Te-Lin Wu

Aarohi Srivastava, Abhinav Rastogi, Abhishek Rao et al. · allen-ai, amazon-science

Language models demonstrate both quantitative improvement and new qualitative capabilities with increasing scale. Despite their potentially transformative impact, these new capabilities are as yet poorly characterized. In order to inform future research, prepare for disruptive new model capabilities, and ameliorate socially harmful effects, it is vital that we understand the present and near-future capabilities and limitations of language models. To address this challenge, we introduce the Beyond the Imitation Game benchmark (BIG-bench). BIG-bench currently consists of 204 tasks, contributed by 450 authors across 132 institutions. Task topics are diverse, drawing problems from linguistics, childhood development, math, common-sense reasoning, biology, physics, social bias, software development, and beyond. BIG-bench focuses on tasks that are believed to be beyond the capabilities of current language models. We evaluate the behavior of OpenAI's GPT models, Google-internal dense transformer architectures, and Switch-style sparse transformers on BIG-bench, across model sizes spanning millions to hundreds of billions of parameters. In addition, a team of human expert raters performed all tasks in order to provide a strong baseline. Findings include: model performance and calibration both improve with scale, but are poor in absolute terms (and when compared with rater performance); performance is remarkably similar across model classes, though with benefits from sparsity; tasks that improve gradually and predictably commonly involve a large knowledge or memorization component, whereas tasks that exhibit "breakthrough" behavior at a critical scale often involve multiple steps or components, or brittle metrics; social bias typically increases with scale in settings with ambiguous context, but this can be improved with prompting.

Hong Chen, Rujun Han, Te-Lin Wu et al.

Story visualization advances the traditional text-to-image generation by enabling multiple image generation based on a complete story. This task requires machines to 1) understand long text inputs and 2) produce a globally consistent image sequence that illustrates the contents of the story. A key challenge of consistent story visualization is to preserve characters that are essential in stories. To tackle the challenge, we propose to adapt a recent work that augments Vector-Quantized Variational Autoencoders (VQ-VAE) with a text-tovisual-token (transformer) architecture. Specifically, we modify the text-to-visual-token module with a two-stage framework: 1) character token planning model that predicts the visual tokens for characters only; 2) visual token completion model that generates the remaining visual token sequence, which is sent to VQ-VAE for finalizing image generations. To encourage characters to appear in the images, we further train the two-stage framework with a character-token alignment objective. Extensive experiments and evaluations demonstrate that the proposed method excels at preserving characters and can produce higher quality image sequences compared with the strong baselines. Codes can be found in https://github.com/sairin1202/VP-CSV

Te-Lin Wu, Caiqi Zhang, Qingyuan Hu et al. · cambridge

The ability to infer pre- and postconditions of an action is vital for comprehending complex instructions, and is essential for applications such as autonomous instruction-guided agents and assistive AI that supports humans to perform physical tasks. In this work, we propose a task dubbed action condition inference, and collecting a high-quality, human annotated dataset of preconditions and postconditions of actions in instructional manuals. We propose a weakly supervised approach to automatically construct large-scale training instances from online instructional manuals, and curate a densely human-annotated and validated dataset to study how well the current NLP models can infer action-condition dependencies in the instruction texts. We design two types of models differ by whether contextualized and global information is leveraged, as well as various combinations of heuristics to construct the weak supervisions. Our experimental results show a >20% F1-score improvement with considering the entire instruction contexts and a >6% F1-score benefit with the proposed heuristics.

