Exploring Multimodal Language Models for Sustainability Disclosure Extraction: A Comparative Study

Pneumothorax requires rapid recognition and accurate interpretation of chest X-rays (CXRs), particularly in acute settings where delays can have serious consequences. With the emergence of advanced image interpretation models capable of visual analysis, their diagnostic reliability in radiology practice remains to be determined. This study aimed to assess the diagnostic performance of three state-of-the-art systems in detecting pneumothorax using a large, well-annotated dataset. A total of 10,675 CXRs from the publicly available SIIM-ACR Pneumothorax Segmentation dataset were analyzed. Three multimodal models (GPT-4o, Gemini 2 Pro, and Claude 4 Sonnet) were evaluated using a uniform, image-based approach. Each model's binary outputs (presence: 1, absence: 0) were compared with reference results to determine accuracy, sensitivity, specificity, precision, and F1 scores. Additional subgroup analyses were conducted across pneumothorax size categories: small, medium, and large. Pairwise statistical comparisons were performed using McNemar's test. Sensitivity, specificity, and overall accuracy are reported with corresponding 95% confidence intervals. The prevalence of pneumothorax in the dataset was 22.3% (n = 2,379). All models demonstrated high specificity (above 0.90) but consistently low sensitivity (0.16-0.36). The best overall performance was observed with Gemini 2, which achieved an accuracy of 0.79 and specificity of 0.95, whereas Claude 4 showed greater sensitivity (0.20-0.34) across lesion-size categories. Diagnostic performance improved with increasing pneumothorax size, but smaller lesions remained difficult to identify. Pairwise comparisons confirmed statistically significant differences among all evaluated systems (P < 0.050). In this large-scale evaluation, the tested models exhibited strong reliability in identifying normal examinations but limited ability to detect subtle or small pneumothoraxes. Despite high specificity, low sensitivity limits the use of current Multimodal large language models as rule-out tools for pneumothorax. With continued refinement, these models may eventually support radiologists by improving workflow efficiency and diagnostic confidence. Automated systems capable of high specificity but low sensitivity should not be relied upon to exclude pneumothorax. However, they may serve as valuable assistants for confirming positive findings and prioritizing urgent cases in busy clinical workflows.

Zero shot learning brings new challenges to referring image segmentation, a typical application to find a segmentation mask given a referring expression. Most researches borrow generalization of multi-modal model, such as Contrastive Language Image Pre-training(CLIP), to realize zero shot inference. However, CLIP also suffers the limitation of insensitive to direction and local information, which is incapable of fine-grained region-text matching. To overcome this limitation, we propose a new framework, to leverage multimodal large language models (MLLM) into segmentation. Specifically, augmentations are considered from two aspects. One is to align local image information to comprehensive linguistic information, under the hypothesis that the textual description guiding local target object in image. Comprehensive linguistic information here are considered from local, global and multi-modal generated description. The other one is to enhance the robustness by blurring and captions from multi-modal model. Comprehensive experiments on three datasets prove the effectiveness of proposed approach.

Sentiment analysis is crucial in understanding user opinions, emotions, and attitudes across various domains. Traditional sentiment analysis methods rely primarily on textual data, limiting their ability to capture the full context of human expression, often including multimodal elements such as images, audio, and videos. Existing approaches struggle with ambiguity, sarcasm, and lack of contextual awareness, reducing accuracy and effectiveness. To address these limitations, we propose a novel framework called Sentiment Analysis using Machine Learning with Multimodal Models (SA-ML-MM), which integrates text, images, and audio inputs using large language models (LLMs) enhanced with multimodal capabilities. Our approach leverages ML-based feature extraction and fusion techniques to improve sentiment classification accuracy. The proposed framework is applied in social media analysis, customer feedback interpretation, and emotion recognition. Experimental results demonstrate that SA-ML-MM significantly outperforms traditional text-based models, achieving higher accuracy and robustness in sentiment prediction. By incorporating multiple data modalities, our approach effectively captures nuanced emotional expressions, making sentiment analysis more precise and reliable.

