arXiv subjects

Reza Khanmohammadi

Publications and source records attributed to Reza Khanmohammadi.

16 recordsLinked to original sources

arXiv2026

Grounded or Guessing? LVLM Confidence Estimation via Blind-Image Contrastive Ranking

Large vision-language models (LVLMs) suffer from visual ungroundedness: they can produce a fluent, confident, and even correct response driven entirely by language priors, with the image contributing nothing to the prediction. Existing confidence estimation methods cannot detect this, as they observe model behavior under normal inference with no mechanism to determine whether a prediction was shaped by the image or by text alone. We introduce BICR (Blind-Image Contrastive Ranking), a model-agnostic confidence estimation framework. The LVLM stays entirely frozen; instead, we extract its hidden states twice per sample, once with the real image-question pair and once with the image blacked out while the question is held fixed. A small, separate probe is then trained on the real-image hidden state, regularized by a ranking loss that penalizes higher confidence on the blacked-out view, teaching it to treat visual grounding as a signal of reliability. At inference only the single real-image pass is used, so BICR adds zero cost beyond a standard single-view probe. Evaluated across five modern LVLMs and seven baselines on a benchmark covering visual question answering, object hallucination detection, medical imaging, and financial document understanding, BICR achieves the best cross-LVLM average on both calibration and discrimination simultaneously, with statistically significant discrimination gains robust to cluster-aware analysis, at 4-18x fewer parameters than the strongest probing baseline.

cs.CL↗

arXiv2026

Calibrated Triage, Not Autonomy: Confidence Estimation for Medical Vision-Language Models

A vision-language model can answer a question about a chest radiograph or a pathology slide fluently and confidently while barely using the image, relying instead on language priors. In medicine this is the failure that matters most: the answer looks trustworthy and is not, and the natural safeguard is a confidence score reliable enough to say when the model should abstain. We ask a deployment question rather than an accuracy one: how much medical imaging work a vision-language model can safely defer on its own, and which confidence signal makes that possible. We evaluate nine confidence estimators, spanning training-free logit baselines, prompt-based self-reports, and trained internal probes, across five open-weight LVLMs and three medical VQA datasets covering broad clinical imaging, radiology, and pathology, every probe trained only on natural images and applied to medicine without adaptation. Recast as bounded selective prediction, the comparison is cautionary. Standard metrics mislead: discrimination barely separates the estimators, and a fixed high-confidence cutoff separates them far less than it appears, because their scores sit on incomparable scales; no estimator is reliably best across domains or models. What can be safely deferred is set at two levels: base-model competence fixes a ceiling, and the confidence layer determines how much of it is reachable. At a 20% error tolerance the strongest estimator defers about a quarter of radiology cases under a distribution-free guarantee and a third under a held-out threshold, and little to none of pathology. The usable role is calibrated triage under clinical oversight, not autonomous deferral: a good estimator makes a competent model defer safely where it is competent, but none manufactures reliability where the base model lacks it. We release all outputs, correctness judgments, and confidence scores, with code.

cs.CL↗

arXiv2026

Confidence Estimation for Financial Vision-Language Models in Chart and Document Understanding

LVLMs are increasingly used to read financial charts, tables, and documents, where a single misread figure can move a decision and the most authoritative-looking answer is sometimes one the model produced without reading the exhibit. The operational question is therefore trust, not accuracy: which answers can be acted on, and which escalated to a reviewer. We evaluate seven confidence estimators, three inference-only and four trained internal probes, across five open-weight LVLMs and four conditions from three financial visual question-answering benchmarks, one bilingual; every probe is trained only on natural images and applied to finance without adaptation, so the results measure out-of-distribution transfer. Three findings hold. First, the scarce property is calibration, not ranking: the inference baselines rank correct above incorrect answers competitively but are badly overconfident, calibration error far above what a threshold can tolerate, and only the trained probes produce a thresholdable score. Second, reliability is structured rather than global, along two axes a practitioner can read directly: the best estimator shifts with both model and task, none leading more than eight of twenty (model, condition) cells, and a controlled bilingual contrast exposes an apparent language robustness as a composition artifact that dissolves once models are read one at a time. Third, cast as deferral under an error budget, how much can be safely automated is set first by the model's competence and only narrowed by its confidence, so deferral clears a real share of the easiest condition and almost none of the hardest, near zero at a strict 5% budget. Two trained probes carry the calibration a deferral policy needs, and among them only the grounding-aware one lowers its confidence on answers a model gives without using the figure, separating detected non-grounding from a fluent guess.

