arXiv subjects

Elham Dolatabadi

Publications and source records attributed to Elham Dolatabadi.

At least 19 recordsLinked to original sources

arXiv2026

Training Therapeutic Judges and Multi-Agent Systems for Human-Aligned Mental Health Support

Large language models show promise for mental health support, yet therapeutic quality improves only when evaluation functions as an actionable control signal rather than a passive metric. We introduce a framework that formulates therapeutic response generation as a decision-refinement problem driven by multi-dimensional, human-aligned evaluation. In Stage I, we introduce TheraJudge, an open-source therapeutic evaluator trained via preference-based optimization on human-annotated data to produce reliable judgments across 7 psychological dimensions. In Stage II, we introduce TheraAgent, which operationalizes TheraJudge's evaluations through a coordinated refinement process with specialized Critic, Coach, and Therapist roles that translate evaluative signals into targeted response revisions. Empirically, TheraJudge achieves strong agreement with clinician ratings, with intraclass correlation coefficients (ICC = 0.87-0.95), surpassing supervised baselines and strong closed-source judges, particularly on critical dimensions such as Safety, Relevance, and Empathy. Acting on these evaluations, TheraAgent yields a +0.43 improvement in human-rated therapeutic quality (on a 5-point scale) under blind evaluation, with 96\% clinician inter-rater reliability. Low-quality responses ($\leq 3$) improve by +2.45 points with a 94\% recovery rate, demonstrating targeted correction of unsafe outputs. Overall, our results indicate that effective alignment of mental-health LLMs stems from acting on human-aligned evaluation, rather than relying solely on stronger generation. We release code at https://github.com/vis-nlp/TheraAlign.

cs.CL↗

arXiv2026

Towards Understanding and Measuring COGNITIVE ATROPHY in LLM Behaviour

Recent incidents involving LLMs used for mental-health support reveal a critical evaluation gap: surface-level safety scores do not capture how models behave across realistic, emotionally sensitive interactions over time. Existing benchmarks measure knowledge, safety, or static response quality, but miss whether LLM interactions help users keep reflecting, coping, and making decisions themselves. We formalize this missing dimension as COGNITIVE ATROPHY, a process-level behavioural measure in AI-mediated mental-health support distinct from safety and helpfulness. To measure it, we introduce COGNITIVE ATROPHY BENCH, a clinically grounded benchmark built from 1,576 fully human-generated counseling conversations, 15,680 turns, and 42,230 responses from five LLMs. Three clinical and neuropsychology experts developed a 20-attribute schema spanning user context, response behaviour, and global risk flags; six trained clinical reviewers applied it with span-grounded evidence, producing 5,324 reviewer judgments. We further introduce the User-Input Risk Index (UIRI), the Cognitive Atrophy Risk Index (ARI), and trajectory summaries. Across five LLMs, models show a consistent moderate-to-high level of atrophy-aligned behaviour across single and multi-turn settings. While models generally respond to overt safety cues, they adapt less reliably when users seek solutions or decisions. The dominant recurring patterns are directive advice, problem-solving, recommendation responses, topic shifts, and forms of validation that may reinforce dependence rather than reflection. Our work makes COGNITIVE ATROPHY measurable and provides a foundation for auditing model behaviour in sensitive LLM conversations.

cs.HC↗

arXiv2026

OpenMedReason: Scientific Reasoning Supervision for Medical Vision-Language Models

