arXiv subjects

Asad Aali

Publications and source records attributed to Asad Aali.

At least 19 recordsLinked to original sources

arXiv2026

Benchmarking the Benchmarks: A Validity Audit of Tool-Calling Evaluation

Tool-calling benchmarks are increasingly used to rank language-model agents, yet their scores are often treated as ground truth without validating the evaluators themselves. We present a systematic validity and reproducibility audit of four major tool-calling benchmark families: BFCL v4, {\tau}2-Bench, LiveMCPBench, and MCP-Atlas. Across 496 expert-reviewed benchmark tasks, we find 92 evaluator-human disagreements, corresponding to an 18.5% misalignment rate. The failures are not isolated annotation mistakes: deterministic benchmarks exhibit brittle state matching, trajectory lock-in, incorrect ground truths, substring-based communication failures, and reward-basis misalignment, while LLM-judge benchmarks exhibit rubric drift, hallucinated completion, answer-only scoring, and substantial run-to-run variance. In LiveMCPBench, 23 repeated evaluations of the same setup produce scores ranging from 57.9% to 76.8%, a spread of 18.9 percentage points, large enough to change leaderboard conclusions. These results show that current tool-calling scores can reflect evaluator artifacts rather than agent capability. We introduce a unified taxonomy of tool-calling evaluation failures, release trace-level audit artifacts and corrected evaluation components, and argue for decomposed metrics that separately measure tool invocation, task completion, and outcome verification. Our findings suggest that progress in tool-using agents requires benchmarks whose evaluators are themselves reproducible, auditable, and aligned with human judgments of task success. We further introduce Tool-Veritas, a configurable benchmark that combines deterministic state verification with optional qualitative judging, and Harness Lab, an open-source system for benchmark execution, trace inspection, repeated-run comparison, and evaluator debugging.

cs.SE↗

arXiv2026

MAVEN: Improving Generalization in Agentic Tool Calling

Generalization across agentic tool-calling environments remains a central challenge for reliable agentic reasoning systems. Although large language models achieve strong results on individual benchmarks, their ability to compose reasoning strategies, preserve intermediate states, and coordinate tools across domains remains underexplored. We present MAVEN (Modular Agentic Verification and Execution Network), a lightweight symbolic reasoning scaffold for structured decomposition, adaptive tool orchestration, and intermediate verification. We evaluate MAVEN across established tool-calling benchmarks, including BFCL v3, TauBench, Tau2Bench, AceBench, and introduce MAVEN-Bench, a stress-test benchmark for multi-step mathematical and physical reasoning with explicit verification and adversarial task composition. MAVEN-Bench exposes a substantial gap between partial reasoning quality and end-to-end task success; in direct MAVEN-Bench runs, MAVEN improves its GPT-OSS-120b base model from 48% to 71% accuracy without additional training. It also remains competitive with frontier proprietary baselines while using an open-weight backbone with an estimated cost ratio of roughly 1/10, suggesting that lightweight verification-centered scaffolds can strengthen compositional reasoning and motivate more process-aware evaluation of agents in the wild.

cs.AI↗

arXiv2026

$S^3$: Stratified Scaling Search for Test-Time in Diffusion Language Models

Test-time scaling investigates whether a fixed diffusion language model (DLM) can generate better outputs when given more inference compute, without additional training. However, naive best-of-$K$ sampling is fundamentally limited because it repeatedly draws from the same base diffusion distribution, whose high-probability regions are often misaligned with high-quality outputs. We propose $S^3$ (Stratified Scaling Search), a classical verifier-guided search method that improves generation by reallocating compute during the denoising process rather than only at the final output stage. At each denoising step, $S^3$ expands multiple candidate trajectories, evaluates them with a lightweight reference-free verifier, and selectively resamples promising candidates while preserving diversity within the search frontier. This procedure effectively approximates a reward-tilted sampling distribution that favors higher-quality outputs while remaining anchored to the model prior. Experiments with LLaDA-8B-Instruct on MATH-500, GSM8K, ARC-Challenge, and TruthfulQA demonstrate that $S^3$ consistently improves performance across benchmarks, achieving the largest gains on mathematical reasoning tasks while leaving the underlying model and decoding schedule unchanged. These results show that classical search over denoising trajectories provides a practical mechanism for test-time scaling in DLMs.