cs.CL↗

arXiv2026

Model Confidence Under Answer-Preserving Attacks: An Informativeness-Manipulability Frontier

Deployed vision-language systems often gate their answers on confidence, making confidence robustness relevant to oversight. We study confidence readouts under white-box, image-only attacks constrained to preserve the generated answer byte-identically. Under a reachability assumption, an unmovable readout cannot outperform the answer-string accuracy prior, whose pooled value is 0.617. Independently of that assumption, a uniform amplitude certificate below a measurable threshold guarantees adversarial discrimination above the same floor. Across four vision-language models, three visual question answering benchmarks, five deployed confidence channels and two defense estimators, direct or surrogate-aimed attacks produce itemwise feasible perturbations that refute this uniform certificate in all 84 estimator-by-cell combinations. Coordinated correctness-label-aware attacks drive adversarial discrimination to or below the answer-string floor in all sixty deployed-channel cells, including all fifty-nine that begin above it. Hidden-state interventions and an open-ended text-model activation-space replication show that comparable confidence movement can be induced at the representation level rather than only through adversarial images. None of four tested defense families establishes a robust alternative under the specific evaluation applied to it. In a confidence-gated simulation, a coordinated token-probability attack transferred to a hidden-state gate causes up to 84.8% of previously rejected wrong answers to become accepted. After reweighting to each benchmark's natural correctness prevalence, accepted accuracy falls below the no-gate baseline in eight of twelve cells under transfer and all twelve under a direct gate-aimed attack. Under the studied threat model and budget, confidence is therefore an integrity-sensitive rather than intrinsically robust oversight signal.

cs.CR↗

arXiv2026

How Reliable are Confidence Estimators for Large Reasoning Models? A Systematic Benchmark on High-Stakes Domains

The miscalibration of Large Reasoning Models (LRMs) undermines their reliability in high-stakes domains, necessitating methods to accurately estimate the confidence of their long-form, multi-step outputs. To address this gap, we introduce the Reasoning Model Confidence estimation Benchmark (RMCB), a public resource of 347,496 reasoning traces from six popular LRMs across different architectural families. The benchmark is constructed from a diverse suite of datasets spanning high-stakes domains, including clinical, financial, legal, and mathematical reasoning, alongside complex general reasoning benchmarks, with correctness annotations provided for all samples. Using RMCB, we conduct a large-scale empirical evaluation of over ten distinct representation-based methods, spanning sequential, graph-based, and text-based architectures. Our central finding is a persistent trade-off between discrimination (AUROC) and calibration (ECE): text-based encoders achieve the best AUROC (0.672), while structurally-aware models yield the best ECE (0.148), with no single method dominating both. Furthermore, we find that increased architectural complexity does not reliably outperform simpler sequential baselines, suggesting a performance ceiling for methods relying solely on chunk-level hidden states. This work provides the most comprehensive benchmark for this task to date, establishing rigorous baselines and demonstrating the limitations of current representation-based paradigms.