High-stakes clinical use of large vision-language models (LVLMs) requires reasoning that is grounded in visual evidence and clinical knowledge, not just correct final answers. We introduce OpenMedReason, a large-scale, open multimodal medical reasoning corpus comprising approximately 450K image-question-answer instances whose reasoning traces are primarily derived from curated biomedical, human-authored scientific articles. OpenMedReason provides high-fidelity supervision beyond synthetic chains of thought, covering diverse medical domain vision modalities such as radiological scans, microscopic images, visible light photographs, charts, and others. We complement it with OpenMedReason-Bench, a held-out benchmark that allows fine-grained evaluation of LVLMs along three complementary axes of capability, including perception, medical knowledge, and rationale, enabling diagnostic evaluation beyond final-answer accuracy. OpenMedReason is a rich training resource that exhibits its effectiveness in both supervised fine-tuning (SFT) and reinforcement-based alignment. Training with OpenMedReason yields a 20% average improvement in VQA accuracy over the base model and achieves performance within 4.2% of the strongest comparable-scale medical LVLMs. Fine-grained performance analysis confirms that the gains are not concentrated in any single axis: OpenMedReason improves perception, medical knowledge, and rationale jointly, and its reasoning traces are preferred over those of the base model in 86.1% of pairwise comparisons. We release the code and dataset at huggingface.co/datasets/neginb/OpenMedReason.

cs.CV↗

arXiv2026

Analyzing and Improving Fine-grained Preference Optimization in Medical LVLMs

Large Vision-Language Models (LVLMs) have achieved strong performance across medical imaging tasks, yet they remain prone to factual inconsistencies, poor visual grounding, and misalignment with clinically meaningful feedback. Existing post-training alignment approaches, including Direct Preference Optimization (DPO) and its variants, face three critical limitations in the medical domain: (1) sequence-level reward signals treat clinically critical tokens identically to generic filler text; (2) reliance on static supervised fine-tuning references as preferred responses introduces an off-policy distribution shift, steering optimization toward stylistic artifacts over clinical correctness; and (3) alignment objectives lack explicit visual grounding constraints, leaving models insensitive to subtle yet diagnostically decisive pathological features. Our method leverages a bidirectional token-wise KL regularizer alongside a visual-contrastive grounding objective that pairs clean and lesion-corrupted images to penalize responses generated without adequate visual evidence. Together, these components form a fine-grained, on-policy alignment framework that constructs preference pairs by minimally editing model-generated outputs, correcting only clinically erroneous spans while preserving the original linguistic style. Extensive experiments across medical imaging tasks and clinical text generation benchmarks validate the effectiveness of our approach.

cs.CV↗

arXiv2026

Keyphrase Generative Representation of Youth Crisis Conversations Beyond Static Taxonomies

Crisis Responders (CRs) rapidly assess thousands of youth SMS conversations each year to identify mental health concerns and guide support. Yet youth distress is increasingly expressed through evolving and context-specific language that often does not fit fixed-label taxonomies. This work analyzed 703,975 de-identified Kids Help Phone conversations (2018-2023) and expanded KHP's 19-label issue taxonomy into a 39-label hierarchical schema. We then introduce Keyphrase Generative Representation (KGR), a constrained LLM generating concise, conversation-specific keyphrases, evaluated across 129 conversations and 387 expert annotations. The expanded taxonomy achieved expert consensus reliability, with an accuracy of 0.96, and expert review found that 81% of keyphrases accurately reflected content and 74% improved clarity. KGR surfaced identity-linked themes absent from the fixed taxonomy, including immigration problems and caregiver burden, and supported a topic-retrieval workflow that increased accuracy from 0.25 to 0.70 (+0.45) over the manual analyst process. KGR marks a shift toward hybrid, interpretable generative representations that extend crisis response beyond static taxonomies to surface emerging and culturally grounded patterns of youth distress.

cs.CL↗

arXiv2026

Fine-Grained Benchmark Generation for Comprehensive Evaluation of Foundation Models

Evaluation of foundation models often rely on aggregate scores from benchmarks that lack comprehensive coverage and metadata for a fine-grained evaluation. We introduce a framework for automated benchmark generation. Our framework generates evaluation problems grounded in reference material, such as textbooks, producing benchmarks with broad coverage, rich metadata, and robustness to contamination. The pipeline employs a multi-agent architecture for problem generation and a solution-graph-driven strategy that significantly improves the reliability of ground truth solutions. Using the framework, we generate three benchmarks in Machine Learning, Corporate Finance, and Personal Finance. Expert review finds a significantly lower ground-truth error rate than previous benchmarks such as MMLU and GSM8K. Evaluation of 12 commercial and open-source models shows that our benchmarks achieve near-uniform competency coverage and surface performance differences across models that existing benchmarks fail to capture. We will open-source the framework and our curated benchmarks soon.