The integration of visual encoders and large language models (LLMs) has driven recent progress in multimodal large language models (MLLMs). However, the scarcity of high-quality instruction-tuning data for vision-language tasks remains a challenge. The current leading paradigm, such as LLaVA, relies on language-only GPT-4 to generate data, which requires pre-annotated image captions and detection bounding boxes, suffering from understanding image details. A practical solution to this problem would be to utilize the available multimodal large language models to generate instruction data for vision-language tasks. However, it's worth noting that the currently accessible MLLMs are not as powerful as their LLM counterparts, as they tend to produce inadequate responses and generate false information. As a solution for addressing the current issue, this paper proposes the Visual Instruction Generation and Correction (VIGC) framework that enables multimodal large language models to generate instruction-tuning data and progressively enhance its quality on-the-fly. Specifically, Visual Instruction Generation (VIG) guides the vision-language model to generate diverse instruction-tuning data. To ensure generation quality, Visual Instruction Correction (VIC) adopts an iterative update mechanism to correct any inaccuracies in data produced by VIG, effectively reducing the risk of hallucination. Leveraging the diverse, high-quality data generated by VIGC, we finetune mainstream models and validate data quality based on various evaluations. Experimental results demonstrate that VIGC not only compensates for the shortcomings of language-only data generation methods, but also effectively enhances the benchmark performance. The models, datasets, and code are available at https://opendatalab.github.io/VIGC

This study aimed to evaluate the diagnostic accuracy of multimodal large language models in classifying superior labial frenulum attachments from intraoral photographs using expert consensus as the reference standard. Five experts (two periodontists and three orthodontists) established the consensus standard by classifying frenulum attachments in 117 intraoral images as mucosal, gingival, papillary, and papilla penetrating. The same photographs were then presented to three multimodal large language models (ChatGPT 4o, Gemini 2.5 Pro, and Microsoft Copilot GPT-4), and their diagnostic performance was evaluated using accuracy, sensitivity, specificity, and F1 score. Reliability was assessed using Fleiss' and Cohen's Kappa, and diagnostic performances were compared using Cochran's Q test. Human raters demonstrated almost perfect agreement (κ = 0.838, p < 0.001), whereas large language models showed poor inter-model agreement (κ = -0.124, p < 0.001). ChatGPT achieved slight significant agreement with the consensus (κ = 0.114, p = 0.019), although its clinical relevance was negligible. Gemini (κ = 0.099) and Copilot (κ = 0.027) showed no significant agreement (p > 0.05). Copilot yielded the highest overall accuracy (46.2%), followed by Gemini (44.5%) and ChatGPT (35.0%). The performance of the large language models varied across frenulum types. Current multimodal large language models demonstrate inconsistent and clinically insufficient accuracy in the classification of superior labial frenulum attachments from photographs. Domain-specific training is essential before large language models can be considered reliable diagnostic tools in dentistry.

Multimodal Large Language Models (MLLMs) promise seamless integration of vision and language understanding. However, despite their strong performance, recent studies reveal that MLLMs often fail to effectively utilize visual information, frequently relying on textual cues instead. This survey provides a comprehensive analysis of the vision component in MLLMs, covering both application-level and architectural aspects. We investigate critical challenges such as weak spatial reasoning, poor fine-grained visual perception, and suboptimal fusion of visual and textual modalities. Additionally, we explore limitations in current vision encoders, benchmark inconsistencies, and their implications for downstream tasks. By synthesizing recent advancements, we highlight key research opportunities to enhance visual understanding, improve cross-modal alignment, and develop more robust and efficient MLLMs. Our observations emphasize the urgent need to elevate vision to an equal footing with language, paving the path for more reliable and perceptually aware multimodal models.

