Building Graphical User Interface (GUI) agents is a promising research direction, which simulates human interaction with computers or mobile phones to perform diverse GUI tasks. However, a major challenge in developing generalized GUI agents is the lack of sufficient trajectory data across various operating systems and applications, mainly due to the high cost of manual annotations. In this paper, we propose the TongUI framework that transforms millions of multimodal web tutorials into GUI trajectories for generalized GUI agents. Concretely, we crawl GUI videos and articles from the Internet and process them into GUI agent trajectory data. Based on this, we construct the GUI-Net-1M dataset, which contains 1 million trajectories across five operating systems and over 280 applications. To the best of our knowledge, this is the largest open-source GUI trajectory dataset. We develop the TongUI agent by fine-tuning Qwen2.5-VL-3B/7B/32B models on GUI-Net-1M, which shows consistent performance improvements on commonly used grounding and navigation benchmarks, outperforming baseline agents by 10\% on multiple benchmarks, showing the effectiveness of the GUI-Net-1M dataset and underscoring the significance of our TongUI framework.

Open-vocabulary 3D scene understanding presents a significant challenge in computer vision, with wide-ranging applications in embodied agents and augmented reality systems. Existing methods adopt neural rendering methods as 3D representations and jointly optimize color and semantic features to achieve rendering and scene understanding simultaneously. In this paper, we introduce Semantic Gaussians, a novel open-vocabulary scene understanding approach based on 3D Gaussian Splatting. Our key idea is to distill knowledge from 2D pretrained models to 3D Gaussians. Unlike existing methods, we design a versatile projection approach that maps various 2D semantic features from pre-trained image encoders into a novel semantic component of 3D Gaussians, which is based on spatial relationships and needs no additional training. We further build a 3D semantic network that directly predicts the semantic component from raw 3D Gaussians for fast inference. The quantitative results on ScanNet segmentation and LERF object localization demonstrate the superior performance of our method. Additionally, we explore several applications of Semantic Gaussians, including object part segmentation, instance segmentation, scene editing, and spatiotemporal segmentation, with better qualitative results over 2D and 3D baselines, highlighting its versatility and effectiveness in supporting diverse downstream tasks.

Existing 3D vision-language (3D-VL) benchmarks fall short in evaluating 3D-VL models, creating a "mist" that obscures rigorous insights into model capabilities and 3D-VL tasks. This mist persists due to three key limitations. First, flawed test data, like ambiguous referential text in the grounding task, can yield incorrect and unreliable test results. Second, oversimplified metrics such as simply averaging accuracy per question answering (QA) pair, cannot reveal true model capability due to their vulnerability to language variations. Third, existing benchmarks isolate the grounding and QA tasks, disregarding the underlying coherence that QA should be based on solid grounding capabilities. To unveil the "mist", we propose Beacon3D, a benchmark for 3D-VL grounding and QA tasks, delivering a perspective shift in the evaluation of 3D-VL understanding. Beacon3D features (i) high-quality test data with precise and natural language, (ii) object-centric evaluation with multiple tests per object to ensure robustness, and (iii) a novel chain-of-analysis paradigm to address language robustness and model performance coherence across grounding and QA. Our evaluation of state-of-the-art 3D-VL models on Beacon3D reveals that (i) object-centric evaluation elicits true model performance and particularly weak generalization in QA; (ii) grounding-QA coherence remains fragile in current 3D-VL models, and (iii) incorporating large language models (LLMs) to 3D-VL models, though as a prevalent practice, hinders grounding capabilities and has yet to elevate QA capabilities. We hope Beacon3D and our comprehensive analysis could benefit the 3D-VL community towards faithful developments.

