ICML 2024 论文列表

Effectively extracting motions from video is a critical and long-standing problem for action recognition. This problem is very challenging because motions (i) do not have an explicit form, (ii) have various concepts such as displacement, velocity, and acceleration, and (iii) often contain noise caused by unstable pixels. Addressing these challenges, we propose the Taylor video, a new video format that highlights the dominate motions (e.g., a waving hand) in each of its frames named the Taylor frame. Taylor video is named after Taylor series, which approximates a function at a given point using important terms. In the scenario of videos, we define an implicit motion-extraction function which aims to extract motions from video temporal block. In this block, using the frames, the difference frames, and higher-order difference frames, we perform Taylor expansion to approximate this function at the starting frame. We show the summation of the higher-order terms in the Taylor series gives us dominant motion patterns, where static objects, small and unstable motions are removed. Experimentally we show that Taylor videos are effective inputs to popular architectures including 2D CNNs, 3D CNNs, and transformers. When used individually, Taylor videos yield competitive action recognition accuracy compared to RGB videos and optical flow. When fused with RGB or optical flow videos, further accuracy improvement is achieved. Additionally, we apply Taylor video computation to human skeleton sequences, resulting in Taylor skeleton sequences that outperform the use of original skeletons for skeleton-based action recognition.

Current vision large language models (VLLMs) exhibit remarkable capabilities yet are prone to generate harmful content and are vulnerable to even the simplest jailbreaking attacks. Our initial analysis finds that this is due to the presence of harmful data during vision-language instruction fine-tuning, and that VLLM fine-tuning can cause forgetting of safety alignment previously learned by the underpinning LLM. To address this issue, we first curate a vision-language safe instruction-following dataset VLGuard covering various harmful categories. Our experiments demonstrate that integrating this dataset into standard vision-language fine-tuning or utilizing it for post-hoc fine-tuning effectively safety aligns VLLMs. This alignment is achieved with minimal impact on, or even enhancement of, the models' helpfulness. The versatility of our safety fine-tuning dataset makes it a valuable resource for safety-testing existing VLLMs, training new models or safeguarding pre-trained VLLMs. Empirical results demonstrate that fine-tuned VLLMs effectively reject unsafe instructions and substantially reduce the success rates of several black-box adversarial attacks, which approach zero in many cases. The code and dataset will be open-sourced.

Model-based methods have significantly contributed to distinguishing task-irrelevant distractors for visual control. However, prior research has primarily focused on heterogeneous distractors like noisy background videos, leaving homogeneous distractors that closely resemble controllable agents largely unexplored, which poses significant challenges to existing methods. To tackle this problem, we propose Implicit Action Generator (IAG) to learn the implicit actions of visual distractors, and present a new algorithm named implicit Action-informed Diverse visual Distractors Distinguisher (AD3), that leverages the action inferred by IAG to train separated world models. Implicit actions effectively capture the behavior of background distractors, aiding in distinguishing the task-irrelevant components, and the agent can optimize the policy within the task-relevant state space. Our method achieves superior performance on various visual control tasks featuring both heterogeneous and homogeneous distractors. The indispensable role of implicit actions learned by IAG is also empirically validated.

The ability to accurately comprehend natural language instructions and navigate to the target location is essential for an embodied agent. Such agents are typically required to execute user instructions in an online manner, leading us to explore the use of unlabeled test samples for effective online model adaptation. However, for online Vision-and-Language Navigation (VLN), due to the intrinsic nature of inter-sample online instruction execution and intra-sample multi-step action decision, frequent updates can result in drastic changes in model parameters, while occasional updates can make the model ill-equipped to handle dynamically changing environments. Therefore, we propose a Fast-Slow Test-Time Adaptation (FSTTA) approach for online VLN by performing joint decomposition-accumulation analysis for both gradients and parameters in a unified framework. Extensive experiments show that our method obtains impressive performance gains on four popular benchmarks. Code is available at https://github.com/Feliciaxyao/ICML2024-FSTTA.

Structured pruning is a promising approach for reducing the inference costs of large vision and language models. By removing carefully chosen structures, e.g., neurons or attention heads, the improvements from this approach can be realized on standard deep learning hardware. In this work, we focus on structured pruning in the one-shot (post-training) setting, which does not require model retraining after pruning. We propose a novel combinatorial optimization framework for this problem, based on a layer-wise reconstruction objective and a careful reformulation that allows for scalable optimization. Moreover, we design a new local combinatorial optimization algorithm, which exploits low-rank updates for efficient local search. Our framework is time and memory-efficient and considerably improves upon state-of-the-art one-shot methods on vision models (e.g., ResNet50, MobileNet) and language models (e.g., OPT-1.3B -- OPT-30B). For language models, e.g., OPT-2.7B, OSSCAR can lead to $125\times$ lower test perplexity on WikiText with $2\times$ inference time speedup in comparison to the state-of-the-art ZipLM approach. Our framework is also $6\times$ -- $8\times$ faster. Notably, our work considers models with tens of billions of parameters, which is up to $100\times$ larger than what has been previously considered in the structured pruning literature. Our code is available at https://github.com/mazumder-lab/OSSCAR.