cs.LG↗

arXiv2026

When Does RL Help Medical VLMs? Disentangling Vision, SFT, and RL Gains

Reinforcement learning (RL) is increasingly used to post-train medical Vision-Language Models (VLMs), yet it remains unclear whether RL improves medical visual reasoning or mainly sharpens behaviors already induced by supervised fine-tuning (SFT). We present a controlled study that disentangles these effects along three axes: vision, SFT, and RL. Using MedMNIST as a multi-modality testbed, we probe visual perception by benchmarking VLM vision towers against vision-only baselines, quantify reasoning support and sampling efficiency via Accuracy@1 versus Pass@K, and evaluate when RL closes the support gap and how gains transfer across modalities. We find that RL is most effective when the model already has non-trivial support (high Pass@K): it primarily sharpens the output distribution, improving Acc@1 and sampling efficiency, while SFT expands support and makes RL effective. Based on these findings, we propose a boundary-aware recipe and instantiate it by RL post-training an OctoMed-initialized model on a small, balanced subset of PMC multiple-choice VQA, achieving strong average performance across six medical VQA benchmarks.

cs.CV↗

arXiv2026

Assessing the Quality of Mental Health Support in LLM Responses through Multi-Attribute Human Evaluation

The escalating global mental health crisis, marked by persistent treatment gaps, availability, and a shortage of qualified therapists, positions Large Language Models (LLMs) as a promising avenue for scalable support. While LLMs offer potential for accessible emotional assistance, their reliability, therapeutic relevance, and alignment with human standards remain challenging to address. This paper introduces a human-grounded evaluation methodology designed to assess LLM generated responses in therapeutic dialogue. Our approach involved curating a dataset of 500 mental health conversations from datasets with real-world scenario questions and evaluating the responses generated by nine diverse LLMs, including closed source and open source models. More specifically, these responses were evaluated by two psychiatric trained experts, who independently rated each on a 5 point Likert scale across a comprehensive 6 attribute rubric. This rubric captures Cognitive Support and Affective Resonance, providing a multidimensional perspective on therapeutic quality. Our analysis reveals that LLMs provide strong cognitive reliability by producing safe, coherent, and clinically appropriate information, but they demonstrate unstable affective alignment. Although closed source models (e.g., GPT-4o) offer balanced therapeutic responses, open source models show greater variability and emotional flatness. We reveal a persistent cognitive-affective gap and highlight the need for failure aware, clinically grounded evaluation frameworks that prioritize relational sensitivity alongside informational accuracy in mental health oriented LLMs. We advocate for balanced evaluation protocols with human in the loop that center on therapeutic sensitivity and provide a framework to guide the responsible design and clinical oversight of mental health oriented conversational AI.

cs.AI↗

arXiv2025

VLDBench Evaluating Multimodal Disinformation with Regulatory Alignment

Detecting disinformation that blends manipulated text and images has become increasingly challenging, as AI tools make synthetic content easy to generate and disseminate. While most existing AI safety benchmarks focus on single modality misinformation (i.e., false content shared without intent to deceive), intentional multimodal disinformation, such as propaganda or conspiracy theories that imitate credible news, remains largely unaddressed. We introduce the Vision-Language Disinformation Detection Benchmark (VLDBench), the first large-scale resource supporting both unimodal (text-only) and multimodal (text + image) disinformation detection. VLDBench comprises approximately 62,000 labeled text-image pairs across 13 categories, curated from 58 news outlets. Using a semi-automated pipeline followed by expert review, 22 domain experts invested over 500 hours to produce high-quality annotations with substantial inter-annotator agreement. Evaluations of state-of-the-art Large Language Models (LLMs) and Vision-Language Models (VLMs) on VLDBench show that incorporating visual cues improves detection accuracy by 5 to 35 percentage points over text-only models. VLDBench provides data and code for evaluation, fine-tuning, and robustness testing to support disinformation analysis. Developed in alignment with AI governance frameworks (e.g., the MIT AI Risk Repository), VLDBench offers a principled foundation for advancing trustworthy disinformation detection in multimodal media. Project: https://vectorinstitute.github.io/VLDBench/ Dataset: https://huggingface.co/datasets/vector-institute/VLDBench Code: https://github.com/VectorInstitute/VLDBench

