cells (two-sided sign test, p approximately 0.003), by 1.5 to 14.5 points in area under the receiver operating characteristic curve (AUC). Paired bootstrap confidence intervals exclude zero in every newly measured cell. The gap varies with task-specific supervision and mismatch between training and serving formats, ranging from -2.2 points for an untuned base model to +13.7 for rationale-format supervision. Analysis of approximately 9,000 rationales identifies two correlates: reduced reliance on the dominant predictive feature and convergence on stock formulations. Probability saturation does not track the gap. A third readout, eliciting a probability before any verdict, improves calibration (Brier score from 0.47 to 0.15) while ranking within noise of scoring, but only for outcome rates represented in training; it is worse than scoring when the scored head is already calibrated. We interpret these differences through the objectives matched by each readout, identify training choices that narrow the gap, and propose retaining generated rationales while sourcing ranking from the scored head.

Although MMLU is widely adopted as a benchmark for calibrating general AI capabilities, we psychometrically demonstrate that its aggregate score primarily evaluates a model's factual retrieval capacity rather than its reasoning ability. By calibrating item difficulty for 1,000 open-weights language models over 14,042 MMLU test items using Item Response Theory, we show that evaluating both abilities via a single test is inherently flawed. Difficulty is then regressed on a deterministic, text-extractable framework of structural complexity. Applying a joint Wald test with subject-clustered covariances demonstrates that the MMLU conflates fundamentally separable constructs. The mapping from structural complexity to difficulty is not invariant across the benchmark's STEM and non-STEM partitions. This finding has practical consequences. Aggregate leaderboard ranks track non-STEM accuracy more closely than STEM accuracy, so selecting a Top-50 model on the aggregate for a reasoning-intensive deployment displaces roughly 22% of the STEM-appropriate choices. Furthermore, when controlling for the multiple-choice guessing floor natively inside the response model, we find that higher-ability models continue to degrade more steeply under increased reasoning depth. The MMLU aggregate therefore weights retrieval capacity and reasoning stability unequally, inadvertently favoring models optimized for retrieval. We release our deterministic framework as a reproducible auditing instrument and recommend disaggregated reporting.

Large language models (LLMs) are prone to hallucinating or misinterpreting facts, which impairs their usability in retrieval-augmented generation or data-to-text systems. We analyse how faithfulness of LLMs to provided context depends on how plausible they perceive the context to be (context-memory conflict). To better identify error patterns, we make use of the increased difficulty of non-English and low-resource language text generation and input data based on local knowledge, only partially captured in models' parametric knowledge. We let the models generate text in English, Czech, Slovak and Upper Sorbian from factual (FA), counterfactual (CFA) and fictional (FI) RDF triples containing local Czech and Slovak data. Contrary to our expectations, we observe only a weak context-memory conflict on the human-annotated sample. For Kimi K3 as an LLM judge, which agrees well with human annotations on the sample, counterfactual inputs receive only slightly lower faithfulness scores than factual ones (-0.05 on a 1-5 scale). We also find that a suboptimal choice of LLM judge would lead to overestimating the strength of the context-memory conflict.

Large language models (LLMs) may abandon correct positions when users push back, exhibiting a failure mode known as sycophancy. Existing evaluations typically use short, pre-specified conversations and may therefore miss failures that emerge under sustained, adaptive disagreement. We introduce SPINE, a benchmark in which an LLM proxy plays a persistent but mistaken user and adaptively challenges a target model for up to 25 turns. We evaluate four production systems and three Olmo3-7b variants on 100 false-presupposition and 100 unethical-query items. Our experimental results show that collapse rates increase with conversation length for every model, short-horizon protocols underestimate sycophancy and resistance under sustained pressure remains unreliable across current models. By analyzing models with accessible reasoning traces, we surprisingly found that the correct position often remains represented in a reasoning trace when the response concedes, suggesting that the model chooses to please a user and sycophancy is not due to lack of knowledge or ignorance. Ablations show that adaptive LLM proxy exposes more sycophantic collapse than pre-generated scripts. Among all tactics, emotional appeals is the most associated with inducing LLM sycophantic behavior. The code and data are released at https://anonymous.4open.science/r/SPINE