Vision Transformers (ViTs) have achieved state-of-the-art performance for various vision tasks. One reason behind the success lies in their ability to provide plausible innate explanations for the behavior of neural architectures. However, ViTs suffer from issues with explanation faithfulness, as their focal points are fragile to adversarial attacks and can be easily changed with even slight perturbations on the input image. In this paper, we propose a rigorous approach to mitigate these issues by introducing Faithful ViTs (FViTs). Briefly speaking, an FViT should have the following two properties: (1) The top-$k$ indices of its self-attention vector should remain mostly unchanged under input perturbation, indicating stable explanations; (2) The prediction distribution should be robust to perturbations. To achieve this, we propose a new method called Denoised Diffusion Smoothing (DDS), which adopts randomized smoothing and diffusion-based denoising. We theoretically prove that processing ViTs directly with DDS can turn them into FViTs. We also show that Gaussian noise is nearly optimal for both $\ell_2$ and $\ell_\infty$-norm cases. Finally, we demonstrate the effectiveness of our approach through comprehensive experiments and evaluations. Results show that FViTs are more robust against adversarial attacks while maintaining the explainability of attention, indicating higher faithfulness.

Reward engineering has long been a challenge in Reinforcement Learning (RL) research, as it often requires extensive human effort and iterative processes of trial-and-error to design effective reward functions. In this paper, we propose RL-VLM-F, a method that automatically generates reward functions for agents to learn new tasks, using only a text description of the task goal and the agent's visual observations, by leveraging feedbacks from vision language foundation models (VLMs). The key to our approach is to query these models to give preferences over pairs of the agent's image observations based on the text description of the task goal, and then learn a reward function from the preference labels, rather than directly prompting these models to output a raw reward score, which can be noisy and inconsistent. We demonstrate that RL-VLM-F successfully produces effective rewards and policies across various domains — including classic control, as well as manipulation of rigid, articulated, and deformable objects — without the need for human supervision, outperforming prior methods that use large pretrained models for reward generation under the same assumptions. Videos can be found on our project website: https://rlvlmf2024.github.io/

This work tackles the problem of geo-localization with a new paradigm using a large vision-language model (LVLM) augmented with human inference knowledge. A primary challenge here is the scarcity of data for training the LVLM - existing street-view datasets often contain numerous low-quality images lacking visual clues, and lack any reasoning inference. To address the data-quality issue, we devise a CLIP-based network to quantify the degree of street-view images being locatable, leading to the creation of a new dataset comprising highly locatable street views. To enhance reasoning inference, we integrate external knowledge obtained from real geo-localization games, tapping into valuable human inference capabilities. The data are utilized to train GeoReasoner, which undergoes fine-tuning through dedicated reasoning and location-tuning stages. Qualitative and quantitative evaluations illustrate that GeoReasoner outperforms counterpart LVLMs by more than 25% at country-level and 38% at city-level geo-localization tasks, and surpasses StreetCLIP performance while requiring fewer training resources. The data and code are available at https://github.com/lingli1996/GeoReasoner.

Leveraging massive knowledge from large language models (LLMs), recent machine learning models show notable successes in general-purpose task solving in diverse domains such as computer vision and robotics. However, several significant challenges remain: (i) most of these models rely on 2D images yet exhibit a limited capacity for 3D input; (ii) these models rarely explore the tasks inherently defined in 3D world, e.g., 3D grounding, embodied reasoning and acting. We argue these limitations significantly hinder current models from performing real-world tasks and approaching general intelligence. To this end, we introduce LEO, an embodied multi-modal generalist agent that excels in perceiving, grounding, reasoning, planning, and acting in the 3D world. LEO is trained with a unified task interface, model architecture, and objective in two stages: (i) 3D vision-language (VL) alignment and (ii) 3D vision-language-action (VLA) instruction tuning. We collect large-scale datasets comprising diverse object-level and scene-level tasks, which require considerable understanding of and interaction with the 3D world. Moreover, we meticulously design an LLM-assisted pipeline to produce high-quality 3D VL data. Through extensive experiments, we demonstrate LEO's remarkable proficiency across a wide spectrum of tasks, including 3D captioning, question answering, embodied reasoning, navigation and manipulation. Our ablative studies and scaling analyses further provide valuable insights for developing future embodied generalist agents. Code and data are available on [project page](https://embodied-generalist.github.io/).

