Mini-Gemini: Mining the Potential of Multi-Modality Vision Language Models.

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.

Recently, Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) have shown promise in instruction following and image understanding. While these models are powerful, they have not yet been developed to comprehend the more challenging 3D geometric and physical scenes, especially when it comes to the sparse outdoor LiDAR data. In this paper, we introduce LiDAR-LLM, which takes raw LiDAR data as input and harnesses the remarkable reasoning capabilities of LLMs to gain a comprehensive understanding of outdoor 3D scenes. The central insight of our LiDAR-LLM is the reformulation of 3D outdoor scene cognition as a language modeling problem, encompassing tasks such as 3D captioning, 3D grounding, 3D question answering, etc. Specifically, due to the scarcity of 3D LiDAR-text pairing data, we introduce a three-stage training strategy and generate relevant datasets, progressively aligning the 3D modality with the language embedding of LLM. Furthermore, we design a Position-Aware Transformer (PAT) to connect the 3D encoder with the LLM, which effectively bridges the modality gap and enhances the LLM's spatial orientation comprehension of visual features. Our experiments demonstrate that LiDAR-LLM effectively comprehends a wide range of instructions related to 3D scenes, achieving a 40.9 BLEU-1 score on the 3D captioning dataset, a Grounded Captioning accuracy of 63.1%, and a BEV mIoU of 14.3%.

AI-enabled language models (LMs) to large language models (LLMs) and multimodal large language models (MLLMs) in drug discovery and development

In recent years, generative artificial intelligence (AI), particularly large language models (LLMs) and their multimodal counterparts, multimodal large language models, including vision language models, have generated considerable interest in the global AI discourse. LLMs, or pre-trained language models (such as ChatGPT, Med-PaLM, LLaMA), are neural network architectures trained on extensive text data, excelling in language comprehension and generation. Multimodal LLMs, a subset of foundation models, are trained on multimodal data sets, integrating text with another modality, such as images, to learn universal representations akin to human cognition better. This versatility enables them to excel in tasks like chatbots, translation, and creative writing while facilitating knowledge sharing through transfer learning, federated learning, and synthetic data creation. Several of these models can have potentially appealing applications in the medical domain, including, but not limited to, enhancing patient care by processing patient data; summarizing reports and relevant literature; providing diagnostic, treatment, and follow-up recommendations; and ancillary tasks like coding and billing. As radiologists enter this promising but uncharted territory, it is imperative for them to be familiar with the basic terminology and processes of LLMs. Herein, we present an overview of the LLMs and their potential applications and challenges in the imaging domain.

The emergence of large language models (LLMs) has transformed research and practice across a wide range of domains. Within the computing education research (CER) domain, LLMs have garnered significant attention, particularly in the context of learning programming. Much of the work on LLMs in CER, however, has focused on applying and evaluating proprietary models. In this article, we evaluate the efficiency of open-source LLMs in generating high-quality feedback for programming assignments and judging the quality of programming feedback, contrasting the results with proprietary models. Our evaluations on a dataset of students' submissions to introductory Python programming exercises suggest that state-of-the-art open-source LLMs are nearly on par with proprietary models in both generating and assessing programming feedback. Additionally, we demonstrate the efficiency of smaller LLMs in these tasks and highlight the wide range of LLMs accessible, even for free, to educators and practitioners.

Recently, Referring Image Segmentation (RIS) frameworks that pair the Multimodal Large Language Model (MLLM) with the Segment Anything Model (SAM) have achieved impressive results. However, adapting MLLM to segmentation is computationally intensive, primarily due to visual token redundancy. We observe that traditional patch-wise visual projectors struggle to strike a balance between reducing the number of visual tokens and preserving semantic clarity, often retaining overly long token sequences to avoid performance drops. Inspired by text tokenizers, we propose a novel semantic visual projector that leverages semantic superpixels generated by SAM to identify “visual words” in an image. By compressing and projecting semantic superpixels as visual tokens, our approach adaptively shortens the token sequence according to scene complexity while minimizing semantic loss in compression. To mitigate loss of information, we propose a semantic superpixel positional embedding to strengthen MLLM’s awareness of superpixel geometry and position, alongside a semantic superpixel aggregator to preserve both fine-grained details inside superpixels and global context outside. Experiments show that our method cuts visual tokens by \(\sim\) 93% without compromising performance, notably speeding up MLLM training and inference, and outperforming existing compressive visual projectors on RIS.