cs.LG↗

arXiv2026

Compositional Neuro-Symbolic Reasoning

We study structured abstraction-based reasoning for the Abstraction and Reasoning Corpus (ARC) and compare its generalization to test-time approaches. Purely neural architectures lack reliable combinatorial generalization, while strictly symbolic systems struggle with perceptual grounding. We therefore propose a neuro-symbolic architecture that extracts object-level structure from grids, uses neural priors to propose candidate transformations from a fixed domain-specific language (DSL) of atomic patterns, and filters hypotheses using cross-example consistency. Instantiated as a compositional reasoning framework based on unit patterns inspired by human visual abstraction, the system augments large language models (LLMs) with object representations and transformation proposals. On ARC-AGI-2, it improves base LLM performance from 16% to 24.4% on the public evaluation set, and to 30.8% when combined with ARC Lang Solver via a meta-classifier. These results demonstrate that separating perception, neural-guided transformation proposal, and symbolic consistency filtering improves generalization without task-specific finetuning or reinforcement learning, while reducing reliance on brute-force search and sampling-based test-time scaling. We open-source the ARC-AGI-2 Reasoner code (https://github.com/CoreThink-AI/arc-agi-2-reasoner).

cs.AI↗

arXiv2026

Clinical Note Bloat Reduction for Efficient LLM Use

Health systems are rapidly deploying large language models (LLMs) that use clinical notes for clinical decision support applications. However, modern documentation practices rely heavily on templates, copy--paste shortcuts, and auto-populated fields, producing extensive duplicated text (``note bloat'') that dilutes clinically meaningful signal and substantially increases the computational cost of LLM use. We introduce TRACE, a scalable preprocessing pipeline that removes note bloat by leveraging EHR attribution metadata to identify templated and copied content and applying frequency-based deduplication when metadata are unavailable. We evaluated TRACE across four real--world clinical cohorts spanning liver transplantation, obstetrics, and inpatient care (5.3 million notes) using blinded physician review and downstream modeling tasks. TRACE removed 47.3% of chart text while preserving performance for information extraction and clinical outcome prediction. At a large academic medical center, this reduction corresponds to an estimated $9.5 million annual decrease in LLM inference costs assuming one query per encounter. These findings show how underutilized EHR metadata can enable more scalable and cost-efficient deployment of LLM-based clinical systems.

cs.CY↗

arXiv2026

Attention Head Entropy of LLMs Predicts Answer Correctness

Large language models (LLMs) often generate plausible yet incorrect answers, posing risks in safety-critical settings such as medicine. Human evaluation is expensive, and LLM-as-judge approaches risk introducing hidden errors. Recent white-box methods detect contextual hallucinations using model internals, focusing on the localization of the attention mass, but two questions remain open: do these approaches extend to predicting answer correctness, and do they generalize out-of-domains? We introduce Head Entropy, a method that predicts answer correctness from attention entropy patterns, specifically measuring the spread of the attention mass. Using sparse logistic regression on per-head 2-Renyi entropies, Head Entropy matches or exceeds baselines in-distribution and generalizes substantially better on out-of-domains, it outperforms the closest baseline on average by +8.5% AUROC. We further show that attention patterns over the question/context alone, before answer generation, already carry predictive signal using Head Entropy with on average +17.7% AUROC over the closest baseline. We evaluate across 5 instruction-tuned LLMs and 3 QA datasets spanning general knowledge, multi-hop reasoning, and medicine.