Reasoning representations are increasingly used as explanations for large language model outputs. Yet they are typically evaluated with model-centric criteria, such as answer accuracy and faithfulness, leaving it unclear whether they help people evaluate model responses. In this work, we study reasoning representations as human-facing interfaces rather than proxies for model reasoning ability. We conduct a controlled human study of six reasoning formats across tasks of varying complexity, supported by a web-based framework that randomizes task domains, problem instances, and representation order. The study collects fine-grained judgments of structural understanding, error detection and localization, and trust calibration. Our study shows a mismatch between perceived preference and support for human evaluation. Participants prefer planning- and decomposition-based representations, but simpler chain-of-thought traces better support verification, trust, and interpretability. Preferred representations also introduce calibration risks, with more false alarms on correct traces and high trust despite low willingness to verify.

Large Language Models (LLMs) demonstrate impressive performance across diverse NLP tasks, yet their ability to exhibit genuine contextual understanding remains uncertain. Traditional evaluation metrics such as perplexity, BiLingual Evaluation Understudy (BLEU), or surface-level accuracy fail to reveal how well LLMs extract, integrate, and reason over contextual information--a gap particularly critical in question answering, where models must align responses with contextually grounded knowledge rather than memorized associations. We propose a novel knowledge graph-based evaluation framework introducing Semantic Structural Similarity for KGs (S3KG), a hybrid similarity measure integrating structural and semantic similarity into a continuous evaluation score, alongside a diagnostic framework for categorizing reasoning errors. To validate this pipeline, we evaluate S3KG against established metrics on a curated question-answer (QA) benchmark, demonstrating its effectiveness in measuring correctness, faithfulness, and interpretability in LLM-generated responses.

Automatic fact-checking systems assess the veracity of claims given evidence from relevant documents. Large Language Models (LLMs) have demonstrated strong performance in fact-checking due to their general reasoning capabilities. However, it remains unclear whether they faithfully make use of the evidence provided to reach veracity judgments or rely on parametric knowledge. To investigate this, we introduce Fact-Ablated Evaluation (FAE), a new evaluation framework that iteratively ablates the cited evidence to assess whether LLMs revise their predictions accordingly. Our empirical results show that current off-the-shelf LLMs as fact-checking systems rely more on their parametric knowledge than on the evidence provided. To bridge this gap between prediction accuracy and evidence grounding, we propose REAL (Rigorous Evidence Ablation Learning), a training framework that promotes evidence-dependent verification through counterfactual evidence supervision for the LLM-as-verifier models. Experiments on four fact-checking datasets across different domains demonstrate that models trained with REAL obtain superior evidence-dependent capabilities compared to standard fine-tuned models. Our findings highlight that strong fact-checking performance can still coexist with weak evidence dependency, while REAL encourages veracity predictions to remain more closely tied to the availability of supporting evidence.

Benchmark scores are a central currency in model releases: they inform purchasing decisions, shape public trust, and influence policy. Yet, a key assumption underlying benchmark scores is that the model performance measured through APIs faithfully reflects the behavior of deployed systems. We challenge this assumption by auditing ChatGPT, Claude, and Gemini across seven systems and nine benchmarks spanning general capability, social bias, and sycophancy. We find systematic API--interface differences in both accuracy and consistency. On average, API evaluations score 3.4 percentage points higher in accuracy and 2.1 percentage points higher in test--retest agreement than corresponding interface evaluations. For ChatGPT, the performance difference between API and interface access exceeds the API-only difference between GPT 5.3 and GPT 5.4. Put differently, switching access surfaces can degrade performance as much as downgrading a full model generation. We further test whether exposed API controls can reproduce interface behavior by varying system prompts, sampling parameters, and reasoning settings. These controls shift behavior in some cases but do not reliably eliminate the gap. Our findings document a context-validity gap: measurements obtained through APIs do not necessarily generalize to corresponding deployed interfaces, complicating the use of API evaluations as proxies for deployed systems.