cs.CL↗

arXiv2025

Open-PMC-18M: A High-Fidelity Large Scale Medical Dataset for Multimodal Representation Learning

In biomedical vision-language modeling, datasets are typically mined from scientific literature, pairing compound figures with captions that are short, context-dependent, and oftern partially informative. Prior work on subfigure extraction has been limited in both dataset size and generalizability. In addition, no existing effort has incorporated rich medical context in image-text pairs. We revisit data curation as a foundational component of effective biomedical representation learning. Our data curation process integrates transformer-based subfigure detection, subcaption extraction, and contextual text enrichment derived from inline references. Our subfigure extraction model, trained on a corpus of 500,000 compound figures, achieves state-of-the-art performance on real and synthetic benchmarks. Using this process, we curate and release Open-PMC-18M, a large-scale high-fidelity biomedical dataset comprising 18 million image-text pairs, spanning radiology, microscopy, and visible light photography. We train vision-language models on our dataset and perform extensive evaluation on 6 retrieval and 19 zero-shot classification tasks across three major modalities. The models trained on our dataset set a new state-of-the-art results in medical representation learning. We release our dataset, models, and code to support reproducible benchmarks and further study into biomedical vision-language modeling and representation learning.

cs.CV↗

arXiv2025

When Can We Trust LLMs in Mental Health? Large-Scale Benchmarks for Reliable LLM Evaluation

Evaluating Large Language Models (LLMs) for mental health support is challenging due to the emotionally and cognitively complex nature of therapeutic dialogue. Existing benchmarks are limited in scale, reliability, often relying on synthetic or social media data, and lack frameworks to assess when automated judges can be trusted. To address the need for large-scale dialogue datasets and judge reliability assessment, we introduce two benchmarks that provide a framework for generation and evaluation. MentalBench-100k consolidates 10,000 one-turn conversations from three real scenarios datasets, each paired with nine LLM-generated responses, yielding 100,000 response pairs. MentalAlign-70k}reframes evaluation by comparing four high-performing LLM judges with human experts across 70,000 ratings on seven attributes, grouped into Cognitive Support Score (CSS) and Affective Resonance Score (ARS). We then employ the Affective Cognitive Agreement Framework, a statistical methodology using intraclass correlation coefficients (ICC) with confidence intervals to quantify agreement, consistency, and bias between LLM judges and human experts. Our analysis reveals systematic inflation by LLM judges, strong reliability for cognitive attributes such as guidance and informativeness, reduced precision for empathy, and some unreliability in safety and relevance. Our contributions establish new methodological and empirical foundations for reliable, large-scale evaluation of LLMs in mental health. We release the benchmarks and codes at: https://github.com/abeerbadawi/MentalBench/

cs.CL↗

arXiv2025

Automated Capability Evaluation of Foundation Models

Current evaluation frameworks for foundation models rely heavily on static, manually curated benchmarks, limiting their ability to capture the full breadth of model capabilities. This paper introduces Active learning for Capability Evaluation (ACE), a novel framework for scalable, automated, and fine-grained evaluation of foundation models. ACE leverages the knowledge embedded in powerful frontier models to decompose a domain into semantically meaningful capabilities and generates diverse evaluation tasks, significantly reducing human effort. In Mathematics, ACE generated 433 capabilities and 11,800 tasks, covering 94% of Wikipedia-defined skills in the domain while introducing novel, coherent ones. To maximize efficiency, ACE fits a capability model in latent semantic space, allowing reliable approximation of a subject model's performance by evaluating only a subset of capabilities via active learning. It reaches within 0.01 RMSE of exhaustive evaluation by evaluating less than half of capabilities. Compared to static datasets, ACE provides more balanced coverage and uncovers fine-grained differences that aggregate metrics fail to capture. Our results demonstrate that ACE provides a more complete and informative picture of model capabilities, which is essential for safe and well-informed deployment of foundation models.