Embodied AI (EAI) research requires high-quality, diverse 3D scenes to effectively support skill acquisition, sim-to-real transfer, and generalization. Achieving these quality standards, however, necessitates the precise replication of real-world object diversity. Existing datasets demon strate that this process heavily relies on artist-driven designs, which demand substantial human effort and present significant scalability challenges. To scalably produce realistic and interactive 3D scenes, we first present MetaScenes, a large-scale simulatable 3D scene dataset constructed from real-world scans, which includes 15366 objects spanning 831 fine-grained categories. Then, we introduce SCAN2SIM, a robust multi-modal alignment model, which enables the automated, high-quality replacement of assets, thereby eliminating the reliance on artist-driven designs for scaling 3D scenes. We further propose two benchmarks to evaluate MetaScenes: a detailed scene synthesis task focused on small item layouts for robotic manipulation and a domain transfer task in vision-and-language navigation (VLN) to validate cross-domain transfer. Results confirm MetaScenes ’s potential to enhance EAI by supporting more generalizable agent learning and sim-to-real applications, introducing new possibilities for EAI research.

Embodied scene understanding requires not only comprehending visual-spatial information that has been observed but also determining where to explore next in the 3D physical world. Existing 3D Vision-Language (3D-VL) models primarily focus on grounding objects in static observations from 3D reconstruction, such as meshes and point clouds, but lack the ability to actively perceive and explore their environment. To address this limitation, we introduce \underline{\textbf{M}}ove \underline{\textbf{t}}o \underline{\textbf{U}}nderstand (\textbf{\model}), a unified framework that integrates active perception with \underline{\textbf{3D}} vision-language learning, enabling embodied agents to effectively explore and understand their environment. This is achieved by three key innovations: 1) Online query-based representation learning, enabling direct spatial memory construction from RGB-D frames, eliminating the need for explicit 3D reconstruction. 2) A unified objective for grounding and exploring, which represents unexplored locations as frontier queries and jointly optimizes object grounding and frontier selection. 3) End-to-end trajectory learning that combines \textbf{V}ision-\textbf{L}anguage-\textbf{E}xploration pre-training over a million diverse trajectories collected from both simulated and real-world RGB-D sequences. Extensive evaluations across various embodied navigation and question-answering benchmarks show that MTU3D outperforms state-of-the-art reinforcement learning and modular navigation approaches by 14\%, 23\%, 9\%, and 2\% in success rate on HM3D-OVON, GOAT-Bench, SG3D, and A-EQA, respectively. \model's versatility enables navigation using diverse input modalities, including categories, language descriptions, and reference images. These findings highlight the importance of bridging visual grounding and exploration for embodied intelligence.

Knowledge editing techniques have emerged as essential tools for updating the factual knowledge of large language models (LLMs) and multimodal models (LMMs), allowing them to correct outdated or inaccurate information without retraining from scratch. However, existing benchmarks for multimodal knowledge editing primarily focus on entity-level knowledge represented as simple triplets, which fail to capture the complexity of real-world multimodal information. To address this issue, we introduce MMKE-Bench, a comprehensive **M**ulti**M**odal **K**nowledge **E**diting Benchmark, designed to evaluate the ability of LMMs to edit diverse visual knowledge in real-world scenarios. MMKE-Bench addresses these limitations by incorporating three types of editing tasks: visual entity editing, visual semantic editing, and user-specific editing. Besides, MMKE-Bench uses free-form natural language to represent and edit knowledge, offering a more flexible and effective format. The benchmark consists of 2,940 pieces of knowledge and 8,363 images across 33 broad categories, with evaluation questions automatically generated and human-verified. We assess five state-of-the-art knowledge editing methods on three prominent LMMs, revealing that no method excels across all criteria, and that visual and user-specific edits are particularly challenging. MMKE-Bench sets a new standard for evaluating the robustness of multimodal knowledge editing techniques, driving progress in this rapidly evolving field.

The advancement of large language models (LLMs) prompts the development of multi-modal agents, which are used as a controller to call external tools, providing a feasible way to solve practical tasks. In this paper, we propose a multi-modal agent tuning method that automatically generates multi-modal tool-usage data and tunes a vision-language model (VLM) as the controller for powerful tool-usage reasoning. To preserve the data quality, we prompt the GPT-4o mini model to generate queries, files, and trajectories, followed by query-file and trajectory verifiers. Based on the data synthesis pipeline, we collect the MM-Traj dataset that contains 20K tasks with trajectories of tool usage. Then, we develop the T3-Agent via Trajectory Tuning on VLMs for Tool usage using MM-Traj. Evaluations on the GTA and GAIA benchmarks show that the T3-Agent consistently achieves improvements on two popular VLMs: MiniCPM-V-8.5B and Qwen2-VL-7B, which outperforms untrained VLMs by 20%, showing the effectiveness of the proposed data synthesis pipeline, leading to high-quality data for tool-usage capabilities.