Large language model (LLM) benchmarks are often treated as fixed datasets with stable scores, yet their outcomes depend on configurable evaluation pipelines. We audit eight cybersecurity benchmarks across 10 proprietary, open-weight, and cybersecurity-specialized LLMs. By modeling benchmarks as measurement pipelines, we identify 15 systematic failure modes and show that a single pipeline choice can change a model's score by more than 80 percentage points and substantially alter model rankings. At the cross-benchmark level, two semantically similar task pairs rank the same models differently because of incompatible evaluation conventions. Under an evaluation harness that standardizes pipeline choices while preserving task semantics, nine of 10 models shift by at least three ranks on at least one benchmark. These results show that cybersecurity LLM benchmark scores are pipeline-dependent and motivate pipeline-aware auditing as a core requirement for reliable model evaluation.

Large language models (LLMs) are increasingly used as backends for intelligent web services, but serving them across the edge continuum requires balancing quality, latency, model footprint, and energy. This paper presents a controlled measurement study of self-hosted LLM inference across edge and near-edge deployment nodes: an NVIDIA Jetson AGX Orin and a near-edge server with CPU-only and GPU-enabled inference modes. We evaluate multiple open-weight LLMs and quantization variants using a fixed question-answering workload, and compare them against GPT-4o as a cloud-hosted accuracy and latency reference. Our benchmarking pipeline reports accuracy, model footprint, per-token decoding latency, prefill latency, and overall execution energy. The results show that GPU-enabled server execution provides the lowest compute-side latency, while Jetson Orin shows lower measured energy, consistent with its lower platform power under our setup. CPU-only execution is consistently dominated in latency for our workload and shows higher measured energy. We also show that parameter count and downloaded weight-file size alone do not reliably predict observed accuracy or latency. Finally, using Pareto-frontier analysis, we study how deployment decisions may change under possible streamed-token delivery overheads, highlighting that compute-side inference metrics alone can lead to suboptimal placement for latency-sensitive interactive web services.

Existing causal-inference benchmarks for LLMs mostly score method descriptions or whether generated code runs, not whether the executed workflow recovers the target causal estimate. CausalVerify studies this verification problem for structured econometric causal-estimation workflows by separating realistic interpretation from verifiable computation. It pairs 259 published economics papers (reconstructed research question, data description, institutional context) with 100 fixed-seed synthetic scenarios that realise CSV datasets for difference-in-differences, event study, instrumental variables, and regression discontinuity designs. Experiment A (real-paper text agreement) scores method-family and direction agreement against four-LLM consensus labels. Experiment B (synthetic execution) runs model-written R code and checks whether the extracted treatment-effect estimate matches a canonical estimator on the same realised dataset; this execution-grounded correctness layer is L2b+, distinct from L2b, which records only whether the code executes. A calibration arm asks whether self-reported confidence separates correct from incorrect workflows. On Experiment B, seven LLMs reach L2b+ pass rates of 10% to 88% at the default 50% tolerance, and 66 of the 426 workflows that execute (15.5%) return a wrong estimate. Execution ranking (L2b) agrees with L2b+ far better than text-direction scoring (L4): Kendall $τ=0.81$ and Spearman $ρ=0.93$, versus Kendall $τ$ between $-0.20$ and $0.10$ for L4. Llama-3.3-70B-Instruct shows the same qualitative gap, and reported confidence does not reliably separate correct from incorrect workflows. The claims are confined to standardized single-shot workflows in these four design families under the evaluated R backend and model panel; the benchmark does not measure general causal-inference ability. Code, data, cached outputs, and a datasheet are released.