Te-Lin Wu, Yu Zhou, Nanyun Peng · meta-ai

The ability to actively ground task instructions from an egocentric view is crucial for AI agents to accomplish tasks or assist humans virtually. One important step towards this goal is to localize and track key active objects that undergo major state change as a consequence of human actions/interactions to the environment without being told exactly what/where to ground (e.g., localizing and tracking the `sponge` in video from the instruction "Dip the `sponge` into the bucket."). While existing works approach this problem from a pure vision perspective, we investigate to which extent the textual modality (i.e., task instructions) and their interaction with visual modality can be beneficial. Specifically, we propose to improve phrase grounding models' ability on localizing the active objects by: (1) learning the role of `objects undergoing change` and extracting them accurately from the instructions, (2) leveraging pre- and post-conditions of the objects during actions, and (3) recognizing the objects more robustly with descriptional knowledge. We leverage large language models (LLMs) to extract the aforementioned action-object knowledge, and design a per-object aggregation masking technique to effectively perform joint inference on object phrases and symbolic knowledge. We evaluate our framework on Ego4D and Epic-Kitchens datasets. Extensive experiments demonstrate the effectiveness of our proposed framework, which leads to>54% improvements in all standard metrics on the TREK-150-OPE-Det localization + tracking task, >7% improvements in all standard metrics on the TREK-150-OPE tracking task, and >3% improvements in average precision (AP) on the Ego4D SCOD task.

Xiaomeng Jin, Jeonghwan Kim, Yu Zhou et al. · meta-ai

In Multimodal Language Models (MLMs), the cost of manually annotating high-quality image-text pair data for fine-tuning and alignment is extremely high. While existing multimodal data augmentation frameworks propose ways to augment image-text pairs, they either suffer from semantic inconsistency between texts and images, or generate unrealistic images, causing knowledge gap with real world examples. To address these issues, we propose Attribute-based Multimodal Data Augmentation (ARMADA), a novel multimodal data augmentation method via knowledge-guided manipulation of visual attributes of the mentioned entities. Specifically, we extract entities and their visual attributes from the original text data, then search for alternative values for the visual attributes under the guidance of knowledge bases (KBs) and large language models (LLMs). We then utilize an image-editing model to edit the images with the extracted attributes. ARMADA is a novel multimodal data generation framework that: (i) extracts knowledge-grounded attributes from symbolic KBs for semantically consistent yet distinctive image-text pair generation, (ii) generates visually similar images of disparate categories using neighboring entities in the KB hierarchy, and (iii) uses the commonsense knowledge of LLMs to modulate auxiliary visual attributes such as backgrounds for more robust representation of original entities. Our empirical results over four downstream tasks demonstrate the efficacy of our framework to produce high-quality data and enhance the model performance. This also highlights the need to leverage external knowledge proxies for enhanced interpretability and real-world grounding.

Te-Lin Wu, Zi-Yi Dou, Qingyuan Hu et al.

Multimodal counterfactual reasoning is a vital yet challenging ability for AI systems. It involves predicting the outcomes of hypothetical circumstances based on vision and language inputs, which enables AI models to learn from failures and explore hypothetical scenarios. Despite its importance, there are only a few datasets targeting the counterfactual reasoning abilities of multimodal models. Among them, they only cover reasoning over synthetic environments or specific types of events (e.g. traffic collisions), making them hard to reliably benchmark the model generalization ability in diverse real-world scenarios and reasoning dimensions. To overcome these limitations, we develop a video question answering dataset, ACQUIRED: it consists of 3.9K annotated videos, encompassing a wide range of event types and incorporating both first and third-person viewpoints, which ensures a focus on real-world diversity. In addition, each video is annotated with questions that span three distinct dimensions of reasoning, including physical, social, and temporal, which can comprehensively evaluate the model counterfactual abilities along multiple aspects. We benchmark our dataset against several state-of-the-art language-only and multimodal models and experimental results demonstrate a significant performance gap (>13%) between models and humans. The findings suggest that multimodal counterfactual reasoning remains an open challenge and ACQUIRED is a comprehensive and reliable benchmark for inspiring future research in this direction.

Cho-Ying Wu, Quankai Gao, Chin-Cheng Hsu et al.