cs.LG↗

arXiv2026

Continuous-Utility Direct Preference Optimization

Large language model reasoning is often treated as a monolithic capability, relying on binary preference supervision that fails to capture partial progress or fine-grained reasoning quality. We introduce Continuous Utility Direct Preference Optimization (CU-DPO), a framework that aligns models to a portfolio of prompt-based cognitive strategies by replacing binary labels with continuous scores that capture fine-grained reasoning quality. We prove that learning with K strategies yields a Theta(K log K) improvement in sample complexity over binary preferences, and that DPO converges to the entropy-regularized utility-maximizing policy. To exploit this signal, we propose a two-stage training pipeline: (i) strategy selection, which optimizes the model to choose the best strategy for a given problem via best-vs-all comparisons, and (ii) execution refinement, which trains the model to correctly execute the selected strategy using margin-stratified pairs. On mathematical reasoning benchmarks, CU-DPO improves strategy selection accuracy from 35-46 percent to 68-78 percent across seven base models, yielding consistent downstream reasoning gains of up to 6.6 points on in-distribution datasets with effective transfer to out-of-distribution tasks.

cs.LG↗

arXiv2026

Training-Free Adaptation of New-Generation LLMs using Legacy Clinical Models

Adapting language models to the clinical domain through continued pretraining and instruction tuning requires costly retraining for each new model generation. We propose Cross-Architecture Proxy Tuning (CAPT), a model-ensembling approach that enables training-free adaptation of state-of-the-art general-domain models using existing clinical models. CAPT supports models with disjoint vocabularies, leveraging contrastive decoding to selectively inject clinically relevant signals while preserving the general-domain model's reasoning and fluency. On six clinical classification and text-generation tasks, CAPT with a new-generation general-domain model and an older-generation clinical model consistently outperforms both models individually and state-of-the-art ensembling approaches (average +17.6\% over UniTE, +41.4\% over proxy tuning across tasks). Through token-level analysis and physician case studies, we demonstrate that CAPT amplifies clinically actionable language, reduces context errors, and increases clinical specificity. This technique especially benefits healthcare institutions with constrained computational capacity that cannot support iterative clinical training and want to adopt emerging general-domain model advances.

cs.CL↗

arXiv2025

Resolution-Agnostic Neural Operators for Multi-Rate Sparse-View CT

Sparse-view Computed Tomography (CT) reconstructs images from a limited number of X-ray projections to reduce radiation and scanning time, which is an ill-posed inverse problem. Existing methods achieve high-fidelity reconstructions but overfit to a fixed acquisition setup, failing to generalize well across sampling rates. For example, convolutional neural networks (CNNs) use the same kernels across resolutions, leading to artifacts when data resolution changes. This is a critical limitation in clinical practice, where acquisition sampling settings vary across organs and diagnostic protocols. We propose Computed Tomography neural Operator (CTO), the first neural operator (NO) framework for CT reconstruction. CTO extends learning from fixed discretized grids to continuous function space, enabling a single model to generalize across measurement sampling rates without retraining. We also propose new NO architectural designs for CT: (i) a dual-domain NO architecture in both sinogram and image spaces, capturing complementary spatial-frequency information, and (ii) rotation-equivariant DIScrete-COntinuous convolutions (DISCO) that exploit the rotational structure inherent in tomographic acquisition. Empirically, CTO outperforms CNNs (> 3.4dB PSNR gain) and other baselines in multi-resolution settings across multiple CT datasets. Compared to state-of-the-art diffusion methods, CTO has 500x faster inference with an average 3dB gain. CTO further demonstrates strong out-of-distribution robustness, maintaining gains under cross-dataset transfer and noisy sinogram conditions. Ablation studies also validate each design choice. CTO establishes neural operators as a principled and practical paradigm for flexible, discretization-agnostic CT reconstruction. Our code is available at https://github.com/neuraloperator/sparse_ct.