Diffusion policies are widely adopted in complex visuomotor tasks for their ability to capture multimodal action distributions. However, the multiple sampling steps required for action generation significantly harm real-time inference efficiency, which limits their applicability in real-time decision-making scenarios. Existing acceleration techniques either require retraining or degrade performance under low sampling steps. Here we propose Falcon, which mitigates this speed-performance trade-off and achieves further acceleration. The core insight is that visuomotor tasks exhibit sequential dependencies between actions. Falcon leverages this by reusing partially denoised actions from historical information rather than sampling from Gaussian noise at each step. By integrating current observations, Falcon reduces sampling steps while preserving performance. Importantly, Falcon is a training-free algorithm that can be applied as a plug-in to further improve decision efficiency on top of existing acceleration techniques. We validated Falcon in 48 simulated environments and 2 real-world robot experiments. demonstrating a 2-7x speedup with negligible performance degradation, offering a promising direction for efficient visuomotor policy design.

Recently, large language models (LLMs) have shown strong potential in facilitating human-robotic interaction and collaboration. However, existing LLM-based systems often overlook the misalignment between human and robot perceptions, which hinders their effective communication and real-world robot deployment. To address this issue, we introduce SYNERGAI, a unified system designed to achieve both perceptual alignment and human-robot collaboration. At its core, SYNERGAI employs 3D Scene Graph (3DSG) as its explicit and innate representation. This enables the system to leverage LLM to break down complex tasks and allocate appropriate tools in intermediate steps to extract relevant information from the 3DSG, modify its structure, or generate responses. Importantly, SYNERGAI incorporates an automatic mechanism that enables perceptual misalignment correction with users by updating its 3DSG with online interaction. SYNERGAI achieves comparable performance with the data-driven models in ScanQA in a zero-shot manner. Through comprehensive experiments across 10 real-world scenes, SYNERGAI demonstrates its effectiveness in establishing common ground with humans, realizing a success rate of 61.9 % in alignment tasks. It also significantly improves the success rate from 3.7% to 45.68 % on novel tasks by transferring the knowledge acquired during alignment.

Multimodal agents, which integrate a controller (e.g., a vision language model) with external tools, have demonstrated remarkable capabilities in tackling complex multimodal tasks. Existing approaches for training these agents, both supervised fine-tuning and reinforcement learning, depend on extensive human-annotated task-answer pairs and tool trajectories. However, for complex multimodal tasks, such annotations are prohibitively expensive or impractical to obtain. In this paper, we propose an iterative tool usage exploration method for multimodal agents without any pre-collected data, namely SPORT, via step-wise preference optimization to refine the trajectories of tool usage. Our method enables multimodal agents to autonomously discover effective tool usage strategies through self-exploration and optimization, eliminating the bottleneck of human annotation. SPORT has four iterative components: task synthesis, step sampling, step verification, and preference tuning. We first synthesize multimodal tasks using language models. Then, we introduce a novel trajectory exploration scheme, where step sampling and step verification are executed alternately to solve synthesized tasks. In step sampling, the agent tries different tools and obtains corresponding results. In step verification, we employ a verifier to provide AI feedback to construct step-wise preference data. The data is subsequently used to update the controller for tool usage through preference tuning, producing a SPORT agent. By interacting with real environments, the SPORT agent gradually evolves into a more refined and capable system. Evaluation in the GTA and GAIA benchmarks shows that the SPORT agent achieves 6.41% and 3.64% improvements, underscoring the generalization and effectiveness introduced by our method.