In this work, we introduce Mini-Gemini, a simple and effective framework enhancing multi-modality Vision Language Models (VLMs). Despite the advancements in VLMs facilitating basic visual dialog and reasoning, a performance gap persists compared to advanced models like GPT-4 and Gemini. We propose a novel approach to narrow the gap by mining the potential of VLMs for better performance across various cross-modal tasks. It tackles the following questions: (1) How can high-resolution visual tokens improve image understanding without lengthening the token sequence? (2) How to improve reasoning and generation abilities of VLM with high-quality data? (3) How to close the gap between open-source VLMs and proprietary models on reasoning-driven generation? In particular, to enhance visual tokens, we propose to utilize an additional visual encoder for high-resolution refinement without increasing the visual token count. We further construct a high-quality dataset that promotes precise image comprehension and reasoning-based generation, expanding the operational scope of current VLMs. In general, Mini-Gemini further mines the potential of VLMs and empowers current frameworks with image understanding, reasoning, and generation simultaneously. The proposed model supports a series of dense and MoE Large Language Models (LLMs) from 2B to 34B, which achieve leading performance in several zero-shot benchmarks and even surpasses the developed private models. It is demonstrated to attain 80.6% accuracy on the MMB benchmark (+5.4 vs Gemini Pro) and 74.1% on TextVQA (+4.6 vs LLaVA-NeXT), achieving leading performance in several zero-shot benchmarks and even surpasses the developed private models. Furthermore, Mini-Gemini is proven to improve consistently with stronger LLM, visual encoder, and data in experiments.

Background: Multimodal large language models (LLMs) are increasingly being explored as clinical support tools, yet their capacity for orthodontic biomechanical reasoning has not been systematically evaluated. This retrospective study assessed their ability to analyze treatment mechanics and explored their potential role in supporting orthodontic decision-making. Methods: Five publicly available models (GPT-o3, Claude 3.7 Sonnet, Gemini 2.5 Pro, GPT-4.0, and Grok) analyzed 56 standardized intraoral photographs illustrating a diverse range of active orthodontic force systems commonly encountered in clinical practice. Three experienced orthodontists independently scored the outputs across four domains—observation, interpretation, biomechanics, and confidence—using a 5-point scale. Inter-rater agreement and consistency were assessed, and statistical comparisons were made between models. Results: GPT-o3 achieved the highest composite score (3.34/5.00; 66.8%), significantly outperforming all other models. The performance ranking was followed by Claude (57.8%), Gemini (52.6%), GPT-4.0 (48.8%), and Grok (38.8%). Inter-rater reliability among the expert evaluators was excellent, with ICC values ranging from 0.786 (Confidence Evaluation) to 0.802 (Observation). Model self-reported confidence showed poor calibration against expert-rated output quality. Conclusions: Multimodal LLMs show emerging potential for assisting orthodontic biomechanical assessment. With expert-guided validation, these models may contribute meaningfully to clinical decision support across diverse biomechanical scenarios encountered in routine orthodontic care.

In maritime surveillance, UAV-based vessel detection is essential for ensuring security and safety at sea. However, limited and non-diverse annotated data often restrict model performance in complex maritime environments. This study introduces a novel data augmentation pipeline using multimodal generative models to enhance training datasets with realistic synthetic images. Scene descriptions are automatically generated from UAV imagery using Gemma, a lightweight multimodal language model, and then used to guide FLUX, a text-to-image diffusion model, in creating diverse vessel-centric scenes under varying environmental conditions. A hybrid annotation strategy combines YOLO-World for initial object proposals with manual refinement to ensure label accuracy. The augmented dataset is integrated with the original data to train a vessel detection model. Experiments on the VESSELImg benchmark demonstrate that the proposed approach improves the YOLOv11 detector’s mean average precision (mAP) from 0.775 to 0.805 at IoU thresholds of 0.50:0.95. These results validate the effectiveness of combining multimodal diffusion and language models for domain-specific data synthesis, offering improved generalization and robustness in UAV-based maritime vessel detection.

