TokenFormer: Rethinking Transformer Scaling with Tokenized Model Parameters

Transformers have become the predominant architecture in foundation models due to their excellent performance across various domains. However, the substantial cost of scaling these models remains a significant concern. This problem arises primarily from their dependence on a fixed number of parameters within linear projections. When architectural modifications (e.g., channel dimensions) are introduced, the entire model typically requires retraining from scratch. As model sizes continue growing, this strategy results in increasingly high computational costs and becomes unsustainable. To overcome this problem, we introduce TokenFormer, a natively scalable architecture that leverages the attention mechanism not only for computations among input tokens but also for interactions between tokens and model parameters, thereby enhancing architectural flexibility. By treating model parameters as tokens, we replace all the linear projections in Transformers with our token-parameter attention layer, where input tokens act as queries and model parameters as keys and values. This reformulation allows for progressive and efficient scaling without necessitating retraining from scratch. Our model scales from 124M to 1.4B parameters by incrementally adding new key-value parameter pairs, achieving performance comparable to Transformers trained from scratch while greatly reducing training costs. Code and models are available at \url{https://github.com/Haiyang-W/TokenFormer}.

Toward a Diffusion-Based Generalist for Dense Vision Tasks

Building generalized models that can solve many computer vision tasks simultaneously is an intriguing direction. Recent works have shown image itself can be used as a natural interface for general-purpose visual perception and demonstrated inspiring results. In this paper, we explore diffusion-based vision generalists, where we unify different types of dense prediction tasks as conditional image generation and re-purpose pre-trained diffusion models for it. However, directly applying off-the-shelf latent diffusion models leads to a quantization issue. Thus, we propose to perform diffusion in pixel space and provide a recipe for finetuning pre-trained text-to-image diffusion models for dense vision tasks. In experiments, we evaluate our method on four different types of tasks and show competitive performance to the other vision generalists.

Read Anywhere Pointed: Layout-aware GUI Screen Reading with Tree-of-Lens Grounding

Graphical User Interfaces (GUIs) are central to our interaction with digital devices. Recently, growing efforts have been made to build models for various GUI understanding tasks. However, these efforts largely overlook an important GUI-referring task: screen reading based on user-indicated points, which we name the Screen Point-and-Read (SPR) task. This task is predominantly handled by rigid accessible screen reading tools, in great need of new models driven by advancements in Multimodal Large Language Models (MLLMs). In this paper, we propose a Tree-of-Lens (ToL) agent, utilizing a novel ToL grounding mechanism, to address the SPR task. Based on the input point coordinate and the corresponding GUI screenshot, our ToL agent constructs a Hierarchical Layout Tree. Based on the tree, our ToL agent not only comprehends the content of the indicated area but also articulates the layout and spatial relationships between elements. Such layout information is crucial for accurately interpreting information on the screen, distinguishing our ToL agent from other screen reading tools. We also thoroughly evaluate the ToL agent against other baselines on a newly proposed SPR benchmark, which includes GUIs from mobile, web, and operating systems. Last but not least, we test the ToL agent on mobile GUI navigation tasks, demonstrating its utility in identifying incorrect actions along the path of agent execution trajectories. Code and data: screen-point-and-read.github.io

MMWorld: Towards Multi-discipline Multi-faceted World Model Evaluation in Videos

Multimodal Language Language Models (MLLMs) demonstrate the emerging abilities of "world models" -- interpreting and reasoning about complex real-world dynamics. To assess these abilities, we posit videos are the ideal medium, as they encapsulate rich representations of real-world dynamics and causalities. To this end, we introduce MMWorld, a new benchmark for multi-discipline, multi-faceted multimodal video understanding. MMWorld distinguishes itself from previous video understanding benchmarks with two unique advantages: (1) multi-discipline, covering various disciplines that often require domain expertise for comprehensive understanding; (2) multi-faceted reasoning, including explanation, counterfactual thinking, future prediction, etc. MMWorld consists of a human-annotated dataset to evaluate MLLMs with questions about the whole videos and a synthetic dataset to analyze MLLMs within a single modality of perception. Together, MMWorld encompasses 1,910 videos across seven broad disciplines and 69 subdisciplines, complete with 6,627 question-answer pairs and associated captions. The evaluation includes 2 proprietary and 10 open-source MLLMs, which struggle on MMWorld (e.g., GPT-4V performs the best with only 52.3\% accuracy), showing large room for improvement. Further ablation studies reveal other interesting findings such as models' different skill sets from humans. We hope MMWorld can serve as an essential step towards world model evaluation in videos.

Semantic Gaussians: Open-Vocabulary Scene Understanding with 3D Gaussian Splatting

Open-vocabulary 3D scene understanding presents a significant challenge in computer vision, withwide-ranging applications in embodied agents and augmented reality systems. Previous approaches haveadopted Neural Radiance Fields (NeRFs) to analyze 3D scenes. In this paper, we introduce SemanticGaussians, a novel open-vocabulary scene understanding approach based on 3D Gaussian Splatting. Our keyidea is distilling pre-trained 2D semantics into 3D Gaussians. We design a versatile projection approachthat maps various 2Dsemantic features from pre-trained image encoders into a novel semantic component of 3D Gaussians, withoutthe additional training required by NeRFs. We further build a 3D semantic network that directly predictsthe semantic component from raw 3D Gaussians for fast inference. We explore several applications ofSemantic Gaussians: semantic segmentation on ScanNet-20, where our approach attains a 4.2% mIoU and 4.0%mAcc improvement over prior open-vocabulary scene understanding counterparts; object part segmentation,sceneediting, and spatial-temporal segmentation with better qualitative results over 2D and 3D baselines,highlighting its versatility and effectiveness on supporting diverse downstream tasks.

