Multimodal Large Language Models (MLLMs) have become increasingly important due to their state-of-the-art performance and ability to integrate multiple data modalities, such as text, images, and audio, to perform complex tasks with high accuracy. This paper presents a comprehensive survey on personalized multimodal large language models, focusing on their architecture, training methods, and applications. We propose an intuitive taxonomy for categorizing the techniques used to personalize MLLMs to individual users, and discuss the techniques accordingly. Furthermore, we discuss how such techniques can be combined or adapted when appropriate, highlighting their advantages and underlying rationale. We also provide a succinct summary of personalization tasks investigated in existing research, along with the evaluation metrics commonly used. Additionally, we summarize the datasets that are useful for benchmarking personalized MLLMs. Finally, we outline critical open challenges. This survey aims to serve as a valuable resource for researchers and practitioners seeking to understand and advance the development of personalized multimodal large language models.

In this paper, we introduce ChatCite, a novel method leveraging large language models (LLMs) for generating comparative literature summaries. The ability to summarize research papers with a focus on key comparisons between studies is an essential task in academic research. Existing summarization models, while effective at generating concise summaries, fail to provide deep comparative insights. ChatCite addresses this limitation by incorporating a multi-step reasoning mechanism that extracts critical elements from papers, incrementally builds a comparative summary, and refines the output through a reflective memory process. We evaluate ChatCite on a custom dataset, CompLit-LongContext, consisting of 1000 research papers with annotated comparative summaries. Experimental results show that ChatCite outperforms several baseline methods, including GPT-4, BART, T5, and CoT, across various automatic evaluation metrics such as ROUGE and the newly proposed G-Score. Human evaluation further confirms that ChatCite generates more coherent, insightful, and fluent summaries compared to these baseline models. Our method provides a significant advancement in automatic literature review generation, offering researchers a powerful tool for efficiently comparing and synthesizing scientific research.

Although large language models (LLMs) have demonstrated their effectiveness in a wide range of applications, they have also been observed to perpetuate unwanted biases present in the training data, potentially leading to harm for marginalized communities. In this paper, we mitigate bias by leveraging small biased and anti-biased expert models to obtain a debiasing signal that will be added to the LLM output at decoding-time. This approach combines resource efficiency with interpretability and can be optimized for mitigating specific types of bias, depending on the target use case. Experiments on mitigating gender, race, and religion biases show a reduction in bias on several local and global bias metrics while preserving language model performance.

There is growing interest in leveraging the capabilities of robust Multi-Modal Large Language Models (MLLMs) directly within autonomous driving contexts. However, the high costs and complexity of designing and training end-to-end autonomous driving models make them challenging for many enterprises and research entities. To address this, our study explores a seamless integration of MLLMs into autonomous driving systems by proposing a Zero-Shot Chain-of-Thought (Zero-Shot-CoT) prompt design named PKRD-CoT. PKRD-CoT is based on the four fundamental capabilities of autonomous driving: perception, knowledge, reasoning, and decision-making. This makes it particularly suitable for understanding and responding to dynamic driving environments by mimicking human thought processes step by step, thus enhancing decision-making in real-time scenarios. Our design enables MLLMs to tackle problems without prior experience, thereby increasing their utility within unstructured autonomous driving environments. In experiments, we demonstrate the exceptional performance of GPT-4.0 with PKRD-CoT across autonomous driving tasks, highlighting its effectiveness in autonomous driving scenarios. Additionally, our benchmark analysis reveals the promising viability of PKRD-CoT for other MLLMs, such as Claude, LLava1.6, and Qwen-VL-Plus. Overall, this study contributes a novel and unified prompt-design framework for GPT-4.0 and other MLLMs in autonomous driving, while also rigorously evaluating the efficacy of these widely recognized MLLMs in the autonomous driving domain through comprehensive comparisons.

3D scene understanding is a long-standing challenge in computer vision and a key component in enabling mixed reality, wearable computing, and embodied AI. Providing a solution to these applications requires a multifaceted approach that covers scene-centric, object-centric, as well as interaction-centric capabilities. While there exist numerous datasets approaching the former two problems, the task of understanding interactable and articulated objects is underrepresented and only partly covered by current works. In this work, we address this shortcoming and introduce (1) an expertly curated dataset in the Universal Scene Description (USD) format, featuring high-quality manual annotations, for instance, segmentation and articulation on 280 indoor scenes; (2) a learning-based model together with a novel baseline capable of predicting part segmentation along with a full specification of motion attributes, including motion type, articulated and interactable parts, and motion parameters; (3) a benchmark serving to compare upcoming methods for the task at hand. Overall, our dataset provides 8 types of annotations - object and part segmentations, motion types, movable and interactable parts, motion parameters, connectivity, and object mass annotations. With its broad and high-quality annotations, the data provides the basis for holistic 3D scene understanding models. All data is provided in the USD format, allowing interoperability and easy integration with downstream tasks. We provide open access to our dataset, benchmark, and method's source code.