3D visual grounding has made notable progress in localizing objects within complex 3D scenes. However, grounding referring expressions beyond objects in 3D scenes remains unexplored. In this paper, we introduce Anywhere3D-Bench, a holistic 3D visual grounding benchmark consisting of 2,886 referring expression-3D bounding box pairs spanning four different grounding levels: human-activity areas, unoccupied space beyond objects, individual objects in the scene, and fine-grained object parts. We assess a range of state-of-the-art 3D visual grounding methods alongside large language models (LLMs) and multimodal LLMs (MLLMs) on Anywhere3D-Bench. Experimental results reveal that space-level and part-level visual grounding pose the greatest challenges: space-level tasks require a more comprehensive spatial reasoning ability, for example, modeling distances and spatial relations within 3D space, while part-level tasks demand fine-grained perception of object composition. Even the best performance model, OpenAI o4-mini, achieves only 23.00% accuracy on space-level tasks and 31.46% on part-level tasks, significantly lower than its performance on area-level and object-level tasks. These findings underscore a critical gap in current models’ capacity to understand and reason about 3D scenes beyond object-level semantics.

Text-guided image editing involves modifying a source image based on a language instruction and, typically, requires changes to only small local regions. However, existing approaches generate the entire target image rather than selectively regenerate only the intended editing areas. This results in (1) unnecessary computational costs and (2) a bias toward reconstructing non-editing regions, which compromises the quality of the intended edits. To resolve these limitations, we propose to formulate image editing as $\textbf{N}$ext $\textbf{E}$diting-token $\textbf{P}$rediction (NEP) based on autoregressive image generation, where only regions that need to be edited are regenerated, thus avoiding unintended modification to the non-editing areas. To enable any-region editing, we propose to pre-train an any-order autoregressive text-to-image (T2I) model. Once trained, it is capable of zero-shot image editing and can be easily adapted to NEP for image editing, which achieves a new state-of-the-art on widely used image editing benchmarks. Moreover, our model naturally supports test-time scaling (TTS) through iteratively refining its generation in a zero-shot manner.

Utilizing large language models (LLMs) to compose off-the-shelf visual tools represents a promising avenue of research for developing robust visual assistants capable of addressing diverse visual tasks. However, these methods often overlook the potential for continual learning, typically by freezing the utilized tools, thus limiting their adaptation to environments requiring new knowledge. To tackle this challenge, we propose CLOVA, a Closed-LOop Visual Assistant, which operates within a framework encompassing inference, reflection, and learning phases. During the inference phase, LLMs generate programs and execute corresponding tools to complete assigned tasks. In the reflection phase, a multimodal global-local reflection scheme analyzes human feedback to determine which tools require updating. Lastly, the learning phase employs three flexible approaches to automatically gather training data and introduces a novel prompt tuning scheme to update the tools, allowing CLOVA to efficiently acquire new knowledge. Experimental findings demonstrate that CLOVA surpasses existing tool-usage methods by 5% in visual question answering and multiple-image reasoning, by 10% in knowledge tagging, and by 20% in image editing. These results under-score the significance of the continual learning capability in general visual assistants.

Current learning models often struggle with human-like systematic generalization, particularly in learning compositional rules from limited data and extrapolating them to novel combinations. We introduce the Neural-Symbolic Recursive Ma- chine ( NSR), whose core is a Grounded Symbol System ( GSS), allowing for the emergence of combinatorial syntax and semantics directly from training data. The NSR employs a modular design that integrates neural perception, syntactic parsing, and semantic reasoning. These components are synergistically trained through a novel deduction-abduction algorithm. Our findings demonstrate that NSR’s design, imbued with the inductive biases of equivariance and compositionality, grants it the expressiveness to adeptly handle diverse sequence-to-sequence tasks and achieve unparalleled systematic generalization. We evaluate NSR’s efficacy across four challenging benchmarks designed to probe systematic generalization capabilities: SCAN for semantic parsing, PCFG for string manipulation, HINT for arithmetic reasoning, and a compositional machine translation task. The results affirm NSR ’s superiority over contemporary neural and hybrid models in terms of generalization and transferability.