Indoor monocular depth estimation has attracted increasing research interest. Most previous works have been focusing on methodology, primarily experimenting with NYU-Depth-V2 (NYUv2) Dataset, and only concentrated on the overall performance over the test set. However, little is known regarding robustness and generalization when it comes to applying monocular depth estimation methods to real-world scenarios where highly varying and diverse functional \textit{space types} are present such as library or kitchen. A study for performance breakdown into space types is essential to realize a pretrained model's performance variance. To facilitate our investigation for robustness and address limitations of previous works, we collect InSpaceType, a high-quality and high-resolution RGBD dataset for general indoor environments. We benchmark 12 recent methods on InSpaceType and find they severely suffer from performance imbalance concerning space types, which reveals their underlying bias. We extend our analysis to 4 other datasets, 3 mitigation approaches, and the ability to generalize to unseen space types. Our work marks the first in-depth investigation of performance imbalance across space types for indoor monocular depth estimation, drawing attention to potential safety concerns for model deployment without considering space types, and further shedding light on potential ways to improve robustness. See \url{https://depthcomputation.github.io/DepthPublic} for data and the supplementary document. The benchmark list on the GitHub project page keeps updates for the lastest monocular depth estimation methods.

Xueqing Wu, Rui Zheng, Jingzhen Sha et al.

Data analysis is a crucial analytical process to generate in-depth studies and conclusive insights to comprehensively answer a given user query for tabular data. In this work, we aim to propose new resources and benchmarks to inspire future research on this crucial yet challenging and under-explored task. However, collecting data analysis annotations curated by experts can be prohibitively expensive. We propose to automatically generate high-quality answer annotations leveraging the code-generation capabilities of LLMs with a multi-turn prompting technique. We construct the DACO dataset, containing (1) 440 databases (of tabular data) collected from real-world scenarios, (2) ~2k query-answer pairs that can serve as weak supervision for model training, and (3) a concentrated but high-quality test set with human refined annotations that serves as our main evaluation benchmark. We train a 6B supervised fine-tuning (SFT) model on DACO dataset, and find that the SFT model learns reasonable data analysis capabilities. To further align the models with human preference, we use reinforcement learning to encourage generating analysis perceived by human as helpful, and design a set of dense rewards to propagate the sparse human preference reward to intermediate code generation steps. Our DACO-RL algorithm is evaluated by human annotators to produce more helpful answers than SFT model in 57.72% cases, validating the effectiveness of our proposed algorithm. Data and code are released at https://github.com/shirley-wu/daco

Cho-Ying Wu, Quankai Gao, Chin-Cheng Hsu et al.

Indoor monocular depth estimation helps home automation, including robot navigation or AR/VR for surrounding perception. Most previous methods primarily experiment with the NYUv2 Dataset and concentrate on the overall performance in their evaluation. However, their robustness and generalization to diversely unseen types or categories for indoor spaces (spaces types) have yet to be discovered. Researchers may empirically find degraded performance in a released pretrained model on custom data or less-frequent types. This paper studies the common but easily overlooked factor-space type and realizes a model's performance variances across spaces. We present InSpaceType Dataset, a high-quality RGBD dataset for general indoor scenes, and benchmark 13 recent state-of-the-art methods on InSpaceType. Our examination shows that most of them suffer from performance imbalance between head and tailed types, and some top methods are even more severe. The work reveals and analyzes underlying bias in detail for transparency and robustness. We extend the analysis to a total of 4 datasets and discuss the best practice in synthetic data curation for training indoor monocular depth. Further, dataset ablation is conducted to find out the key factor in generalization. This work marks the first in-depth investigation of performance variances across space types and, more importantly, releases useful tools, including datasets and codes, to closely examine your pretrained depth models. Data and code: https://depthcomputation.github.io/DepthPublic/

Xueqing Wu, Zongyu Lin, Songyan Zhao et al.