Large Language Models (LLMs) are frequently confident, eloquent, and well versed. A natural question arises: do they know what they don't know? To answer this question, we borrow the concept of epistemic honesty and develop a novel metric to systematically evaluate whether an LLM appropriately acknowledges the boundaries of its knowledge. In this work, we introduce the Epistemic Honesty Quotient (EHQ), which reports three observable sub-scores across two operational axes (epistemic restraint and substantive-answer calibration), and construct EHQ-3000, a 3,000-question benchmark spanning Fabricated Entity, Post-Cutoff Event, Hyper-Niche True, and Context-Conditioned Questions. From a frozen registry of 21 model API routes, 15 completed the protocol after endpoint and eligibility checks; 14 entered the confirmatory analysis because severe provider-side truncation made one route's score indeterminate. The study reveals substantial variation across models, including a difference that can not be explained by their capability to extract explicitly available information. Composite EHQ ranges from 0.31 to 0.81 across the analysed panel, despite near-ceiling performance on the document-grounded capability probe. The two restraint criteria overlap strongly under the present category composition, whereas substantive-answer calibration varies across models and does not reliably co-vary with restraint; however, the small panel leaves substantial uncertainty. Thus, EHQ reveals behavioral differences that are not visible to conventional correctness-based assessment, while also showing why dataset composition, provider behavior, and confidence elicitation must remain part of the interpretation.

Large language models are increasingly used as sources of advice and information, including in high-stakes settings, yet little is known about how they respond to user disagreement. We study how a model manages its epistemic authority, referring here to its claim to knowledge, competence, or the right to advise, once a user challenges its answer. Building on Conversation Analysis, we introduce a taxonomy of six challenge types and a four-layer framework for analysing each response: whether the original claim is maintained or changed, where authority is located, how the disagreement is socially managed, and what kind of evidential support is offered. We construct a new dataset of 2,310 controlled challenge scenarios and 32,340 corresponding responses from 14 models, and analyse them using our framework with an LLM-as-judge pipeline, providing a vocabulary which future evaluation and benchmark design can build on. We find that models show conflicting behaviour: they validate users in 85% of responses but maintain their original claim in 65%. They explicitly apologise in 33% of responses, yet 59% of those apologies accompany maintenance of the original claim. They transfer authority most often in advice tasks, doing so in 28% of responses and reaching 57% in health advice and 49% in legal advice, compared with 6% in fact and 3% in explanation tasks. Abandonment of the original claim ranges from 0.8% for GPT-5.2 to 40% for DeepSeek 7B, while complete replacement of the original claim is rare overall at 1.5%.

The application of large language models (LLMs) to personalized medical assistants has garnered growing interest. However, existing medical benchmarks largely rely on static question answering with pre-selected evidence, leaving unclear whether LLMs can make reliable clinical decisions from real longitudinal electronic health records (EHRs). To bridge this gap, we introduce ObGynLongBench, a rule-grounded long-context EHR benchmark for obstetric and gynecologic decision-making, comprising 1,500 clinical decision-point cases from 976 real pregnancy EHR histories and traceable rules. Each case is anchored to a patient, a pregnancy-timeline point, and a pre-decision information boundary, enabling Evidence-only, Visit-level EHR, and History-level EHR evaluation. Evaluating 17 LLMs reveals a substantial Evidence-to-EHR Gap: models perform well when evidence is directly provided, but accuracy drops when evidence must be extracted from same-day records or full pre-decision EHR histories. Further analyses identify evidence utilization as a key bottleneck: performance decreases with longer EHR contexts and more complex evidence requirements, and earlier failures often predict later failures within the same patient history. Finally, active-search agents perform best among EHR access strategies, highlighting patient-specific evidence utilization as a central challenge for reliable personalized medical assistants. Resources are available at https://github.com/xiangjun2003/ObgynLongbench.

We introduce FramingQA, a benchmark that measures the model sensitivity to question framing across law, medicine, finance, and robotic simulations. Large language models (LLMs) often change their responses to subtle rephrasings that align with an implied stance by users. This can leave users with advice tainted by how they happened to phrase a question rather than by the underlying facts, and the consequences are highly costly in high-stakes domains. Because in the realistic scenarios, both expert practitioners and non-expert users frequently ask LLMs questions containing incomplete or misleading assumptions, models are highly susceptible to those framings. To test this, we inject the framing bias across three nested levels: a framing-biased question phrasing (root), an injected framing-biased premise prepended to a neutral question (propositional), and a premise paired with a framing-biased question (global). Evaluating nine open models (3.8B-70B) across four families, we find that strong per-variant accuracy does not guarantee the robustness across differently phrased questions under the fixed factual information.