Large Vision-Language Models (LVLMs) show significant strides in general-propose multimodal applications such as visual dialogue and embodied navigation. However, existing multimodal evaluation benchmarks cover a limited number of multimodal tasks testing rudimentary capabilities, falling short in tracking LVLM development. In this study, we present MMT-Bench, a comprehensive benchmark designed to assess LVLMs across massive multimodal tasks requiring expert knowledge and deliberate visual recognition, localization, and reasoning. MMT-Bench comprises $31,325$ meticulously curated multi-choice visual questions from various multimodal scenarios such as vehicle driving and embodied navigation, covering $32$ core meta-tasks and $162$ subtasks in multimodal understanding. Due to its extensive task coverage, MMT-Bench enables the evaluation of LVLMs using a task map, facilitating the discovery of in- and out-of-domain tasks. Evaluation results involving $20$ publicly available LVLMs such as the proprietary GeminiProVision model, underscore the significant challenges posed by MMT-Bench. We anticipate that MMT-Bench will inspire the community to develop next-generation multimodal foundation models aimed at achieving general-purpose multimodal intelligence.

Compositional reasoning capabilities are usually considered as fundamental skills to characterize human perception. Recent studies show that current Vision Language Models (VLMs) surprisingly lack sufficient knowledge with respect to such capabilities. To this end, we propose to thoroughly diagnose the composition representations encoded by VLMs, systematically revealing the potential cause for this weakness. Specifically, we propose evaluation methods from a novel game-theoretic view to assess the vulnerability of VLMs on different aspects of compositional understanding, e.g., relations and attributes. Extensive experimental results demonstrate and validate several insights to understand the incapabilities of VLMs on compositional reasoning, which provide useful and reliable guidance for future studies. The deliverables will be updated [here](https://vlms-compositionality-gametheory.github.io/).

Recently, various pre-training methods have been introduced in vision-based Reinforcement Learning (RL). However, their generalization ability remains unclear due to evaluations being limited to in-distribution environments and non-unified experimental setups. To address this, we introduce the Atari Pre-training Benchmark (Atari-PB), which pre-trains a ResNet-50 model on 10 million transitions from 50 Atari games and evaluates it across diverse environment distributions. Our experiments show that pre-training objectives focused on learning task-agnostic features (e.g., identifying objects and understanding temporal dynamics) enhance generalization across different environments. In contrast, objectives focused on learning task-specific knowledge (e.g., identifying agents and fitting reward functions) improve performance in environments similar to the pre-training dataset but not in varied ones. We publicize our codes, datasets, and model checkpoints at https://github.com/dojeon-ai/Atari-PB.

Transformer models have been gaining substantial interest in the field of computer vision tasks nowadays. Although a vision transformer contains two important components which are self-attention module and feedforward network (FFN) module, the majority of research tends to concentrate on modifying the former while leaving the latter in its original form. In this paper, we focus on improving the FFN module within the vision transformer. Through theoretical analysis, we demonstrate that the effect of the FFN module primarily lies in providing non-linearity, whose degree corresponds to the hidden dimensions. Thus, the computational cost of the FFN module can be reduced by enhancing the degree of non-linearity in the nonlinear function. Leveraging this insight, we propose an improved FFN (IFFN) module for vision transformers which involves the usage of the arbitrary GeLU (AGeLU) function and integrating multiple instances of it to augment non-linearity so that the number of hidden dimensions can be effectively reduced. Besides, a spatial enhancement part is involved to further enrich the non-linearity in the proposed IFFN module. Experimental results show that we can apply our method to a wide range of state-of-the-art vision transformer models irrespective of how they modify their self-attention part and the overall architecture, and reduce FLOPs and parameters without compromising classification accuracy on the ImageNet dataset.

Although pre-trained models such as Contrastive Language-Image Pre-Training (CLIP) show impressive generalization results, their robustness is still limited under Out-of-Distribution (OOD) scenarios. Instead of undesirably leveraging human annotation as commonly done, it is possible to leverage the visual understanding power of Multi-modal Large Language Models (MLLMs). However, MLLMs struggle with vision problems due to task incompatibility, thus hindering their effectiveness. In this paper, we propose to effectively leverage MLLMs via Machine Vision Therapy which aims to rectify erroneous predictions of specific vision models. By supervising vision models using MLLM predictions, visual robustness can be boosted in a nearly unsupervised manner. Moreover, we propose a Denoising In-Context Learning (DICL) strategy to solve the incompatibility issue. Concretely, by examining the noise probability of each example through a transition matrix, we construct an instruction containing a correct exemplar and a probable erroneous one, which enables MLLMs to detect and rectify the incorrect predictions of vision models. Under mild assumptions, we theoretically show that our DICL method is guaranteed to find the ground truth. Through extensive experiments on various OOD datasets, our method demonstrates powerful capabilities for enhancing visual robustness under many OOD scenarios.