Deformable object manipulation remains a key challenge in developing autonomous robotic systems that can be successfully deployed in real-world scenarios. In this work, we explore the challenges of deformable object manipulation through the task of sculpting clay into 3D shapes. We propose the first coarse-to-fine autonomous sculpting system in which the sculpting agent first selects how many and where to place discrete chunks of clay into the workspace to create a coarse shape, and then iteratively refines the shape with sequences of deformation actions. We leverage large language models for sub-goal generation, and train a point cloud region-based action model to predict robot actions from the desired point cloud sub-goals. Additionally, our method is the first autonomous sculpting system that is a real-world text-to-3D shaping pipeline without any explicit 3D goals or sub-goals provided to the system. We demonstrate our method is able to successfully create a set of simple shapes solely from text-based prompting. Furthermore, we explore rigorously how to best quantify success for the text-to-3D sculpting task, and compare existing text-image and text-point cloud similarity metrics to human evaluations for this task. For experimental videos, human evaluation details, and full prompts, please see our project website: https://sites.google.com/andrew.cmu.edu/hierarchicalsculpting

Agents are provided with a specification document outlining the library's API as well as a suite of interactive unit tests, with the goal of producing an implementation of this API accordingly. The implementation is validated through running these unit tests. Our experiments demonstrate that while current agents can pass some unit tests, none can yet fully reproduce full libraries. Results also show that interactive feedback is quite useful for models to generate code that passes more unit tests, validating the benchmarks that facilitate its use. AI agents have been increasing rapidly in ability, particularly in domains such as problem-solving, math, and coding. Tasks related to software development have been particularly promising areas due to both their clarity of evaluation and economic value. This has motivated the release of several coding benchmarks in recent years (Hendrycks et al., 2021a; Chen et al., 2021; Zhuo et al., 2024). A notable example is SWE-bench (Jimenez et al., 2024), which assesses the ability of agents to generate patches to resolve real-world GitHub issues. While critical, these tasks generally remain within the skill set of an experienced software engineer. If LLM systems continue to improve at current rates, these tasks will be completely solvable. We are interested in benchmarks that exist further beyond both the frontier of expert human ability as well as current model ability.

Large language models (LLMs) exhibit exceptional performance across various downstream tasks. However, they encounter limitations due to slow inference speeds stemming from their extensive parameters. The early exit (EE) is an approach that aims to accelerate auto-regressive decoding. EE generates outputs from intermediate layers instead of using the whole model, which offers a promising solution to this challenge. However, additional output layers and joint optimization used in conventional EE hinder the application of EE in LLMs. In this paper, we explore the possibility of LLMs EE without additional output layers and joint optimization. Our findings indicate that EE is a natural capability within transformer-based models. While joint optimization does not give model EE capability, it must be employed to address challenges by improving the accuracy of locating the optimal EE layer through gating functions. Additionally, our study reveals patterns in EE behavior from a sub-word perspective based on the LLaMA model and the potential possibility for EE based on sub-layers.

Today's largest foundation models have increasingly general capabilities, yet when used as agents, they often struggle with simple reasoning and decision-making tasks, even though they possess good factual knowledge of the task and how to solve it. In this paper, we present a benchmark to pressure-test these models' multimodal decision-making capabilities in the very long-context regime (up to one million tokens) and investigate whether they can learn from a large number of expert demonstrations in their context. We evaluate a wide range of state-of-the-art frontier models as policies across a battery of simple interactive decision-making tasks: playing tic-tac-toe, chess, and Atari, navigating grid worlds, solving crosswords, and controlling a simulated cheetah. We measure the performance of Claude 3.5 Sonnet, Gemini 1.5 Flash, Gemini 1.5 Pro, GPT-4o, o1-mini, and o1-preview under increasing amounts of expert demonstrations in the context $\unicode{x2013}$ from no demonstrations up to 512 full episodes, pushing these models' multimodal long-context reasoning capabilities to their limits. Across our tasks, today's frontier models rarely manage to fully reach expert performance, showcasing the difficulty of our benchmark. Presenting more demonstrations often has little effect, but some models steadily improve with more demonstrations on a few tasks. We investigate the effect of encoding observations as text or images and the impact of chain-of-thought prompting. Overall, our results suggest that even today's most capable models often struggle to imitate desired behavior by generalizing purely from in-context demonstrations. To help quantify the impact of other approaches and future innovations aiming to tackle this problem, we open source our benchmark that covers the zero-, few-, and many-shot regimes in a unified evaluation.

3D affordance segmentation aims to link human instructions to touchable regions of 3D objects for embodied manipulations. Existing efforts typically adhere to single-object, single-affordance paradigms, where each affordance type or explicit instruction strictly corresponds to a specific affordance region and are unable to handle long-horizon tasks. Such a paradigm cannot actively reason about complex user intentions that often imply sequential affordances. In this paper, we introduce the Sequential 3D Affordance Reasoning task, which extends the traditional paradigm by reasoning from cumbersome user intentions and then decomposing them into a series of segmentation maps. Toward this, we construct the first instruction-based affordance segmentation benchmark that includes reasoning over both single and sequential affordances, comprising 180K instruction-point cloud pairs. Based on the benchmark, we propose our model, SeqAfford, to unlock the 3D multi-modal large language model with additional affordance segmentation abilities, which ensures reasoning with world knowledge and fine-grained affordance grounding in a cohesive framework. We further introduce a multi-granular language-point integration module to endow 3D dense prediction. Extensive experimental evaluations show that our model excels over well-established methods and exhibits open-world generalization with sequential reasoning abilities.