We introduce Bongard-OpenWorld, a new benchmark for evaluating real-world few-shot reasoning for machine vision. It originates from the classical Bongard Problems (BPs): Given two sets of images (positive and negative), the model needs to identify the set that query images belong to by inducing the visual concepts, which is exclusively depicted by images from the positive set. Our benchmark inherits the few-shot concept induction of the original BPs while adding the two novel layers of challenge: 1) open-world free-form concepts, as the visual concepts in Bongard-OpenWorld are unique compositions of terms from an open vocabulary, ranging from object categories to abstract visual attributes and commonsense factual knowledge; 2) real-world images, as opposed to the synthetic diagrams used by many counterparts. In our exploration, Bongard-OpenWorld already imposes a significant challenge to current few-shot reasoning algorithms. We further investigate to which extent the recently introduced Large Language Models (LLMs) and Vision-Language Models (VLMs) can solve our task, by directly probing VLMs, and combining VLMs and LLMs in an interactive reasoning scheme. We even conceived a neuro-symbolic reasoning approach that reconciles LLMs & VLMs with logical reasoning to emulate the human problem-solving process for Bongard Problems. However, none of these approaches manage to close the human-machine gap, as the best learner achieves 64% accuracy while human participants easily reach 91%. We hope Bongard-OpenWorld can help us better understand the limitations of current visual intelligence and facilitate future research on visual agents with stronger few-shot visual reasoning capabilities.

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/).

Neuro-symbolic reinforcement learning (NS-RL) has emerged as a promising paradigm for explainable decision-making, characterized by the interpretability of symbolic policies. NS-RL entails structured state representations for tasks with visual observations, but previous methods cannot refine the structured states with rewards due to a lack of efficiency. Accessibility also remains an issue, as extensive domain knowledge is required to interpret symbolic policies. In this paper, we present a neuro-symbolic framework for jointly learning structured states and symbolic policies, whose key idea is to distill the vision foundation model into an efficient perception module and refine it during policy learning. Moreover, we design a pipeline to prompt GPT-4 to generate textual explanations for the learned policies and decisions, significantly reducing users’ cognitive load to understand the symbolic policies. We verify the efficacy of our approach on nine Atari tasks and present GPT-generated explanations for policies and decisions.

Vision language models (VLMs) have achieved impressive progress in diverse applications, becoming a prevalent research direction. In this paper, we build FIRE, a feedback-refinement dataset, consisting of 1.1M multi-turn conversations that are derived from 27 source datasets, empowering VLMs to spontaneously refine their responses based on user feedback across diverse tasks. To scale up the data collection, FIRE is collected in two components: FIRE-100K and FIRE-1M, where FIRE-100K is generated by GPT-4V, and FIRE-1M is freely generated via models trained on FIRE-100K. Then, we build FIRE-Bench, a benchmark to comprehensively evaluate the feedback-refining capability of VLMs, which contains 11K feedback-refinement conversations as the test data, two evaluation settings, and a model to provide feedback for VLMs. We develop the FIRE-LLaVA model by fine-tuning LLaVA on FIRE-100K and FIRE-1M, which shows remarkable feedback-refining capability on FIRE-Bench and outperforms untrained VLMs by 50%, making more efficient user-agent interactions and underscoring the significance of the FIRE dataset.

This paper presents OmniJARVIS, a novel Vision-Language-Action (VLA) model for open-world instruction-following agents in Minecraft. Compared to prior works that either emit textual goals to separate controllers or produce the control command directly, OmniJARVIS seeks a different path to ensure both strong reasoning and efficient decision-making capabilities via unified tokenization of multimodal interaction data. First, we introduce a self-supervised approach to learn a behavior encoder that produces discretized tokens for behavior trajectories $\tau = \{o_0, a_0, \dots\}$ and an imitation learning policy decoder conditioned on these tokens. These additional behavior tokens will be augmented to the vocabulary of pretrained Multimodal Language Models. With this encoder, we then pack long-term multimodal interactions involving task instructions, memories, thoughts, observations, textual responses, behavior trajectories, etc into unified token sequences and model them with autoregressive transformers. Thanks to the semantically meaningful behavior tokens, the resulting VLA model, OmniJARVIS, can reason (by producing chain-of-thoughts), plan, answer questions, and act (by producing behavior tokens for the imitation learning policy decoder). OmniJARVIS demonstrates excellent performances on a comprehensive collection of atomic, programmatic, and open-ended tasks in open-world Minecraft. Our analysis further unveils the crucial design principles in interaction data formation, unified tokenization, and its scaling potentials. The dataset, models, and code will be released at https://craftjarvis.org/OmniJARVIS.