cs.CL↗

arXiv2025

Calibrating LLM Confidence by Probing Perturbed Representation Stability

Miscalibration in Large Language Models (LLMs) undermines their reliability, highlighting the need for accurate confidence estimation. We introduce CCPS (Calibrating LLM Confidence by Probing Perturbed Representation Stability), a novel method analyzing internal representational stability in LLMs. CCPS applies targeted adversarial perturbations to final hidden states, extracts features reflecting the model's response to these perturbations, and uses a lightweight classifier to predict answer correctness. CCPS was evaluated on LLMs from 8B to 32B parameters (covering Llama, Qwen, and Mistral architectures) using MMLU and MMLU-Pro benchmarks in both multiple-choice and open-ended formats. Our results show that CCPS significantly outperforms current approaches. Across four LLMs and three MMLU variants, CCPS reduces Expected Calibration Error by approximately 55% and Brier score by 21%, while increasing accuracy by 5 percentage points, Area Under the Precision-Recall Curve by 4 percentage points, and Area Under the Receiver Operating Characteristic Curve by 6 percentage points, all relative to the strongest prior method. CCPS delivers an efficient, broadly applicable, and more accurate solution for estimating LLM confidence, thereby improving their trustworthiness.

cs.CL↗

arXiv2024

Hybrid Student-Teacher Large Language Model Refinement for Cancer Toxicity Symptom Extraction

Large Language Models (LLMs) offer significant potential for clinical symptom extraction, but their deployment in healthcare settings is constrained by privacy concerns, computational limitations, and operational costs. This study investigates the optimization of compact LLMs for cancer toxicity symptom extraction using a novel iterative refinement approach. We employ a student-teacher architecture, utilizing Zephyr-7b-beta and Phi3-mini-128 as student models and GPT-4o as the teacher, to dynamically select between prompt refinement, Retrieval-Augmented Generation (RAG), and fine-tuning strategies. Our experiments on 294 clinical notes covering 12 post-radiotherapy toxicity symptoms demonstrate the effectiveness of this approach. The RAG method proved most efficient, improving average accuracy scores from 0.32 to 0.73 for Zephyr-7b-beta and from 0.40 to 0.87 for Phi3-mini-128 during refinement. In the test set, both models showed an approximate 0.20 increase in accuracy across symptoms. Notably, this improvement was achieved at a cost 45 times lower than GPT-4o for Zephyr and 79 times lower for Phi-3. These results highlight the potential of iterative refinement techniques in enhancing the capabilities of compact LLMs for clinical applications, offering a balance between performance, cost-effectiveness, and privacy preservation in healthcare settings.

cs.CL↗

arXiv2024

Iterative Prompt Refinement for Radiation Oncology Symptom Extraction Using Teacher-Student Large Language Models

This study introduces a novel teacher-student architecture utilizing Large Language Models (LLMs) to improve prostate cancer radiotherapy symptom extraction from clinical notes. Mixtral, the student model, initially extracts symptoms, followed by GPT-4, the teacher model, which refines prompts based on Mixtral's performance. This iterative process involved 294 single symptom clinical notes across 12 symptoms, with up to 16 rounds of refinement per epoch. Results showed significant improvements in extracting symptoms from both single and multi-symptom notes. For 59 single symptom notes, accuracy increased from 0.51 to 0.71, precision from 0.52 to 0.82, recall from 0.52 to 0.72, and F1 score from 0.49 to 0.73. In 375 multi-symptom notes, accuracy rose from 0.24 to 0.43, precision from 0.6 to 0.76, recall from 0.24 to 0.43, and F1 score from 0.20 to 0.44. These results demonstrate the effectiveness of advanced prompt engineering in LLMs for radiation oncology use.

cs.CL↗

arXiv2024

The Influence of Biomedical Research on Future Business Funding: Analyzing Scientific Impact and Content in Industrial Investments