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.

Addressing the gap in understanding visual comprehension in Large Language Models (LLMs), we designed a challenge-response study, subjecting Google Bard to 64 visual tasks, spanning categories like "Visual Situational Reasoning" and "Next Scene Prediction." Previous models, such as GPT4, leaned heavily on optical character recognition tools like Tesseract, whereas Bard, akin to Google Lens and Visual API, employs deep learning techniques for visual text recognition. However, our findings spotlight Bard's limitations: while proficient in solving visual CAPTCHAs that stump ChatGPT, it falters in recreating visual elements like ASCII art or analyzing Tic Tac Toe grids, suggesting an over-reliance on educated visual guesses. The prediction problem based on visual inputs appears particularly challenging with no commonsense guesses for next scene forecasting based on current “next- token” multimodal models. This study provides experimental insights into the current capacities and areas for improvement in multimodal LLMs.

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.

The Ministry of Food and Drug Safety (MFDS) of the Republic of Korea, similar to the United States Food and Drug Administration and the United Kingdom's Medicines and Healthcare products Regulatory Agency, issued specific regulatory guidelines on January 24, 2025, for approving generative artificial intelligence (AI) technologies as medical devices [1].Although these guidelines use the term 'generative AI,' they predominantly focus on the approval of AI software tools based on large language models (LLMs) and large multimodal models (LMMs) [2], the latter of which can process various types of input data, such as texts, images, videos, audio, and bio-signals.Unlike the broader guidelines for approving AI models as medical devices, specific regulatory guidelines for LLMs/LMMs have arguably not yet been proposed in other countries.

Purpose This study aims to explore the integration of large language models (LLMs) and vision-language models (VLMs) in robotics, highlighting their potential benefits and the safety challenges they introduce, including robustness issues, adversarial vulnerabilities, privacy concerns and ethical implications. Design/methodology/approach This survey conducts a comprehensive analysis of the safety risks associated with LLM- and VLM-powered robotic systems. The authors review existing literature, analyze key challenges, evaluate current mitigation strategies and propose future research directions. Findings The study identifies that ensuring the safety of LLM-/VLM-driven robots requires a multi-faceted approach. While current mitigation strategies address certain risks, gaps remain in real-time monitoring, adversarial robustness and ethical safeguards. Originality/value This study offers a structured and comprehensive overview of the safety challenges in LLM-/VLM-driven robotics. It contributes to ongoing discussions by integrating technical, ethical and regulatory perspectives to guide future advancements in safe and responsible artificial intelligence-driven robotics.