cs.LG↗

arXiv2025

Improving Automatic Evaluation of Large Language Models (LLMs) in Biomedical Relation Extraction via LLMs-as-the-Judge

Large Language Models (LLMs) have demonstrated impressive performance in biomedical relation extraction, even in zero-shot scenarios. However, evaluating LLMs in this task remains challenging due to their ability to generate human-like text, often producing synonyms or abbreviations of gold-standard answers, making traditional automatic evaluation metrics unreliable. On the other hand, while human evaluation is more reliable, it is costly and time-consuming, making it impractical for real-world applications. This paper investigates the use of LLMs-as-the-Judge as an alternative evaluation method for biomedical relation extraction. We benchmark 8 LLMs as judges to evaluate the responses generated by 5 other LLMs across 3 biomedical relation extraction datasets. Unlike other text-generation tasks, we observe that LLM-based judges perform quite poorly (usually below 50% accuracy) in the biomedical relation extraction task. Our findings reveal that it happens mainly because relations extracted by LLMs do not adhere to any standard format. To address this, we propose structured output formatting for LLM-generated responses that helps LLM-Judges to improve their performance by about 15% (on average). We also introduce a domain adaptation technique to further enhance LLM-Judge performance by effectively transferring knowledge between datasets. We release both our human-annotated and LLM-annotated judgment data (36k samples in total) for public use here: https://github.com/tahmedge/llm_judge_biomedical_re.

cs.CL↗

arXiv2025

Position: Beyond Assistance -- Reimagining LLMs as Ethical and Adaptive Co-Creators in Mental Health Care

This position paper argues for a fundamental shift in how Large Language Models (LLMs) are integrated into the mental health care domain. We advocate for their role as co-creators rather than mere assistive tools. While LLMs have the potential to enhance accessibility, personalization, and crisis intervention, their adoption remains limited due to concerns about bias, evaluation, over-reliance, dehumanization, and regulatory uncertainties. To address these challenges, we propose two structured pathways: SAFE-i (Supportive, Adaptive, Fair, and Ethical Implementation) Guidelines for ethical and responsible deployment, and HAAS-e (Human-AI Alignment and Safety Evaluation) Framework for multidimensional, human-centered assessment. SAFE-i provides a blueprint for data governance, adaptive model engineering, and real-world integration, ensuring LLMs align with clinical and ethical standards. HAAS-e introduces evaluation metrics that go beyond technical accuracy to measure trustworthiness, empathy, cultural sensitivity, and actionability. We call for the adoption of these structured approaches to establish a responsible and scalable model for LLM-driven mental health support, ensuring that AI complements, rather than replaces, human expertise.

cs.HC↗

arXiv2025

Consistency-Guided Asynchronous Contrastive Tuning for Few-Shot Class-Incremental Tuning of Foundation Models