This paper presents UltraEdit, a large-scale (~ 4M editing samples), automatically generated dataset for instruction-based image editing. Our key idea is to address the drawbacks in existing image editing datasets like InstructPix2Pix and MagicBrush, and provide a systematic approach to producing massive and high-quality image editing samples: 1) UltraEdit includes more diverse editing instructions by combining LLM creativity and in-context editing examples by human raters; 2) UltraEdit is anchored on real images (photographs or artworks), which offers more diversity and less biases than those purely synthesized by text-to-image models; 3) UltraEdit supports region-based editing with high-quality, automatically produced region annotations. Our experiments show that canonical diffusion-based editing baselines trained on UltraEdit set new records on challenging MagicBrush and Emu-Edit benchmarks, respectively. Our analysis further confirms the crucial role of real image anchors and region-based editing data. The dataset, code, and models will be made public.

3D vision-language grounding (3D-VL) is an emerging field that aims to connect the 3D physical world with natural language, which is crucial for achieving embodied intelligence. Current 3D-VL models rely heavily on sophisticated modules, auxiliary losses, and optimization tricks, which calls for a simple and unified model. In this paper, we propose 3D-VisTA, a pre-trained Transformer for 3D Vision and Text Alignment that can be easily adapted to various downstream tasks. 3D-VisTA simply utilizes self-attention layers for both single-modal modeling and multi-modal fusion without any sophisticated task-specific design. To further enhance its performance on 3D-VL tasks, we construct ScanScribe, the first large-scale 3D scene-text pairs dataset for 3D-VL pre-training. ScanScribe contains 2,995 RGB-D scans for 1,185 unique indoor scenes originating from ScanNet and 3R-Scan datasets, along with paired 278K scene descriptions generated from existing 3D-VL tasks, templates, and GPT-3. 3D-VisTA is pre-trained on ScanScribe via masked language/object modeling and scene-text matching. It achieves state-of-the-art results on various 3D-VL tasks, ranging from visual grounding and dense captioning to question answering and situated reasoning. Moreover, 3D-VisTA demonstrates superior data efficiency, obtaining strong performance even with limited annotations during downstream task fine-tuning.

Conventional imitation learning assumes access to the actions of demonstrators, but these motor signals are often non-observable in naturalistic settings. Additionally, sequential decision-making behaviors in these settings can deviate from the assumptions of a standard Markov Decision Process (MDP). To address these challenges, we explore deep generative modeling of state-only sequences with non-Markov Decision Process (nMDP), where the policy is an energy-based prior in the latent space of the state transition generator. We develop maximum likelihood estimation to achieve model-based imitation, which involves short-run MCMC sampling from the prior and importance sampling for the posterior. The learned model enables $\textit{decision-making as inference}$: model-free policy execution is equivalent to prior sampling, model-based planning is posterior sampling initialized from the policy. We demonstrate the efficacy of the proposed method in a prototypical path planning task with non-Markovian constraints and show that the learned model exhibits strong performances in challenging domains from the MuJoCo suite.

Previous neural solvers of math word problems (MWPs) are learned with full supervision and fail to generate diverse solutions. In this paper, we address this issue by introducing a weakly-supervised paradigm for learning MWPs. Our method only requires the annotations of the final answers and can generate various solutions for a single problem. To boost weakly-supervised learning, we propose a novel learning-by-fixing (LBF) framework, which corrects the misperceptions of the neural network via symbolic reasoning. Specifically, for an incorrect solution tree generated by the neural network, the fixing mechanism propagates the error from the root node to the leaf nodes and infers the most probable fix that can be executed to get the desired answer. To generate more diverse solutions, tree regularization is applied to guide the efficient shrinkage and exploration of the solution space, and a memory buffer is designed to track and save the discovered various fixes for each problem. Experimental results on the Math23K dataset show the proposed LBF framework significantly outperforms reinforcement learning baselines in weakly-supervised learning. Furthermore, it achieves comparable top-1 and much better top-3/5 answer accuracies than fully-supervised methods, demonstrating its strength in producing diverse solutions.