To evaluate the impact of constraining proprietary and open large language models (LLMs) to structured outputs in processing radiology request forms (RRF). We evaluated five LLMs-two proprietary (GPT-5-Thinking, Gemini 2.5 Pro) and three open (Qwen3-235B-A22B-Thinking, gpt-oss-120b, medgemma-27b-it)-on 100 RRFs (50 computed tomography, 50 magnetic resonance imaging). Each model processed cases with and without constraints to structured outputs. Endpoints included accuracy for modality, anatomical region, contrast phase, urgency, "all correct" (all four categories correct), and "indication improved" (clarity of rewritten text). Outputs were evaluated against a reference standard defined by board-certified radiologists and compared with two radiology residents (first-year and third-year). Accuracies with 95% confidence intervals were calculated. Constraining to structured outputs had model-dependent effects: it improved Gemini 2.5 Pro (all correct: from 53.0% [43.3-62.5] to 66.0% [56.3-74.5]) but reduced GPT-5-Thinking accuracy (from 76.0% [66.8-83.3] to 53.0% [43.3-62.5]), with minimal influence on open models. Both proprietary LLMs outperformed the best open models (up to 41.0% [31.9-50.8]). All LLMs exceeded the unassisted first-year residents' performance (19.0% [12.5-27.8]). LLM assistance improved first-year residents' accuracy to 65.0% [55.3-73.6], approaching the third-year residents' performance (80.0% [71.1-86.7]), who performed comparably to the best LLMs. Across models, performance was highest for modality and anatomical region, and lowest for urgency. Indication reformulation was judged clearer in > 90% of cases across all models without hallucinations. Constraining to structured outputs exerted model-specific effects. Proprietary LLMs achieved the highest accuracy in RRF-based protocol selection and improved first-year resident performance to an experienced-resident level. LLMs may serve as valuable decision-support tools for radiology workflow. Constraining LLMs to structured outputs produced divergent, model-specific effects in radiology protocol selection-improving Gemini 2.5 Pro, reducing GPT-5-Thinking, and minimally affecting open models-highlighting the need for model-specific prompting strategies before adopting LLMs in radiology decision support. Structured output constraints affect LLM performance differently. Gemini 2.5 Pro benefits from structured prompting, while GPT-5-Thinking declines. Open-weight models show minimal impact from output constraining. Proprietary models outperform open models in radiology protocol selection.

A chief goal of artificial intelligence is to build machines that think like people. Yet it has been argued that deep neural network architectures fail to accomplish this. Researchers have asserted these models’ limitations in the domains of causal reasoning, intuitive physics and intuitive psychology. Yet recent advancements, namely the rise of large language models, particularly those designed for visual processing, have rekindled interest in the potential to emulate human-like cognitive abilities. This paper evaluates the current state of vision-based large language models in the domains of intuitive physics, causal reasoning and intuitive psychology. Through a series of controlled experiments, we investigate the extent to which these modern models grasp complex physical interactions, causal relationships and intuitive understanding of others’ preferences. Our findings reveal that, while some of these models demonstrate a notable proficiency in processing and interpreting visual data, they still fall short of human capabilities in these areas. Our results emphasize the need for integrating more robust mechanisms for understanding causality, physical dynamics and social cognition into modern-day, vision-based language models, and point out the importance of cognitively inspired benchmarks.

Handling long tail corner cases is a major challenge faced by autonomous vehicles (AVs). While large language models (LLMs) hold great potentials to handle the corner cases with excellent generalization and explanation capabilities and received increasing research interest on application to autonomous driving, there are still technical barriers to be tackled, such as strict model performance and huge computing resource requirements of LLMs, which are difficult to be met locally at AVs. In this paper, we investigate a new approach of applying remote or edge LLMs to support autonomous driving. With this approach connected autonomous vehicles (CAVs) send driving assistance requests to the LLMs. LLMs deployed at the edge of the networks or remote clouds process the requests and generate driving assistance instructions for the CAVs. A key issue for such LLM assisted driving system is the assessment of LLMs on their understanding of driving theory and skills, ensuring they are qualified to undertake safety critical driving assistance tasks for CAVs. As there is no published work on assessing LLM of driving theory and skills, we design and run driving theory tests for several proprietary LLM models (OpenAI GPT models, Baidu Ernie and Ali QWen) and open-source LLM models (Tsinghua MiniCPM-2B and MiniCPM-Llama3-V2.5) with more than 500 multiple-choices theory test questions. These questions are close to the official UK driving theory test ones. Model accuracy, cost and processing latency are measured from the experiments. Experiment results show that while model GPT-4 passes the test with improved domain knowledge and Ernie has an accuracy of 85% (just below the 86% passing threshold), other LLM models including GPT-3.5 fail the test. For the test questions with images, the multimodal model GPT4-o has an excellent accuracy result of 96%, and the MiniCPM-Llama3-V2.5 achieves an accuracy of 76%. While GPT-4 holds stronger potential for CAV driving assistance applications, the cost of using model GPT4 is much higher, almost 50 times of that of using GPT3.5. The results can help make decision on the use of the existing LLMs for CAV applications and balancing on the model performance and cost.