Visual programs are executable code generated by large language models to address visual reasoning problems. They decompose complex questions into multiple reasoning steps and invoke specialized models for each step to solve the problems. However, these programs are prone to logic errors, with our preliminary evaluation showing that 58% of the total errors are caused by program logic errors. Debugging complex visual programs remains a major bottleneck for visual reasoning. To address this, we introduce VDebugger, a novel critic-refiner framework trained to localize and debug visual programs by tracking execution step by step. VDebugger identifies and corrects program errors leveraging detailed execution feedback, improving interpretability and accuracy. The training data is generated through an automated pipeline that injects errors into correct visual programs using a novel mask-best decoding technique. Evaluations on six datasets demonstrate VDebugger's effectiveness, showing performance improvements of up to 3.2% in downstream task accuracy. Further studies show VDebugger's ability to generalize to unseen tasks, bringing a notable improvement of 2.3% on the unseen COVR task. Code, data and models are made publicly available at https://github.com/shirley-wu/vdebugger/

Shikhar Singh, Nuan Wen, Yu Hou et al.

Commonsense reasoning is intuitive for humans but has been a long-term challenge for artificial intelligence (AI). Recent advancements in pretrained language models have shown promising results on several commonsense benchmark datasets. However, the reliability and comprehensiveness of these benchmarks towards assessing model's commonsense reasoning ability remains unclear. To this end, we introduce a new commonsense reasoning benchmark dataset comprising natural language true/false statements, with each sample paired with its complementary counterpart, resulting in 4k sentence pairs. We propose a pairwise accuracy metric to reliably measure an agent's ability to perform commonsense reasoning over a given situation. The dataset is crowdsourced and enhanced with an adversarial model-in-the-loop setup to incentivize challenging samples. To facilitate a systematic analysis of commonsense capabilities, we design our dataset along the dimensions of knowledge domains, reasoning scenarios and numeracy. Experimental results demonstrate that our strongest baseline (UnifiedQA-3B), after fine-tuning, achieves ~71% standard accuracy and ~51% pairwise accuracy, well below human performance (~95% for both metrics). The dataset is available at https://github.com/PlusLabNLP/Com2Sense.

Yu Zhou, Bingxuan Li, Mohan Tang et al. · meta-ai

Large multimodal models (LMMs) often struggle to recognize novel concepts, as they rely on pre-trained knowledge and have limited ability to capture subtle visual details. Domain-specific knowledge gaps in training also make them prone to confusing visually similar, commonly misrepresented, or low-resource concepts. To help LMMs better align nuanced visual features with language, improving their ability to recognize and reason about novel or rare concepts, we propose a Contrastive visual Data Augmentation (CoDA) strategy. CoDA extracts key contrastive textual and visual features of target concepts against the known concepts they are misrecognized as, and then uses multimodal generative models to produce targeted synthetic data. Automatic filtering of extracted features and augmented images is implemented to guarantee their quality, as verified by human annotators. We show the effectiveness and efficiency of CoDA on low-resource concept and diverse scene recognition datasets including INaturalist and SUN. We additionally collect NovelSpecies, a benchmark dataset consisting of newly discovered animal species that are guaranteed to be unseen by LMMs. LLaVA-1.6 1-shot updating results on these three datasets show CoDA significantly improves SOTA visual data augmentation strategies by 12.3% (NovelSpecies), 5.1% (SUN), and 6.0% (iNat) absolute gains in accuracy.

Te-Lin Wu, Alex Spangher, Pegah Alipoormolabashi et al.

The ability to sequence unordered events is an essential skill to comprehend and reason about real world task procedures, which often requires thorough understanding of temporal common sense and multimodal information, as these procedures are often communicated through a combination of texts and images. Such capability is essential for applications such as sequential task planning and multi-source instruction summarization. While humans are capable of reasoning about and sequencing unordered multimodal procedural instructions, whether current machine learning models have such essential capability is still an open question. In this work, we benchmark models' capability of reasoning over and sequencing unordered multimodal instructions by curating datasets from popular online instructional manuals and collecting comprehensive human annotations. We find models not only perform significantly worse than humans but also seem incapable of efficiently utilizing the multimodal information. To improve machines' performance on multimodal event sequencing, we propose sequentiality-aware pretraining techniques that exploit the sequential alignment properties of both texts and images, resulting in > 5% significant improvements.