eess.IV↗

arXiv2025

Structured Prompts Improve Evaluation of Language Models

As language models (LMs) are increasingly adopted across domains, high-quality benchmarking frameworks are essential for guiding deployment decisions. In practice, however, frameworks such as Holistic Evaluation of Language Models (HELM) typically evaluate models under a single static prompt configuration, even though model behavior depends strongly on prompt choice. As a result, reported scores can reflect prompt choice as much as model capability. Declarative prompting frameworks such as DSPy offer a scalable way to evaluate models under a set of structured prompting strategies rather than a static prompt configuration. We present a reproducible DSPy+HELM framework for studying how prompt choice impacts reported benchmark outcomes. Using five prompting methods, we evaluate four frontier and two open-source LMs across seven benchmarks against existing HELM baseline scores. By evaluating LMs across a family of prompt configurations, we find that prompt choice can materially impact leaderboard outcomes. In particular, structured prompting improves performance (by 6% on average), alters comparisons (leaderboard rankings shift on 5/7 benchmarks), with most gains coming from introducing chain-of-thought, and little additional benefit from more advanced optimizers. To our knowledge, this is the first study to systematically integrate structured prompting into an established evaluation framework and quantify how prompt choice alone can impact benchmark conclusions. We open-source (i) DSPy+HELM Evaluation (https://github.com/stanford-crfm/helm/pull/3893) and (ii) Prompt Optimization Pipeline (https://github.com/StanfordMIMI/dspy-helm).

cs.CL↗

arXiv2025

Prompt Triage: Structured Optimization Enhances Vision-Language Model Performance on Medical Imaging Benchmarks

Vision-language foundation models (VLMs) show promise for diverse imaging tasks but often underperform on medical benchmarks. Prior efforts to improve performance include model finetuning, which requires large domain-specific datasets and significant compute, or manual prompt engineering, which is hard to generalize and often inaccessible to medical institutions seeking to deploy these tools. These challenges motivate interest in approaches that draw on a model's embedded knowledge while abstracting away dependence on human-designed prompts to enable scalable, weight-agnostic performance improvements. To explore this, we adapt the Declarative Self-improving Python (DSPy) framework for structured automated prompt optimization in medical vision-language systems through a comprehensive, formal evaluation. We implement prompting pipelines for five medical imaging tasks across radiology, gastroenterology, and dermatology, evaluating 10 open-source VLMs with four prompt optimization techniques. Optimized pipelines achieved a median relative improvement of 53% over zero-shot prompting baselines, with the largest gains ranging from 300% to 3,400% on tasks where zero-shot performance is low. These results highlight the substantial potential of applying automated prompt optimization to medical AI systems, demonstrating significant gains for vision-based applications requiring accurate clinical image interpretation. By reducing dependence on prompt design to elicit intended outputs, these techniques allow clinicians to focus on patient care and clinical decision-making. Furthermore, our experiments offer scalability and preserve data privacy, demonstrating performance improvement on open-source VLMs. We publicly release our evaluation pipelines to support reproducible research on specialized medical tasks, available at https://github.com/DaneshjouLab/prompt-triage-lab.

cs.CV↗

arXiv2025

Patch-Based Diffusion for Data-Efficient, Radiologist-Preferred MRI Reconstruction

Magnetic resonance imaging (MRI) requires long acquisition times, raising costs, reducing accessibility, and making scans more susceptible to motion artifacts. Diffusion probabilistic models that learn data-driven priors can potentially assist in reducing acquisition time. However, they typically require large training datasets that can be prohibitively expensive to collect. Patch-based diffusion models have shown promise in learning effective data-driven priors over small real-valued datasets, but have not yet demonstrated clinical value in MRI. We extend the Patch-based Diffusion Inverse Solver (PaDIS) to complex-valued, multi-coil MRI reconstruction, and compare it against a state-of-the-art whole-image diffusion baseline (FastMRI-EDM) for 7x undersampled MRI reconstruction on the FastMRI brain dataset. We show that PaDIS-MRI models trained on small datasets of as few as 25 k-space images outperform FastMRI-EDM on image quality metrics (PSNR, SSIM, NRMSE), pixel-level uncertainty, cross-contrast generalization, and robustness to severe k-space undersampling. In a blinded study with three radiologists, PaDIS-MRI reconstructions were chosen as diagnostically superior in 91.7% of cases, compared to baselines (i) FastMRI-EDM and (ii) classical convex reconstruction with wavelet sparsity. These findings highlight the potential of patch-based diffusion priors for high-fidelity MRI reconstruction in data-scarce clinical settings where diagnostic confidence matters.

eess.IV↗

arXiv2025

Conditional Prior-based Non-stationary Channel Estimation Using Accelerated Diffusion Models