This paper investigates the relationship between scientific innovation in biomedical sciences and its impact on industrial activities, focusing on how the historical impact and content of scientific papers influenced future funding and innovation grant application content for small businesses. The research incorporates bibliometric analyses along with SBIR (Small Business Innovation Research) data to yield a holistic view of the science-industry interface. By evaluating the influence of scientific innovation on industry across 10,873 biomedical topics and taking into account their taxonomic relationships, we present an in-depth exploration of science-industry interactions where we quantify the temporal effects and impact latency of scientific advancements on industrial activities, spanning from 2010 to 2021. Our findings indicate that scientific progress substantially influenced industrial innovation funding and the direction of industrial innovation activities. Approximately 76% and 73% of topics showed a correlation and Granger-causality between scientific interest in papers and future funding allocations to relevant small businesses. Moreover, around 74% of topics demonstrated an association between the semantic content of scientific abstracts and future grant applications. Overall, the work contributes to a more nuanced and comprehensive understanding of the science-industry interface, opening avenues for more strategic resource allocation and policy developments aimed at fostering innovation.

cs.CE↗

arXiv2023

An Introduction to Natural Language Processing Techniques and Framework for Clinical Implementation in Radiation Oncology

Natural Language Processing (NLP) is a key technique for developing Medical Artificial Intelligence (AI) systems that leverage Electronic Health Record (EHR) data to build diagnostic and prognostic models. NLP enables the conversion of unstructured clinical text into structured data that can be fed into AI algorithms. The emergence of the transformer architecture and large language models (LLMs) has led to remarkable advances in NLP for various healthcare tasks, such as entity recognition, relation extraction, sentence similarity, text summarization, and question answering. In this article, we review the major technical innovations that underpin modern NLP models and present state-of-the-art NLP applications that employ LLMs in radiation oncology research. However, these LLMs are prone to many errors such as hallucinations, biases, and ethical violations, which necessitate rigorous evaluation and validation before clinical deployment. As such, we propose a comprehensive framework for assessing the NLP models based on their purpose and clinical fit, technical performance, bias and trust, legal and ethical implications, and quality assurance, prior to implementation in clinical radiation oncology. Our article aims to provide guidance and insights for researchers and clinicians who are interested in developing and using NLP models in clinical radiation oncology.

cs.CL↗

arXiv2023

The Broad Impact of Feature Imitation: Neural Enhancements Across Financial, Speech, and Physiological Domains

Initialization of neural network weights plays a pivotal role in determining their performance. Feature Imitating Networks (FINs) offer a novel strategy by initializing weights to approximate specific closed-form statistical features, setting a promising foundation for deep learning architectures. While the applicability of FINs has been chiefly tested in biomedical domains, this study extends its exploration into other time series datasets. Three different experiments are conducted in this study to test the applicability of imitating Tsallis entropy for performance enhancement: Bitcoin price prediction, speech emotion recognition, and chronic neck pain detection. For the Bitcoin price prediction, models embedded with FINs reduced the root mean square error by around 1000 compared to the baseline. In the speech emotion recognition task, the FIN-augmented model increased classification accuracy by over 3 percent. Lastly, in the CNP detection experiment, an improvement of about 7 percent was observed compared to established classifiers. These findings validate the broad utility and potency of FINs in diverse applications.

cs.LG↗

arXiv2022

MambaNet: A Hybrid Neural Network for Predicting the NBA Playoffs

In this paper, we present Mambanet: a hybrid neural network for predicting the outcomes of Basketball games. Contrary to other studies, which focus primarily on season games, this study investigates playoff games. MambaNet is a hybrid neural network architecture that processes a time series of teams' and players' game statistics and generates the probability of a team winning or losing an NBA playoff match. In our approach, we utilize Feature Imitating Networks to provide latent signal-processing feature representations of game statistics to further process with convolutional, recurrent, and dense neural layers. Three experiments using six different datasets are conducted to evaluate the performance and generalizability of our architecture against a wide range of previous studies. Our final method successfully predicted the AUC from 0.72 to 0.82, beating the best-performing baseline models by a considerable margin.