Large language models (LLMs) are used to simulate human participants in psychological research. We asked whether LLMs that reproduce human evaluations of a whistleblower's moral character also reproduce the motive attributions that accompany them. Five LLMs and two human samples (N = 125 and N = 742) evaluated a physician who either remained silent about fraudulent billing or reported it to a hospital, regulator, or newspaper. Models reproduced the human ranking of the physician's moral character but portrayed whistleblowers as more helpful, less self-interested, and less hostile. In four of five models, competitive motives were less strongly associated with moral-character judgments. Model ratings changed little when prompts reproduced the narratives and demographic profiles of both human samples, although this comparison cannot isolate a perspective effect. Thus, agreement in average ratings can conceal differences in attributed motives, relationships among judgments, and sensitivity to context. Validating LLMs as simulated participants therefore requires testing psychologically informative response patterns, not average agreement alone.

Robustness evaluation of large language models (LLMs) remains a critical challenge, particularly in assessing their sensitivity to perturbations in input data. In this work, we systematically evaluate LLM robustness across multiple dimensions, including word error rate, character repetition and duplication, modifications in choices, and variability in instruction following. To facilitate this evaluation, we construct a synthetic and augmented dataset encompassing a diverse set of LLM benchmarks, specifically targeting multiple-choice question (MCQ) datasets and instruction-following tasks. We conduct extensive experiments on LLMs of varying scales-small, medium, and large-as well as across base and instruction-tuned variants. Our analysis quantifies the variability in model responses under perturbed conditions and highlights discrepancies relative to baseline models. The findings provide insights into the stability of LLMs across different evaluation scenarios contributing to the development of more robust and reliable language models as well as robust evaluation methodologies.

The rapid digital transformation of manufacturing, often referred to as Industry 4.0, relies on seamless interoperability between physical and software assets. A central enabler is the Asset Administration Shell (AAS), a standardized digital representation of such assets. Recent advances in large language models (LLMs) enable the generation of AAS submodels from unstructured sources such as product datasheets but raise challenges for quality assurance. In particular, unexpected errors, the lack of ground truth references, and the absence of standardized quality metrics hinder reliable adoption. In this work, we evaluate quality metrics for AI-generated AAS using a perturbation-based evaluation framework. By systematically degrading AAS generation along multiple dimensions, we assess how well different metrics reflect quality changes. Based on a dataset of 200 products from multiple manufacturers, we generate 6,400 AAS instances using GPT-4o-mini, Qwen3, and DeepSeek-R1. Our results show that metrics based on exact matching of property names and similarity-based soft matching of property values, in particular value-based recall and name-based F1 score, provide the most reliable indicators of quality degradation. Furthermore, we quantify the impact of different perturbation types and analyze differences across model families and product segments. These findings support the selection of suitable metrics, the tuning of LLM-based pipelines, and the integration of AI-generated AAS into industrial applications.

Minimal-pair benchmarks such as BLiMP evaluate linguistic knowledge by testing whether language models (LMs) prefer acceptable sentences over minimally different unacceptable ones. However, these benchmarks largely ignore lexical frequency variation, despite lexical frequency being a pervasive and highly skewed property of natural language use. Consequently, existing evaluations do not test whether grammatical preferences remain stable when contrasts involve rare lexical items. We introduce FreqBLiMP, a frequency-controlled extension of BLiMP that regenerates all 67 paradigms under explicit Zipf-frequency regimes while preserving each minimal-pair's grammatical contrast. Evaluating multiple open-weight LLM families across scales, we find that decreasing lexical frequency produces a consistent, monotonic decrease in sentence likelihood, but only a modest reduction in overall contrastive acceptability accuracy. However, this aggregate stability masks substantial variation across linguistic phenomena, with LLMs remaining robust on overt morphosyntactic generalization while degrading on phenomena that require lemma-specific information.