Mingyu Derek Ma, Muhao Chen, Te-Lin Wu et al.

Taxonomies are valuable resources for many applications, but the limited coverage due to the expensive manual curation process hinders their general applicability. Prior works attempt to automatically expand existing taxonomies to improve their coverage by learning concept embeddings in Euclidean space, while taxonomies, inherently hierarchical, more naturally align with the geometric properties of a hyperbolic space. In this paper, we present HyperExpan, a taxonomy expansion algorithm that seeks to preserve the structure of a taxonomy in a more expressive hyperbolic embedding space and learn to represent concepts and their relations with a Hyperbolic Graph Neural Network (HGNN). Specifically, HyperExpan leverages position embeddings to exploit the structure of the existing taxonomies, and characterizes the concept profile information to support the inference on unseen concepts during training. Experiments show that our proposed HyperExpan outperforms baseline models with representation learning in a Euclidean feature space and achieves state-of-the-art performance on the taxonomy expansion benchmarks.

Te-Lin Wu, Cheng Li, Mingyang Zhang et al.

Document layout comprises both structural and visual (eg. font-sizes) information that is vital but often ignored by machine learning models. The few existing models which do use layout information only consider textual contents, and overlook the existence of contents in other modalities such as images. Additionally, spatial interactions of presented contents in a layout were never really fully exploited. To bridge this gap, we parse a document into content blocks (eg. text, table, image) and propose a novel layout-aware multimodal hierarchical framework, LAMPreT, to model the blocks and the whole document. Our LAMPreT encodes each block with a multimodal transformer in the lower-level and aggregates the block-level representations and connections utilizing a specifically designed transformer at the higher-level. We design hierarchical pretraining objectives where the lower-level model is trained similarly to multimodal grounding models, and the higher-level model is trained with our proposed novel layout-aware objectives. We evaluate the proposed model on two layout-aware tasks -- text block filling and image suggestion and show the effectiveness of our proposed hierarchical architecture as well as pretraining techniques.

Te-Lin Wu, Shikhar Singh, Sayan Paul et al.

We introduce a new dataset, MELINDA, for Multimodal biomEdicaL experImeNt methoD clAssification. The dataset is collected in a fully automated distant supervision manner, where the labels are obtained from an existing curated database, and the actual contents are extracted from papers associated with each of the records in the database. We benchmark various state-of-the-art NLP and computer vision models, including unimodal models which only take either caption texts or images as inputs, and multimodal models. Extensive experiments and analysis show that multimodal models, despite outperforming unimodal ones, still need improvements especially on a less-supervised way of grounding visual concepts with languages, and better transferability to low resource domains. We release our dataset and the benchmarks to facilitate future research in multimodal learning, especially to motivate targeted improvements for applications in scientific domains.

Amir R. Zamir, Te-Lin Wu, Lin Sun et al.

Currently, the most successful learning models in computer vision are based on learning successive representations followed by a decision layer. This is usually actualized through feedforward multilayer neural networks, e.g. ConvNets, where each layer forms one of such successive representations. However, an alternative that can achieve the same goal is a feedback based approach in which the representation is formed in an iterative manner based on a feedback received from previous iteration's output. We establish that a feedback based approach has several fundamental advantages over feedforward: it enables making early predictions at the query time, its output naturally conforms to a hierarchical structure in the label space (e.g. a taxonomy), and it provides a new basis for Curriculum Learning. We observe that feedback networks develop a considerably different representation compared to feedforward counterparts, in line with the aforementioned advantages. We put forth a general feedback based learning architecture with the endpoint results on par or better than existing feedforward networks with the addition of the above advantages. We also investigate several mechanisms in feedback architectures (e.g. skip connections in time) and design choices (e.g. feedback length). We hope this study offers new perspectives in quest for more natural and practical learning models.