Solving algebra story problems remains a challenging task in artificial intelligence, which requires a detailed understanding of real-world situations and a strong mathematical reasoning capability. Previous neural solvers of math word problems directly translate problem texts into equations, lacking an explicit interpretation of the situations, and often fail to handle more sophisticated situations. To address such limits of neural solvers, we introduce the concept of a situation model, which originates from psychology studies to represent the mental states of humans in problem-solving, and propose SMART, which adopts attributed grammar as the representation of situation models for algebra story problems. Specifically, we first train an information extraction module to extract nodes, attributes and relations from problem texts and then generate a parse graph based on a pre-defined attributed grammar. An iterative learning strategy is also proposed to further improve the performance of SMART. To study this task more rigorously, we carefully curate a new dataset named ASP6.6k. Experimental results on ASP6.6k show that the proposed model outperforms all previous neural solvers by a large margin, while preserving much better interpretability. To test these models' generalization capability, we also design an out-of-distribution (OOD) evaluation, in which problems are more complex than those in the training set. Our model exceeds state-of-the-art models by 17% in the OOD evaluation, demonstrating its superior generalization ability.

We study the machine’s understanding of embodied reference: One agent uses both language and gesture to refer to an object to another agent in a shared physical environment. Of note, this new visual task requires understanding multimodal cues with perspective-taking to identify which object is being referred to. To tackle this problem, we introduce YouRefIt, a new crowd-sourced dataset of embodied reference collected in various physical scenes; the dataset contains 4,195 unique reference clips in 432 indoor scenes. To the best of our knowledge, this is the first embodied reference dataset that allows us to study referring expressions in daily physical scenes to understand referential behavior, human communication, and human-robot interaction. We further devise two benchmarks for image-based and video-based embodied reference understanding. Comprehensive baselines and extensive experiments provide the very first result of machine perception on how the referring expressions and gestures affect the embodied reference understanding. Our results provide essential evidence that gestural cues are as critical as language cues in understanding the embodied reference.

Cognitive grammar suggests that the acquisition of language grammar is grounded within visual structures. While grammar is an essential representation of natural language, it also exists ubiquitously in vision to represent the hierarchical part-whole structure. In this work, we study grounded grammar induction of vision and language in a joint learning framework. Specifically, we present VLGrammar, a method that uses compound probabilistic context-free grammars (compound PCFGs) to induce the language grammar and the image grammar simultaneously. We propose a novel contrastive learning framework to guide the joint learning of both modules. To provide a benchmark for the grounded grammar induction task, we collect a large-scale dataset, PARTIT, which contains human-written sentences that describe part-level semantics for 3D objects. Experiments on the PARTIT dataset show that VLGrammar outperforms all baselines in image grammar induction and language grammar induction. The learned VLGrammar naturally benefits related downstream tasks. Specifically, it improves the image unsupervised clustering accuracy by 30%, and performs well in image retrieval and text retrieval. Notably, the induced grammar shows superior generalizability by easily generalizing to unseen categories. Code and pre-trained models are released at https://github.com/evelinehong/VLGrammar.

The goal of neural-symbolic computation is to integrate the connectionist and symbolist paradigms. Prior methods learn the neural-symbolic models using reinforcement learning (RL) approaches, which ignore the error propagation in the symbolic reasoning module and thus converge slowly with sparse rewards. In this paper, we address these issues and close the loop of neural-symbolic learning by (1) introducing the grammar model as a symbolic prior to bridge neural perception and symbolic reasoning, and (2) proposing a novel back-search algorithm which mimics the top-down human-like learning procedure to propagate the error through the symbolic reasoning module efficiently. We further interpret the proposed learning framework as maximum likelihood estimation using Markov chain Monte Carlo sampling and the back-search algorithm as a Metropolis-Hastings sampler. The experiments are conducted on two weakly-supervised neural-symbolic tasks: (1) handwritten formula recognition on the newly introduced HWF dataset; (2) visual question answering on the CLEVR dataset. The results show that our approach significantly outperforms the RL methods in terms of performance, converging speed, and data efficiency. Our code and data are released at https://liqing-ustc.github.io/NGS.