The increasing complexity of STEM learning tasks and the scale of digital education environments have exposed fundamental limitations in traditional automated feedback systems. Most existing approaches rely on unimodal inputs or rule-based logic, providing surface-level feedback that fails to capture underlying learner misconceptions and reasoning processes. This study proposes and evaluates a multimodal large language model–driven diagnostic feedback framework designed to deliver accurate, explainable, and instructionally aligned feedback in STEM learning platforms. The framework integrates heterogeneous learner data, including text responses, symbolic mathematics, diagrams, code submissions, and interaction traces, through modality-specific encoders and attention-based fusion strategies. Diagnostic reasoning is performed using a multimodal large language model constrained by curricular objectives and enhanced with explainability mechanisms such as rationale tracing and attention visualization. Empirical evaluation across mathematics, physics, and computer science tasks demonstrates significant improvements over baseline systems in diagnostic accuracy, learning gains, error correction rates, learner engagement, and trust. The findings indicate that multimodal LLM-driven diagnostic feedback can operationalize formative assessment principles at scale, offering a robust pathway toward more transparent, adaptive, and pedagogically meaningful AI-supported learning in STEM education.

Pests and diseases significantly impact the growth and development of crops. When attempting to precisely identify disease characteristics in crop images through dialogue, existing multimodal models face numerous challenges, often leading to misinterpretation and incorrect feedback regarding disease information. This paper proposed a large language model for multimodal identification of crop diseases and pests, which can be called LLMI-CDP. It builds up on the VisualGLM model and introduces improvements to achieve precise identification of agricultural crop disease and pest images, along with providing professional recommendations for relevant preventive measures. The use of Low-Rank Adaptation (LoRA) technology, which adjusts the weights of pre-trained models, achieves significant performance improvements with a minimal increase in parameters. This ensures the precise capture and efficient identification of crop pest and disease characteristics, greatly enhancing the model’s application flexibility and accuracy in the field of pest and disease recognition. Simultaneously, the model incorporates the Q-Former framework for effective modal alignment between language models and image features. Through this approach, the LLMI-CDP model is able to more deeply understand and process the complex relationships between language and visual information, further enhancing its performance in multimodal recognition tasks. Experiments are carried out in the homemade datasets, The results demonstrate that the LLMI-CDP model surpasses five leading multimodal large language models in relevant evaluation metrics, confirming its outstanding performance in Chinese multimodal dialogues related to agriculture.

Accurate tinnitus diagnosis is crucial for enabling timely therapeutic intervention and longitudinal treatment monitoring. While non-invasive neuroimaging modalities-particularly electroencephalography (EEG) with millisecond temporal resolution and functional magnetic resonance imaging (fMRI) with millimeter spatial resolution- provide complementary neural features, existing diagnostic approaches remain constrained to unimodal analysis of EEG or fMRI data, inherently limiting diagnostic precision and clinical generalizability. This paper introduces TinnitusLLM, the first multimodal large language model (LLM) framework that synergistically integrates EEG and fMRI features for tinnitus diagnosis. To enable LLM-based interpretation of neural signals, this framework integrates three key components: (1) a neuroinspired positional encoding mechanism that injects neurophysiological priors into the embedding space, enabling neurologically grounded, dynamic positional mapping of EEG and fMRI tokens; (2) multimodal autoregressive pretraining on more than 500 hours of EEG and 250 hours of fMRI data to learn causally informed predictive representations; and (3) fine-tuning with a cross-modal, subject-invariant adversarial learning strategy that enforces subject-independent constraints in the shared cross-modal feature space, thereby substantially improving diagnostic robustness across subjects. We validate TinnitusLLM through comprehensive experiments on a rigorously collected multimodal dataset containing 20 participants. Quantitative evaluations demonstrate that TinnitusLLM achieves superior cross-subject diagnostic accuracy compared to the state-of-the-art baseline methods. These results underscore TinnitusLLM's potential as a clinically viable framework for objective tinnitus assessment through multimodal neural decoding.