Deeply phenotyped cohorts combine clinical, imaging, molecular, and wearable observations across timescales from seconds to years, but heterogeneous analyses are not directly comparable. We present PhenoBench, an executable benchmark that turns deep-phenotyping measurements into explicit questions and controlled comparisons of information sources and predictive models. It is built around the Human Phenotype Project, with more than 13,000 participants at the initial visit. Each question fixes the target, population, timing, and allowed information; its evaluation contract specifies the split, metric, baseline, and claim boundary. PhenoBench defines 90 clinically grounded tasks across 15 domains and 26 input modalities. Across 160 matched regression comparisons spanning 52 tasks, six pretrained tabular models ranked above the evaluated task-specific baselines, including XGBoost and CatBoost, under a fixed single-estimator protocol with bounded tuning. Giving each task equal weight, their mean advantage over ridge was 0.0103 $R^2$ (95% task-bootstrap interval, 0.0071-0.0136). We also evaluated 14 language models, collectively covering 40 tasks spanning phenotype recovery, classification, follow-up forecasting, and participant ordering. Without cohort-specific fitting, language models made informative predictions on some tasks but showed task-specific capability gaps, shared failures of scale, and rarely surpassed task-specific ridge or logistic regression models fitted on the same input fields. PhenoBench provides a versioned, auditable evaluation system where new questions, measurements, and models can be added without redefining existing comparisons.

Large language models (LLMs) are increasingly employed to predict molecular properties. However, prediction error alone cannot distinguish prediction from retrieval of published values. We audit 22 frontier models on 12 molecular regression datasets in a zero-shot setting, assessed against a molecule-blind reference derived from each dataset's labels. Significant retrieval is concentrated on 5 datasets, with isolated flagged LLMs elsewhere. Increasing the reasoning setting raises the number of flagged model--dataset combinations from 47 to 89 of 264. An in-context blinding experiment reduces retrieval but leaves a quarter of the combinations flagged. Blinding changes model rankings and increases errors. Because blinding also removes chemically interpretable structure, the error increase can only be partially attributed to reduced retrieval.

Assessing the impacts of social policy changes is a widely acknowledged challenge for policymakers. Econometric methods can be unreliable when extrapolating to hypothetical scenarios, while field pilot programs are highly costly. In this paper, we propose using large language models (LLMs) as policy-assessment tools adapted from general-purpose models. We present FlexPension-LLM, the first domain-specialized large language model for a hierarchical pension-enrollment prediction task among flexible workers in China, and introduce DKI-RDistill, which injects policy-grounded cues into the prompt, including Probit-derived marginal effects and hukou-province pension rules. The method then uses LoRA/SFT to distill rationale-augmented supervision into an open-weight MoE student, with teacher errors corrected by regenerating those cases under ground-truth labels. On a CHFS 2019 blind split, FlexPension-LLM achieves 0.9316 Composite F1, surpassing its Claude Sonnet 4.5 teacher and 15 of 17 baselines, and is statistically indistinguishable from Claude Opus 4.6. Across four external surveys, it averages 0.7549 Composite F1 and shows the narrowest performance range among the strongest systems. Component analysis shows that gains come mainly from policy-grounded cue injection and error-filtered supervision, while rationales provide decision traces that can be checked against policy rules.

Scientific papers require models to integrate evidence across text, equations, figures, tables, code, and datasets while preserving its provenance. Beyond answer correctness, scientific reading requires verifiable outputs from operations such as evidence localization, definition extraction, and consistency checking. We introduce SciDocBench, a workflow-centered benchmark targeting these operations through 124 expert-authored and difficulty-screened questions across seven research-assistant capability groups, 19 subtasks, and five scientific domains. Each question is instantiated in four matched settings formed by pairing its bilingual variants with the All Images First and Markdown Interleaved document representations, yielding 496 evaluation instances. The strongest evaluated model, Claude-Opus-5, scores 62.6 out of 100, with remaining gaps in evidence localization, structured information extraction, cross-document synthesis, and robustness to document representation. To convert these diagnostics into scalable training signals, we introduce SciDocIR, a structured representation of scientific document objects, layout and cross-reference relations, and provenance. Using SciDocIR, we construct SciDocDataset, which contains 4K supervised fine-tuning instances and 10K reinforcement-learning instances, built on 14 verifiable training subtasks. Post-training Qwen3.6-27B on task-aligned data improves its SciDocBench score from 40.03 to 45.33 with supervised fine-tuning and to 45.74 with subsequent reinforcement learning. Both adapted models preserve DocVQA and InfoVQA performance and improve ChartQA accuracy over the original model by 0.80 and 3.40 points, respectively. Together, SciDocBench, SciDocIR, and SciDocDataset connect capability diagnosis with verifiable training-data construction for scientific-document assistants.