We propose Consistency-guided Asynchronous Contrastive Tuning (CoACT), a novel method for continuously tuning foundation models to learn new classes in few-shot settings. CoACT consists of three key components:(i) asynchronous contrastive tuning, which learns new classes by including LoRA modules in the pre-trained encoder while enforcing consistency between two asynchronous encoders; (ii) controlled fine-tuning, which facilitates effective tuning of a subset of the foundation model; and (iii) consistency-guided incremental tuning, which enforces additional regularization during later sessions to reduce forgetting of the learned classes. We evaluate our proposed solution on Few-Shot Class-Incremental Learning (FSCIL) as well as a new and more challenging setup called Few-Shot Class-Incremental Tuning (FSCIT), which facilitates the continual tuning of vision foundation models to learn new classes with only a few samples per class. Unlike traditional FSCIL, FSCIT does not require a large in-distribution base session for initial fully supervised training prior to the incremental few-shot sessions. We conduct extensive evaluations across 16 diverse datasets, demonstrating the effectiveness of CoACT in both FSCIL and FSCIT setups. CoACT outperforms existing methods by up to 5.02% in FSCIL and up to 12.51% in FSCIT for individual datasets, with an average improvement of 2.47%. Furthermore, CoACT exhibits reduced forgetting and enhanced robustness in low-shot experiments. Detailed ablation and sensitivity studies highlight the contribution of each component of CoACT. We make our code publicly available at https://github.com/ShuvenduRoy/CoACT-FSCIL.

cs.CV↗

arXiv2025

A Flexible Fairness Framework with Surrogate Loss Reweighting for Addressing Sociodemographic Disparities

This paper presents a new algorithmic fairness framework called $\boldsymbolα$-$\boldsymbolβ$ Fair Machine Learning ($\boldsymbolα$-$\boldsymbolβ$ FML), designed to optimize fairness levels across sociodemographic attributes. Our framework employs a new family of surrogate loss functions, paired with loss reweighting techniques, allowing precise control over fairness-accuracy trade-offs through tunable hyperparameters $\boldsymbolα$ and $\boldsymbolβ$. To efficiently solve the learning objective, we propose Parallel Stochastic Gradient Descent with Surrogate Loss (P-SGD-S) and establish convergence guarantees for both convex and nonconvex loss functions. Experimental results demonstrate that our framework improves overall accuracy while reducing fairness violations, offering a smooth trade-off between standard empirical risk minimization and strict minimax fairness. Results across multiple datasets confirm its adaptability, ensuring fairness improvements without excessive performance degradation.

cs.LG↗

arXiv2025

Advancing Medical Representation Learning Through High-Quality Data

Despite the growing scale of medical Vision-Language datasets, the impact of dataset quality on model performance remains under-explored. We introduce Open-PMC, a high-quality medical dataset from PubMed Central, containing 2.2 million image-text pairs, enriched with image modality annotations, subfigures, and summarized in-text references. Notably, the in-text references provide richer medical context, extending beyond the abstract information typically found in captions. Through extensive experiments, we benchmark Open-PMC against larger datasets across retrieval and zero-shot classification tasks. Our results show that dataset quality-not just size-drives significant performance gains. We complement our benchmark with an in-depth analysis of feature representation. Our findings highlight the crucial role of data curation quality in advancing multimodal medical AI. We release Open-PMC, along with the trained models and our codebase.

eess.IV↗

arXiv2025

Similarity-Aware Token Pruning: Your VLM but Faster

The computational demands of Vision Transformers (ViTs) and Vision-Language Models (VLMs) remain a significant challenge due to the quadratic complexity of self-attention. While token pruning offers a promising solution, existing methods often introduce training overhead or fail to adapt dynamically across layers. We present SAINT, a training-free token pruning framework that leverages token similarity and a graph-based formulation to dynamically optimize pruning rates and redundancy thresholds. Through systematic analysis, we identify a universal three-stage token evolution process (aligner-explorer-aggregator) in transformers, enabling aggressive pruning in early stages without sacrificing critical information. For ViTs, SAINT doubles the throughput of ViT-H/14 at 224px with only 0.6% accuracy loss on ImageNet-1K, surpassing the closest competitor by 0.8%. For VLMs, we apply SAINT in three modes: ViT-only, LLM-only, and hybrid. SAINT reduces LLaVA-13B's tokens by 75%, achieving latency comparable to LLaVA-7B with less than 1% performance loss across benchmarks. Our work establishes a unified, practical framework for efficient inference in ViTs and VLMs.

cs.CV↗