Background To systematically compare the diagnostic accuracy of five contemporary multimodal large language models (MLLMs: Gemini-2.5-Pro, Grok-4, GPT-4o, GPT-5, and Qwen-VL-Max) in evaluating the Mayo Endoscopic Score (MES) for ulcerative colitis (UC), and to explore their consistency and performance across various intestinal segments and MES categories. Methods 402 authentic endoscopic images from patients with UC were collected, covering the entire colon location from the ileocecal region to the rectum. Three experienced inflammatory bowel disease (IBD) experts independently reviewed and unanimously graded these images and finally 283 images with consensus on grades. These images were among the second stage of research and the grades as the reference standard. These images were randomly presented to MLLMs and two senior IBD physicians without specifying the intestinal segment, and then randomly presented to MLLMs with segmental information before grade. Model and physician performance were compared, and stratified analyses were conducted by intestinal segment and MES grade. Results The diagnostic accuracies (Acc) of the two IBD physicians were 81.6% and 78.4%, respectively, with strong inter-observer agreement (κ = 0.692). Among these MLLMs, GPT-5 achieved the highest overall performance (F1: GPT-5 0.720 &amp;gt; GPT-4o 0.602 &amp;gt; Gemini-2.5-Pro 0.480 &amp;gt; Grok-4 0.415 &amp;gt; Qwen-VL-Max 0.338), and its diagnostic accuracy was comparable to that of human physicians (GPT-5 Acc 71.7% vs. Senior Physician 2 Acc 78.4%, P = 0.068). The other models exhibited significantly poor diagnostic performance compared with experienced IBD physicians (all P &amp;lt; 0.001). The sigmoid colon was the most accurately assessed region (mean F1 across models 0.682), whereas the rectum and ileocecal region remained the most challenging (0.447 and 0.493, respectively). The provision of segmental information significantly enhanced the performance of the poor-performing models. Moreover, both the models and human physicians showed the lowest accuracy at MES=1(physicians mean Acc≈60.3 %, models mean Acc≈39.4 %), indicating that the mild-activity grade remains the most challenging to classify due to its inherent subjectivity. Conclusion GPT-5 demonstrated diagnostic performance comparable to that of senior IBD physicians in MES grading, whereas other MLLMs require further optimization. In terms of intestinal segments, the rectum and ileocecal region, and in terms of disease severity, mild activity (MES = 1), represented common challenges for both physicians and models. Future efforts should be focused on targeted training for these challenging segments and grades, and integrating additional clinical multimodal data to advance the clinical implementation of intelligent endoscopic assessment.

Estimating urban flooding depth by integrating multimodal image-text data: A segment-level direct preference optimization-based multimodal large language model

This study aimed to assess how multimodal large language models (MLLM) diagnose and differentiate Pentosan Polysulfate (PPS) Maculopathy from other phenotypic mimics. A retrospective review of clinical records and multimodal retinal imaging was conducted with patients from the Shiley Eye Institute and Casey Eye Institute. Four MLLMs (ChatGPT-4o, Claude 3.5 Sonnet, Google Gemini 1.5 Pro, Perplexity Llama 3.1 Sonar/Default) along with human retinal specialists answered prompts based on retinal imaging and demographic data. Performance was evaluated using accuracy, sensitivity and specificity estimates. The study included 126 eyes from 63 patients, with 36 eyes with PPS maculopathy, 50 eyes with Stargardt disease, and 40 eyes with PRPH2-associated multifocal pattern dystrophy. MLLMs showed improved accuracy and sensitivity when answer choices were restricted, with ChatGPT consistently performing best when all imaging modalities were prompted together. The inclusion of demographic data further enhanced performance in prompts with limited answer choices. Human retinal specialist evaluations aligned with MLLM performance trends and also improved with demographic data. While MLLMs show diagnostic potential, further refinement is needed before clinical implementation. These findings highlight the importance of prompt design and demographic data to optimize MLLM performance with retinal imaging modalities.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-025-29299-2.