VideoAgent: A Memory-augmented Multimodal Agent for Video Understanding

We explore how reconciling several foundation models (large language models and vision-language models) with a novel unified memory mechanism could tackle the challenging video understanding problem, especially capturing the long-term temporal relations in lengthy videos. In particular, the proposed multimodal agent VideoAgent: 1) constructs a structured memory to store both the generic temporal event descriptions and object-centric tracking states of the video; 2) given an input task query, it employs tools including video segment localization and object memory querying along with other visual foundation models to interactively solve the task, utilizing the zero-shot tool-use ability of LLMs. VideoAgent demonstrates impressive performances on several long-horizon video understanding benchmarks, an average increase of 6.6% on NExT-QA and 26.0% on EgoSchema over baselines, closing the gap between open-sourced models and private counterparts including Gemini 1.5 Pro.

Parameter-Efficient Fine-Tuning for Pre-Trained Vision Models: A Survey

Large-scale pre-trained vision models (PVMs) have shown great potential for adaptability across various downstream vision tasks. However, with state-of-the-art PVMs growing to billions or even trillions of parameters, the standard full fine-tuning paradigm is becoming unsustainable due to high computational and storage demands. In response, researchers are exploring parameter-efficient fine-tuning (PEFT), which seeks to exceed the performance of full fine-tuning with minimal parameter modifications. This survey provides a comprehensive overview and future directions for visual PEFT, offering a systematic review of the latest advancements. First, we provide a formal definition of PEFT and discuss model pre-training methods. We then categorize existing methods into three categories: addition-based, partial-based, and unified-based. Finally, we introduce the commonly used datasets and applications and suggest potential future research challenges. A comprehensive collection of resources is available at https://github.com/synbol/Awesome-Parameter-Efficient-Transfer-Learning.

Muffin or Chihuahua? Challenging Large Vision-Language Models with Multipanel VQA

Multipanel images, commonly seen as web screenshots, posters, etc., pervade our daily lives. These images, characterized by their composition of multiple subfigures in distinct layouts, effectively convey information to people. Toward building advanced multimodal AI applications, such as agents that understand complex scenes and navigate through webpages, the skill of multipanel visual reasoning is essential, and a comprehensive evaluation of models in this regard is important. Therefore, our paper introduces Multipanel Visual Question Answering (MultipanelVQA), a novel benchmark that specifically challenges models in comprehending multipanel images. The benchmark comprises 6,600 questions and answers related to multipanel images. While these questions are straightforward for average humans, achieving nearly perfect correctness, they pose significant challenges to the state-of-the-art Large Vision Language Models (LVLMs) we tested. In our study, we utilized synthetically curated multipanel images specifically designed to isolate and evaluate the impact of diverse factors on model performance, revealing the sensitivity of LVLMs to various interferences in multipanel images, such as adjacent subfigures and layout complexity. As a result, MultipanelVQA highlights the need and direction for improving LVLMs' ability to understand complex visual-language contexts. Code and data are released at https://sites.google.com/view/multipanelvqa/home.

SSB: Simple but Strong Baseline for Boosting Performance of Open-Set Semi-Supervised Learning

Semi-supervised learning (SSL) methods effectively leverage unlabeled data to improve model generalization. However, SSL models often underperform in open-set scenarios, where unlabeled data contain outliers from novel categories that do not appear in the labeled set. In this paper, we study the challenging and realistic open-set SSL setting, where the goal is to both correctly classify inliers and to detect outliers. Intuitively, the inlier classifier should be trained on inlier data only. However, we find that inlier classification performance can be largely improved by incorporating high-confidence pseudo-labeled data, regardless of whether they are inliers or outliers. Also, we propose to utilize non-linear transformations to separate the features used for inlier classification and outlier detection in the multi-task learning framework, preventing adverse effects between them. Additionally, we introduce pseudo-negative mining, which further boosts outlier detection performance. The three ingredients lead to what we call Simple but Strong Baseline (SSB) for open-set SSL. In experiments, SSB greatly improves both inlier classification and outlier detection performance, outperforming existing methods by a large margin. Our code will be released at https://github.com/YUE-FAN/SSB.

Evaluating Multi-Agent Coordination Abilities in Large Language Models

A pivotal aim in contemporary AI research is to develop agents proficient in multi-agent coordination, enabling effective collaboration with both humans and other systems. Large Language Models (LLMs), with their notable ability to understand, generate, and interpret language in a human-like manner, stand out as promising candidates for the development of such agents. In this study, we build and assess the effectiveness of agents crafted using LLMs in various coordination scenarios. We introduce the LLM-Coordination (LLM-Co) Framework, specifically designed to enable LLMs to play coordination games. With the LLM-Co framework, we conduct our evaluation with three game environments and organize the evaluation into five aspects: Theory of Mind, Situated Reasoning, Sustained Coordination, Robustness to Partners, and Explicit Assistance. First, the evaluation of the Theory of Mind and Situated Reasoning reveals the capabilities of LLM to infer the partner's intention and reason actions accordingly. Then, the evaluation around Sustained Coordination and Robustness to Partners further showcases the ability of LLMs to coordinate with an unknown partner in complex long-horizon tasks, outperforming Reinforcement Learning baselines. Lastly, to test Explicit Assistance, which refers to the ability of an agent to offer help proactively, we introduce two novel layouts into the Overcooked-AI benchmark, examining if agents can prioritize helping their partners, sacrificing time that could have been spent on their tasks. This research underscores the promising capabilities of LLMs in sophisticated coordination environments and reveals the potential of LLMs in building strong real-world agents for multi-agent coordination.