True intelligence hinges on the ability to uncover and leverage hidden causal relations. Despite significant progress in AI and computer vision (CV), there remains a lack of benchmarks for assessing models' abilities to infer latent causality from complex visual data. In this paper, we introduce \textsc{\textbf{Causal3D}}, a novel and comprehensive benchmark that integrates structured data (tables) with corresponding visual representations (images) to evaluate causal reasoning. Designed within a systematic framework, Causal3D comprises 19 3D-scene datasets capturing diverse causal relations, views, and backgrounds, enabling evaluations across scenes of varying complexity. We assess multiple state-of-the-art methods, including classical causal discovery, causal representation learning, and large/vision-language models (LLMs/VLMs). Our experiments show that as causal structures grow more complex without prior knowledge, performance declines significantly, highlighting the challenges even advanced methods face in complex causal scenarios. Causal3D serves as a vital resource for advancing causal reasoning in CV and fostering trustworthy AI in critical domains.

This research investigates the integration of emotional diversity into Large Language Models (LLMs) to enhance collective intelligence. Inspired by the human wisdom of crowds phenomenon, where group decisions often outperform individual judgments, we fine-tuned the DarkIdol-Llama-3.1-8B model using Google's GoEmotions dataset and Low-Rank Adaptation (LoRA) to simulate emotionally diverse responses. Evaluating the model on a distance estimation task between Fargo, ND, and Seattle, WA, across 15,064 unique persona configurations, we analyzed how emotional states and social attributes influence decision-making. Our findings demonstrate that emotional integration shapes response patterns while maintaining acceptable prediction accuracy, revealing its potential to enhance artificial collective intelligence. This study provides valuable insights into the interplay of emotional diversity and decision-making in LLMs, suggesting pathways for creating emotionally aware AI systems that balance emotional depth with analytical precision.

Generative AI (GenAI) represents a transformative leap in artificial intelligence, using advanced models such as Transformers, GPT-4, and Gemini to generate human-like content in multiple modalities [1],[2]. Unlike traditional AI models that focus on classification or predictive tasks using predefined patterns, GenAI utilizes deep learning architectures like Transformer-based Large Language Models (LLMs)[2] to create text, images, code, and audio. The most prominent GenAI models include GPT-4 for advanced text generation[1] and Google Gemini for multimodal AI applications[2]. These models leverage massive data sets and training methodologies such as Reinforcement Learning with Human Feedback (RLHF)[3] and retrieval-augmented generation (RAG)[4] to improve their contextual understanding and adaptability. These AI models, trained on large-scale data, can understand context, generate creative outputs, automate workflows, and drive innovation across industries. GenAI is transforming fields such as healthcare (AI-assisted diagnosis and drug discovery[5]), finance (automated risk analysis and fraud detection[6]), customer support (intelligent virtual assistants[7]), and software development (AI-driven code generation[8]). The emergence of multimodal AI which enables models to process and generate text, images, and audio simultaneously is further unlocking new possibilities in automation, personalization, and decision-making.

Large Language Models (LLMs) require instruction fine-tuning to perform different downstream tasks. However, the instruction fine-tuning phase still demands significant computational resources and labeled data, lacking a paradigm that can improve model performance without additional computational power and data. Model merging aims to enhance performance by combining the parameters of different models, but the lack of a clear optimization direction during the merging process does not always guarantee improved performance. In this paper, we attempt to provide a clear optimization direction for model merging. We first validate the effectiveness of the model extrapolation method during the instruction fine-tuning phase. Then, we propose Extrapolation Merging, a paradigm that can continue improving model performance without requiring extra computational resources or data. Using the extrapolation method, we provide a clear direction for model merging, achieving local optimization search, and consequently enhancing the merged model's performance. We conduct experiments on seven different tasks, and the results show that our method can consistently improve the model's performance after fine-tuning.

This work investigates the capabilities of large language models (LLMs) in detecting and understanding human emotions through text. Drawing upon emotion models from psychology, we adopt an interdisciplinary perspective that integrates computational and affective sciences insights. The main goal is to assess how accurately they can identify emotions expressed in textual interactions and compare different models on this specific task. This research contributes to broader efforts to enhance human-computer interaction, making artificial intelligence technologies more responsive and sensitive to users' emotional nuances. By employing a methodology that involves comparisons with a state-of-the-art model on the GoEmotions dataset, we aim to gauge LLMs' effectiveness as a system for emotional analysis, paving the way for potential applications in various fields that require a nuanced understanding of human language.

While Large Language Models (LLMs) are fundamentally next-token prediction systems, their practical applications extend far beyond this basic function. From natural language processing and text generation to conversational assistants and software use, LLMs have numerous use-cases, and have already acquired a significant degree of enterprise adoption. To evaluate such models, static evaluation datasets, consisting of a set of prompts and their corresponding ground truths, are often used to benchmark the efficacy of the model for a particular task. In this paper, we provide the basis for a more comprehensive evaluation framework, based upon a traditional game and tool-based architecture that enables a more overarching measurement of a model's capabilities. For simplicity, we provide a generalized foundation that can be extended, without significant alteration, to numerous scenarios, from specific use cases such as supply chain management or financial reasoning, to abstract measurements such as ethics or safety.