Fine-tuning pre-trained vision-language models (VLMs), e.g., CLIP, for the open-world generalization has gained increasing popularity due to its practical value. However, performance advancements are limited when relying solely on intricate algorithmic designs for a single model, even one exhibiting strong performance, e.g., CLIP-ViT-B/16. This paper, for the first time, explores the collaborative potential of leveraging much weaker VLMs to enhance the generalization of a robust single model. The affirmative findings motivate us to address the generalization problem from a novel perspective, i.e., ensemble of pre-trained VLMs. We introduce three customized ensemble strategies, each tailored to one specific scenario. Firstly, we introduce the zero-shot ensemble, automatically adjusting the logits of different models based on their confidence when only pre-trained VLMs are available. Furthermore, for scenarios with extra few-shot samples, we propose the training-free and tuning ensemble, offering flexibility based on the availability of computing resources. The code is available at https://github.com/zhiheLu/Ensemble_VLM.git.

Training general-purpose vision models on purely sequential visual data, eschewing linguistic inputs, has heralded a new frontier in visual understanding. These models are intended to not only comprehend but also seamlessly transit to out-of-domain tasks. However, current endeavors are hamstrung by an over-reliance on colossal models, exemplified by models with upwards of 3B parameters, and the necessity for an extensive corpus of visual data, often comprising a staggering 400B tokens. In this paper, we delve into the development of an efficient, autoregression-based vision model, innovatively architected to operate on a limited dataset. We meticulously demonstrate how this model achieves proficiency in a spectrum of visual tasks spanning both high-level and low-level semantic understanding during the testing phase. Our empirical evaluations underscore the model's agility in adapting to various tasks, heralding a significant reduction in the parameter footprint, and a marked decrease in training data requirements, thereby paving the way for more sustainable and accessible advancements in the field of generalist vision models. The code is available at https://github.com/ggjy/DeLVM.

We present Premier-TACO, a multitask feature representation learning approach designed to improve few-shot policy learning efficiency in sequential decision-making tasks. Premier-TACO leverages a subset of multitask offline datasets for pretraining a general feature representation, which captures critical environmental dynamics and is fine-tuned using minimal expert demonstrations. It advances the temporal action contrastive learning (TACO) objective, known for state-of-the-art results in visual control tasks, by incorporating a novel negative example sampling strategy. This strategy is crucial in significantly boosting TACO’s computational efficiency, making large-scale multitask offline pretraining feasible. Our extensive empirical evaluation in a diverse set of continuous control benchmarks including Deepmind Control Suite, MetaWorld, and LIBERO demonstrate Premier-TACO’s effective- ness in pretraining visual representations, significantly enhancing few-shot imitation learning of novel tasks.

It has recently been discovered that using a pre-trained *vision-language model* (VLM), e.g., CLIP, to align a whole query image with several finer text descriptions generated by a large language model can significantly enhance zero-shot performance. However, in this paper, we empirically find that the finer descriptions tend to align more effectively with *local areas of the query image* rather than the whole image, and then we theoretically validate this finding. Thus, we present a method called *weighted visual-text cross alignment* (WCA). This method begins with a *localized visual prompting* technique, designed to identify local visual areas within the query image. The local visual areas are then *cross-aligned* with the finer descriptions by creating a similarity matrix using the pre-trained VLM. To determine how well a query image aligns with each category, we develop a score function based on the weighted similarities in this matrix. Extensive experiments demonstrate that our method significantly improves zero-shot performance across various datasets, achieving results that are even comparable to few-shot learning methods.

In this paper, we introduce MAAD, a novel, sample-efficient on-policy algorithm for Imitation Learning from Observations. MAAD utilizes a surrogate reward signal, which can be derived from various sources such as adversarial games, trajectory matching objectives, or optimal transport criteria. To compensate for the non-availability of expert actions, we rely on an inverse dynamics model that infers plausible actions distribution given the expert’s state-state transitions; we regularize the imitator’s policy by aligning it to the inferred action distribution. MAAD leads to significantly improved sample efficiency and stability. We demonstrate its effectiveness in a number of MuJoCo environments, both int the OpenAI Gym and the DeepMind Control Suite. We show that it requires considerable fewer interactions to achieve expert performance, outperforming current state-of-the-art on-policy methods. Remarkably, MAAD often stands out as the sole method capable of attaining expert performance levels, underscoring its simplicity and efficacy.