cs.LG↗

arXiv2022

Prose2Poem: The Blessing of Transformers in Translating Prose to Persian Poetry

Persian Poetry has consistently expressed its philosophy, wisdom, speech, and rationale on the basis of its couplets, making it an enigmatic language on its own to both native and non-native speakers. Nevertheless, the notice able gap between Persian prose and poem has left the two pieces of literature medium-less. Having curated a parallel corpus of prose and their equivalent poems, we introduce a novel Neural Machine Translation (NMT) approach to translate prose to ancient Persian poetry using transformer-based Language Models in an extremely low-resource setting. More specifically, we trained a Transformer model from scratch to obtain initial translations and pretrained different variations of BERT to obtain final translations. To address the challenge of using masked language modelling under poeticness criteria, we heuristically joined the two models and generated valid poems in terms of automatic and human assessments. Final results demonstrate the eligibility and creativity of our novel heuristically aided approach among Literature professionals and non-professionals in generating novel Persian poems.

cs.CL↗

arXiv2021

Fetal Gender Identification using Machine and Deep Learning Algorithms on Phonocardiogram Signals

Phonocardiogram (PCG) signal analysis is a critical, widely-studied technology to noninvasively analyze the heart's mechanical activity. Through evaluating heart sounds, this technology has been chiefly leveraged as a preliminary solution to automatically diagnose Cardiovascular diseases among adults; however, prenatal tasks such as fetal gender identification have been relatively less studied using fetal Phonocardiography (FPCG). In this work, we apply common PCG signal processing techniques on the gender-tagged Shiraz University Fetal Heart Sounds Database and study the applicability of previously proposed features in classifying fetal gender using both Machine Learning and Deep Learning models. Even though PCG data acquisition's cost-effectiveness and feasibility make it a convenient method of Fetal Heart Rate (FHR) monitoring, the contaminated nature of PCG signals with the noise of various types makes it a challenging modality. To address this problem, we experimented with both static and adaptive noise reduction techniques such as Low-pass filtering, Denoising Autoencoders, and Source Separators. We apply a wide range of previously proposed classifiers to our dataset and propose a novel ensemble method of Fetal Gender Identification (FGI). Our method substantially outperformed the baseline and reached up to 91% accuracy in classifying fetal gender of unseen subjects.

eess.SP↗

arXiv2021

COPER: a Query-adaptable Semantics-based Search Engine for Persian COVID-19 Articles

With the surge of pretrained language models, a new pathway has been opened to incorporate Persian text contextual information. Meanwhile, as many other countries, including Iran, are fighting against COVID-19, a plethora of COVID-19 related articles has been published in Iranian Healthcare magazines to better inform the public of the situation. However, finding answers in this sheer volume of information is an extremely difficult task. In this paper, we collected a large dataset of these articles, leveraged different BERT variations as well as other keyword models such as BM25 and TF-IDF, and created a search engine to sift through these documents and rank them, given a user's query. Our final search engine consists of a ranker and a re-ranker, which adapts itself to the query. We fine-tune our models using Semantic Textual Similarity and evaluate them with standard task metrics. Our final method outperforms the rest by a considerable margin.

cs.IR↗

arXiv2021

PGST: a Polyglot Gender Style Transfer method

Recent developments in Text Style Transfer have led this field to be more highlighted than ever. The task of transferring an input's style to another is accompanied by plenty of challenges (e.g., fluency and content preservation) that need to be taken care of. In this research, we introduce PGST, a novel polyglot text style transfer approach in the gender domain, composed of different constitutive elements. In contrast to prior studies, it is feasible to apply a style transfer method in multiple languages by fulfilling our method's predefined elements. We have proceeded with a pre-trained word embedding for token replacement purposes, a character-based token classifier for gender exchange purposes, and a beam search algorithm for extracting the most fluent combination. Since different approaches are introduced in our research, we determine a trade-off value for evaluating different models' success in faking our gender identification model with transferred text. To demonstrate our method's multilingual applicability, we applied our method on both English and Persian corpora and ended up defeating our proposed gender identification model by 45.6% and 39.2%, respectively. While this research's focus is not limited to a specific language, our obtained evaluation results are highly competitive in an analogy among English state of the art methods.

cs.CL↗