Existing research has primarily focused on static analyses that assess the disparities in the existing knowledge and capabilities of LLMs across languages. However, LLMs are continuously evolving, acquiring new knowledge to generate up-to-date, domain-specific responses. Investigating linguistic inequalities within this dynamic process is, therefore, also essential. In this paper, we explore inequalities in new knowledge learning by LLMs across different languages and four key dimensions: effectiveness, transferability, prioritization, and robustness. Through extensive experiments under two settings (in-context learning and fine-tuning) using both proprietary and open-source models, we demonstrate that low-resource languages consistently face disadvantages across all four dimensions. By shedding light on these disparities, we aim to raise awareness of linguistic inequities in LLMs' new knowledge learning, fostering the development of more inclusive and equitable future LLMs. This transformation is both inevitable and global in scale. One notable example is ChatGPT, which, as of December 2024, serves 300 million weekly active users worldwide (6, 7). Given such widespread adoption, it is crucial to study fairness in multilingual environments to ensure that users of different languages can benefit equally from these systems (9). Existing research on multilingual equality in LLMs primarily focuses on static analyses that evaluate disparities in the knowledge and capabilities of LLMs across different languages (10, 11, 12, 13, 14, 15, 16, 17). Some studies, for example, have examined the amount of factual knowledge encoded in different languages and revealed significant variations. In particular, they reveal that knowledge available in low-resource languages remains limited due to the lack of pre-training data in these languages (18, 19, 20). These studies have significantly advanced our understanding of the extent and nature of multilingual inequalities in LLMs' existing knowledge and capabilities. However, we still lack an understanding of inequalities in the process of acquiring new knowledge, an evolving perspective in research on LLMs. Learning new knowledge is crucial for LLMs, as illustrated in Figure 1a. On the one hand, general-purpose LLMs are pre-trained on static datasets that were collected prior to training and may not include real-time or recent information. As a result, these models do not possess new knowledge, and their knowledge base can quickly become outdated.

Large language models (LLMs) have become important tools in solving biological problems, offering improvements in accuracy and adaptability over conventional methods. Several benchmarks have been proposed to evaluate the performance of these LLMs. However, current benchmarks can hardly evaluate the performance of these models across diverse tasks effectively. In this paper, we introduce a comprehensive prompting-based benchmarking framework, termed Bio-benchmark, which includes 30 key bioinformatics tasks covering areas such as proteins, RNA, drugs, electronic health records, and traditional Chinese medicine. Using this benchmark, we evaluate six mainstream LLMs, including GPT-4o and Llama-3.1-70b, etc., using 0-shot and few-shot Chain-of-Thought (CoT) settings without fine-tuning to reveal their intrinsic capabilities. To improve the efficiency of our evaluations, we demonstrate BioFinder, a new tool for extracting answers from LLM responses, which increases extraction accuracy by round 30% compared to existing methods. Our benchmark results show the biological tasks suitable for current LLMs and identify specific areas requiring enhancement. Furthermore, we propose targeted prompt engineering strategies for optimizing LLM performance in these contexts. Based on these findings, we provide recommendations for the development of more robust LLMs tailored for various biological applications. This work offers a comprehensive evaluation framework and robust tools to support the application of LLMs in bioinformatics.

Recently, Multimodal Large Language Models (MLLMs) have gained significant attention for their remarkable ability to process and analyze non-textual data, such as images, videos, and audio. Notably, several adaptations of general-domain MLLMs to the medical field have been explored, including LLaVA-Med. However, these medical adaptations remain insufficiently advanced in understanding and interpreting retinal images. In contrast, medical experts emphasize the importance of quantitative analyses for disease detection and interpretation. This underscores a gap between general-domain and medical-domain MLLMs: while general-domain MLLMs excel in broad applications, they lack the specialized knowledge necessary for precise diagnostic and interpretative tasks in the medical field. To address these challenges, we introduce \textit{RetinalGPT}, a multimodal conversational assistant for clinically preferred quantitative analysis of retinal images. Specifically, we achieve this by compiling a large retinal image dataset, developing a novel data pipeline, and employing customized visual instruction tuning to enhance both retinal analysis and enrich medical knowledge. In particular, RetinalGPT outperforms MLLM in the generic domain by a large margin in the diagnosis of retinal diseases in 8 benchmark retinal datasets. Beyond disease diagnosis, RetinalGPT features quantitative analyses and lesion localization, representing a pioneering step in leveraging LLMs for an interpretable and end-to-end clinical research framework. The code is available at https://github.com/Retinal-Research/RetinalGPT

Understanding and reasoning over non-speech sounds and music are crucial for both humans and AI agents to interact effectively with their environments. In this paper, we introduce Audio Flamingo 2 (AF2), an Audio-Language Model (ALM) with advanced audio understanding and reasoning capabilities. AF2 leverages (i) a custom CLAP model, (ii) synthetic Audio QA data for fine-grained audio reasoning, and (iii) a multi-stage curriculum learning strategy. AF2 achieves state-of-the-art performance with only a 3B parameter small language model, surpassing large open-source and proprietary models across over 20 benchmarks. Next, for the first time, we extend audio understanding to long audio segments (30 secs to 5 mins) and propose LongAudio, a large and novel dataset for training ALMs on long audio captioning and question-answering tasks. Fine-tuning AF2 on LongAudio leads to exceptional performance on our proposed LongAudioBench, an expert annotated benchmark for evaluating ALMs on long audio understanding capabilities. We conduct extensive ablation studies to confirm the efficacy of our approach. Project Website: https://research.nvidia.com/labs/adlr/AF2/.