Wireless channels in motion-rich urban microcell (UMi) settings are non-stationary; mobility and scatterer dynamics shift the distribution over time, degrading classical and deep estimators. This work proposes conditional prior diffusion for channel estimation, which learns a history-conditioned score to denoise noisy channel snapshots. A temporal encoder with cross-time attention compresses a short observation window into a context vector, which captures the channel's instantaneous coherence and steers the denoiser via feature-wise modulation. In inference, an SNR-matched initialization selects the diffusion step whose marginal aligns with the measured input SNR, and the process follows a shortened, geometrically spaced schedule, preserving the signal-to-noise trajectory with far fewer iterations. Temporal self-conditioning with the previous channel estimate and a training-only smoothness penalty further stabilizes evolution without biasing the test-time estimator. Evaluations on a 3GPP benchmark show lower NMSE across all SNRs than LMMSE, GMM, LSTM, and LDAMP baselines, demonstrating stable performance and strong high SNR fidelity.

cs.DC↗

arXiv2025

MedFactEval and MedAgentBrief: A Framework and Workflow for Generating and Evaluating Factual Clinical Summaries

Evaluating factual accuracy in Large Language Model (LLM)-generated clinical text is a critical barrier to adoption, as expert review is unscalable for the continuous quality assurance these systems require. We address this challenge with two complementary contributions. First, we introduce MedFactEval, a framework for scalable, fact-grounded evaluation where clinicians define high-salience key facts and an "LLM Jury"--a multi-LLM majority vote--assesses their inclusion in generated summaries. Second, we present MedAgentBrief, a model-agnostic, multi-step workflow designed to generate high-quality, factual discharge summaries. To validate our evaluation framework, we established a gold-standard reference using a seven-physician majority vote on clinician-defined key facts from inpatient cases. The MedFactEval LLM Jury achieved almost perfect agreement with this panel (Cohen's kappa=81%), a performance statistically non-inferior to that of a single human expert (kappa=67%, P < 0.001). Our work provides both a robust evaluation framework (MedFactEval) and a high-performing generation workflow (MedAgentBrief), offering a comprehensive approach to advance the responsible deployment of generative AI in clinical workflows.

cs.CL↗

arXiv2025

MedVAL: Toward Expert-Level Medical Text Validation with Language Models

With the growing use of language models (LMs) in clinical environments, there is an immediate need to evaluate the accuracy and safety of LM-generated medical text. Currently, such evaluation relies solely on manual physician review. However, detecting errors in LM-generated text is challenging because 1) manual review is costly and 2) expert-composed reference outputs are often unavailable in real-world settings. While the "LLM-as-a-judge" paradigm offers scalable evaluation, even frontier LMs can miss subtle but clinically significant errors. We propose MedVAL, a novel, self-supervised, data-efficient distillation method that leverages synthetic data to train evaluator LMs to assess whether LM-generated medical outputs are factually consistent with inputs, without requiring physician labels or reference outputs. To evaluate LM performance, we introduce MedVAL-Bench, a dataset of 840 physician-annotated outputs across 6 diverse medical tasks capturing real-world challenges. Across 10 state-of-the-art LMs spanning open-source and proprietary models, MedVAL distillation significantly improves (p < 0.001) alignment with physicians across seen and unseen tasks, increasing average F1 scores from 66% to 83%. Despite strong baseline performance, MedVAL improves the best-performing proprietary LM (GPT-4o) by 8% without training on physician-labeled data, demonstrating a performance statistically non-inferior to a single human expert on a subset annotated by multiple physicians (p < 0.001). To support a scalable, risk-aware pathway towards clinical integration, we open-source: 1) Codebase (https://github.com/StanfordMIMI/MedVAL), 2) MedVAL-Bench (https://huggingface.co/datasets/stanfordmimi/MedVAL-Bench), 3) MedVAL-4B (https://huggingface.co/stanfordmimi/MedVAL-4B). Our benchmark provides evidence of LMs approaching expert-level ability in validating AI-generated medical text.