We introduce the first visual privacy dataset originating from people who are blind in order to better understand their privacy disclosures and to encourage the development of algorithms that can assist in preventing their unintended disclosures. It includes 8,862 regions showing private content across 5,537 images taken by blind people. Of these, 1,403 are paired with questions and 62\% of those directly ask about the private content. Experiments demonstrate the utility of this data for predicting whether an image shows private information and whether a question asks about the private content in an image. The dataset is publicly-shared at http://vizwiz.org/data/.

Visual question answering is the task of returning the answer to a question about an image. A challenge is that different people often provide different answers to the same visual question. To our knowledge, this is the first work that aims to understand why. We propose a taxonomy of nine plausible reasons, and create two labelled datasets consisting of ~45,000 visual questions indicating which reasons led to answer differences. We then propose a novel problem of predicting directly from a visual question which reasons will cause answer differences as well as a novel algorithm for this purpose. Experiments demonstrate the advantage of our approach over several related baselines on two diverse datasets. We publicly share the datasets and code at https://vizwiz.org.

The study of algorithms to automatically answer visual questions currently is motivated by visual question answering (VQA) datasets constructed in artificial VQA settings. We propose VizWiz, the first goal-oriented VQA dataset arising from a natural VQA setting. VizWiz consists of over 31,000 visual questions originating from blind people who each took a picture using a mobile phone and recorded a spoken question about it, together with 10 crowdsourced answers per visual question. VizWiz differs from the many existing VQA datasets because (1) images are captured by blind photographers and so are often poor quality, (2) questions are spoken and so are more conversational, and (3) often visual questions cannot be answered. Evaluation of modern algorithms for answering visual questions and deciding if a visual question is answerable reveals that VizWiz is a challenging dataset. We introduce this dataset to encourage a larger community to develop more generalized algorithms that can assist blind people.

In Visual Question Answering, most existing approaches adopt the pipeline of representing an image via pre-trained CNNs, and then using the uninterpretable CNN features in conjunction with the question to predict the answer.Although such end-to-end models might report promising performance, they rarely provide any insight, apart from the answer, into the VQA process.In this work, we propose to break up the end-to-end VQA into two steps: explaining and reasoning, in an attempt towards a more explainable VQA by shedding light on the intermediate results between these two steps.To that end, we first extract attributes and generate descriptions as explanations for an image.Next, a reasoning module utilizes these explanations in place of the image to infer an answer.The advantages of such a breakdown include: (1) the attributes and captions can reflect what the system extracts from the image, thus can provide some insights for the predicted answer; (2) these intermediate results can help identify the inabilities of the image understanding or the answer inference part when the predicted answer is wrong.We conduct extensive experiments on a popular VQA dataset and our system achieves comparable performance with the baselines, yet with added benefits of explanability and the inherent ability to further improve with higher quality explanations.

Recognizing actions in videos is a challenging task as video is an information-intensive media with complex variations. Most existing methods have treated video as a flat data sequence while ignoring the intrinsic hierarchical structure of the video content. In particular, an action may span different granularities in this hierarchy including, from small to large, a single frame, consecutive frames (motion), a short clip, and the entire video. In this paper, we present a novel framework to boost action recognition by learning a deep spatio-temporal video representation at hierarchical multi-granularity. Specifically, we model each granularity as a single stream by 2D (for frame and motion streams) or 3D (for clip and video streams) convolutional neural networks (CNNs). The framework therefore consists of multi-stream 2D or 3D CNNs to learn both the spatial and temporal representations. Furthermore, we employ the Long Short-Term Memory (LSTM) networks on the frame, motion, and clip streams to exploit long-term temporal dynamics. With a softmax layer on the top of each stream, the classification scores can be predicted from all the streams, followed by a novel fusion scheme based on the multi-granular score distribution. Our networks are learned in an end-to-end fashion. On two video action benchmarks of UCF101 and HMDB51, our framework achieves promising performance compared with the state-of-the-art.