Performance modeling is central to hardware design and software optimization, yet constructing these models requires structured reasoning about computation, data reuse, storage, and movement. We introduce PerfReasoning, a benchmark that evaluates LLMs both as direct performance reasoners and as generators of analytical performance-model code. Given workload, architecture, and mapping specifications, models compare mappings and predict off-chip traffic and buffer requirements. The strongest closed-source models exceed 90% on reasoning-based Q&A, and the best open-weight model reaches 82.4%. However, model construction is substantially harder: while GPT-5.6 Sol exceeds 80% pass rate, all other model configurations average below 45% and vary markedly across runs. Task-specific RL raises a 4B model's mapping-reasoning accuracy by 15.7 points, whereas feedback-free multi-round self-revision prompting is not reliably effective. PerfReasoning exposes the gap between plausible architectural reasoning and reliable performance-model construction. We will publicly release the benchmark to support reproducible evaluation and track future progress.

Language-model judges now gate training data, score generations, and drive leaderboards. The judge is then a measurement instrument, resting on one rarely stated assumption: the same request, sent to the same model name, reads the same tomorrow. We audited that assumption in two preregistered campaigns with every threshold fixed in advance; neither got past validating its instrument. Across 52,988 audited request attempts, same-window repeat rankings agreed at Spearman 0.400 against a required 0.90, and byte-identical next-day replays agreed at 0.78 against a required 0.99, each time with the execution record at ceiling. Three mechanisms explain the gap: a label-to-meaning mapping that biased readouts as strongly as the signal; candidate gaps seven orders of magnitude below the instrument's own noise floor; and byte-identical inputs returning different rankings, a noise that exact-permutation readouts compound. Neither metric substitution nor sampling repaired it on the tested grid. Preregistered follow-ups bound the problem: waiting did not help on the days sampled (0.805 versus 0.800, replicated over five further days); switching providers did not help (four providers share the floor, medians 0.74 to 0.88, predicted by none of the metadata fields they expose); self-hosting on batch-invariant kernels helped only while the server was quiet; and on constructed errors with known gaps, the readout's separation tracks error type, not size. We distill the evidence into a three-level snapshot-identity ladder, eight design rules, and a reporting checklist; a pilot at roughly 2% of the study's call volume would have exposed both unreachable gates in advance. All results concern externally measured behaviour on shared serving infrastructure. On a shared endpoint, a model name is not a frozen instrument; a preregistered evaluation must measure its instrument before freezing any gate on it.

Several studies have evaluated the ability of Large Language Models (LLMs) for meal planning, yielding positive outcomes. These models can process natural language inputs and leverage learned knowledge from their pretraining to generate meal plans. In this work, we investigate the ability of LLMs to analyze the suitability of given recipes for diabetes. The primary challenge for LLMs is to retrieve relevant dietary guidelines for diabetes, decompose recipes into ingredients and cooking methods, and apply these guidelines to determine the recipe's suitability. To study these challenges, we employ three kinds of prompts namely, (i) Direct Query Prompt (ii) Context-Guided Prompt, and (iii) Exemplary Context Prompt that incorporate different levels of diabetes dietary guidelines from medical sources. We introduce a benchmark dataset curated for this investigation consisting of 7607 recipes that include 3807 recipes suitable for diabetes and 3800 recipes not suitable for diabetes. Our results demonstrate that most LLMs are cautious in predicting recipes as suitable to prevent detrimental outcomes. Further, the models that can reason using the dietary guidelines performed better in predicting the suitability of recipes for diabetes. Overall, Mistral-7B and Llama 70B showed superior performance to their counterparts.