cs.CL↗

arXiv2025

MedHELM: Holistic Evaluation of Large Language Models for Medical Tasks

While large language models (LLMs) achieve near-perfect scores on medical licensing exams, these evaluations inadequately reflect the complexity and diversity of real-world clinical practice. We introduce MedHELM, an extensible evaluation framework for assessing LLM performance for medical tasks with three key contributions. First, a clinician-validated taxonomy spanning 5 categories, 22 subcategories, and 121 tasks developed with 29 clinicians. Second, a comprehensive benchmark suite comprising 35 benchmarks (17 existing, 18 newly formulated) providing complete coverage of all categories and subcategories in the taxonomy. Third, a systematic comparison of LLMs with improved evaluation methods (using an LLM-jury) and a cost-performance analysis. Evaluation of 9 frontier LLMs, using the 35 benchmarks, revealed significant performance variation. Advanced reasoning models (DeepSeek R1: 66% win-rate; o3-mini: 64% win-rate) demonstrated superior performance, though Claude 3.5 Sonnet achieved comparable results at 40% lower estimated computational cost. On a normalized accuracy scale (0-1), most models performed strongly in Clinical Note Generation (0.73-0.85) and Patient Communication & Education (0.78-0.83), moderately in Medical Research Assistance (0.65-0.75), and generally lower in Clinical Decision Support (0.56-0.72) and Administration & Workflow (0.53-0.63). Our LLM-jury evaluation method achieved good agreement with clinician ratings (ICC = 0.47), surpassing both average clinician-clinician agreement (ICC = 0.43) and automated baselines including ROUGE-L (0.36) and BERTScore-F1 (0.44). Claude 3.5 Sonnet achieved comparable performance to top models at lower estimated cost. These findings highlight the importance of real-world, task-specific evaluation for medical use of LLMs and provides an open source framework to enable this.

cs.CL↗

arXiv2025

Splitwiser: Efficient LM inference with constrained resources

Efficient inference of LLMs remains a crucial challenge, with two main phases: a compute-intensive prompt computation and a memory-intensive token generation. Despite existing batching and scheduling techniques, token generation phases fail to fully utilize compute resources, especially when compared to prompt computation phases. To address these challenges, we propose Splitwiser, a methodology that splits the two phases of an LLM inference request onto the same GPU, thereby reducing overhead and improving memory access and cache utilization. By eliminating the need to transfer data across devices, Splitwiser aims to minimize network-related overheads. In this report, we describe the basic structure of our proposed pipeline while sharing preliminary results and analysis. We implement our proposed multiprocessing design on two widely-used and independent LLM architectures: Huggingface and vLLM. We open-source our code for the respective implementations: 1) Huggingface (https://github.com/asad-aali/splitwiser), and 2) vLLM (https://github.com/adney11/vllm-sysml).

cs.AR↗

arXiv2024

Robust multi-coil MRI reconstruction via self-supervised denoising

We study the effect of incorporating self-supervised denoising as a pre-processing step for training deep learning (DL) based reconstruction methods on data corrupted by Gaussian noise. K-space data employed for training are typically multi-coil and inherently noisy. Although DL-based reconstruction methods trained on fully sampled data can enable high reconstruction quality, obtaining large, noise-free datasets is impractical. We leverage Generalized Stein's Unbiased Risk Estimate (GSURE) for denoising. We evaluate two DL-based reconstruction methods: Diffusion Probabilistic Models (DPMs) and Model-Based Deep Learning (MoDL). We evaluate the impact of denoising on the performance of these DL-based methods in solving accelerated multi-coil magnetic resonance imaging (MRI) reconstruction. The experiments were carried out on T2-weighted brain and fat-suppressed proton-density knee scans. We observed that self-supervised denoising enhances the quality and efficiency of MRI reconstructions across various scenarios. Specifically, employing denoised images rather than noisy counterparts when training DL networks results in lower normalized root mean squared error (NRMSE), higher structural similarity index measure (SSIM) and peak signal-to-noise ratio (PSNR) across different SNR levels, including 32dB, 22dB, and 12dB for T2-weighted brain data, and 24dB, 14dB, and 4dB for fat-suppressed knee data. Overall, we showed that denoising is an essential pre-processing technique capable of improving the efficacy of DL-based MRI reconstruction methods under diverse conditions. By refining the quality of input data, denoising enables training more effective DL networks, potentially bypassing the need for noise-free reference MRI scans.

eess.IV↗