Existing works on long-term open-domain dialogues focus on evaluating model responses within contexts spanning no more than five chat sessions. Despite advancements in long-context large language models (LLMs) and retrieval augmented generation (RAG) techniques, their efficacy in very long-term dialogues remains unexplored. To address this research gap, we introduce a machine-human pipeline to generate high-quality, very long-term dialogues by leveraging LLM-based agent architectures and grounding their dialogues on personas and temporal event graphs. Moreover, we equip each agent with the capability of sharing and reacting to images. The generated conversations are verified and edited by human annotators for long-range consistency and grounding to the event graphs. Using this pipeline, we collect LoCoMo, a dataset of very long-term conversations, each encompassing 600 turns and 16K tokens on avg., over up to 32 sessions. Based on LoCoMo, we present a comprehensive evaluation benchmark to measure long-term memory in models, encompassing question answering, event summarization, and multi-modal dialogue generation tasks. Our experimental results indicate that LLMs exhibit challenges in understanding lengthy conversations and comprehending long-range temporal and causal dynamics within dialogues. Employing strategies like long-context LLMs or RAG can offer improvements but these models still substantially lag behind human performance.

Most existing debiasing methods for multimodal models, including causal intervention and inference methods, utilize approximate heuristics to represent the biases, such as shallow features from early stages of training or unimodal features for multimodal tasks like VQA, etc., which may not be accurate. In this paper, we study bias arising from confounders in a causal graph for multimodal data, and examine a novel approach that leverages causally-motivated information minimization to learn the confounder representations. Robust predictive features contain diverse information that helps a model generalize to out-of-distribution data. Hence, minimizing the information content of features obtained from a pretrained biased model helps learn the simplest predictive features that capture the underlying data distribution. We treat these features as confounder representations and use them via methods motivated by causal theory to remove bias from models. We find that the learned confounder representations indeed capture dataset biases and the proposed debiasing methods improve out-of-distribution (OOD) performance on multiple multimodal datasets without sacrificing in-distribution performance. Additionally, we introduce a novel metric to quantify the sufficiency of spurious features in models’ predictions that further demonstrates the effectiveness of our proposed methods.

Commonsense reasoning tasks follow a standard paradigm of finetuning pretrained language models on the target task data, where samples are introduced to the model in a random order during training. However, recent research suggests that data order can have a significant impact on the performance of finetuned models for natural language understanding. Hence, we examine the effect of a human-like easy-to-difficult curriculum during finetuning of language models for commonsense reasoning tasks. We use paced curriculum learning to rank data and sample training mini-batches with increasing levels of difficulty from the ranked dataset during finetuning. Further, we investigate the effect of an adaptive curriculum, i.e., the data ranking is dynamically updated during training based on the current state of the learner model. We use a teacher model to measure difficulty of each sample and experiment with three measures based on question answering probability, variability and out-of-distribution. To understand the effectiveness of curriculum learning in various scenarios, we apply it on full model fine-tuning as well as parameter-efficient prompt-tuning settings. Our results show that fixed as well as adaptive curriculum learning significantly improve performance for five commonsense reasoning tasks, i.e., SocialIQA, CosmosQA, CODAH, HellaSwag, WinoGrande in both tuning settings. Further, we find that prioritizing the difficult samples in the tail end of training improves generalization to unseen in-domain data as well as out-of-domain data. Our work provides evidence and encourages research into curriculum learning for commonsense reasoning.

Recent advances in commonsense reasoning have been fueled by the availability of large-scale human annotated datasets. Manual annotation of such datasets, many of which are based on existing knowledge bases, is expensive and not scalable. Moreover, it is challenging to build augmentation data for commonsense reasoning because the synthetic questions need to adhere to real-world scenarios. Hence, we present GraDA, a graph-generative data augmentation framework to synthesize factual data samples from knowledge graphs for commonsense reasoning datasets. First, we train a graph-to-text model for conditional generation of questions from graph entities and relations. Then, we train a generator with GAN loss to generate distractors for synthetic questions. Our approach improves performance for SocialIQA, CODAH, HellaSwag and CommonsenseQA, and works well for generative tasks like ProtoQA. We show improvement in robustness to semantic adversaries after training with GraDA and provide human evaluation of the quality of synthetic datasets in terms of factuality and answerability. Our work provides evidence and encourages future research into graph-based generative data augmentation.

Individuals, educational institutions, and businesses are prolific at generating instructional video content such as “how-to” and tutorial guides. While significant progress has been made in basic video understanding tasks, identifying procedural intent within these instructional videos is a challenging and important task that remains unexplored but essential to video summarization, search, and recommendations. This paper introduces the problem of instructional intent identification and extraction from software instructional livestreams. We construct and present a new multimodal dataset consisting of software instructional livestreams and containing manual annotations for both detailed and abstract procedural intent that enable training and evaluation of joint video and text understanding models. We then introduce a multimodal cascaded cross-attention model to efficiently combine the weaker and noisier video signal with the more discriminative text signal. Our experiments show that our proposed model brings significant gains compared to strong baselines, including large-scale pretrained multimodal models. Our analysis further identifies that the task benefits from spatial as well as motion features extracted from videos, and provides insight on how the video signal is preferentially used for intent discovery. We also show that current models struggle to comprehend the nature of abstract intents, revealing important gaps in multimodal understanding and paving the way for future work.

Recent advances in commonsense reasoning have been fueled by the availability of large-scale human annotated datasets. Manual annotation of such datasets, many of which are based on existing knowledge bases, is expensive and not scalable. Moreover, it is challenging to build augmentation data for commonsense reasoning because the synthetic questions need to adhere to real-world scenarios. Hence, we present GraDA, a graph-generative data augmentation framework to synthesize factual data samples from knowledge graphs for commonsense reasoning datasets. First, we train a graph-to-text model for conditional generation of questions from graph entities and relations. Then, we train a generator with GAN loss to generate distractors for synthetic questions. Our approach improves performance for SocialIQA, CODAH, HellaSwag and CommonsenseQA, and works well for generative tasks like ProtoQA. We show improvement in robustness to semantic adversaries after training with GraDA and provide human evaluation of the quality of synthetic datasets in terms of factuality and answerability. Our work provides evidence and encourages future research into graph-based generative data augmentation.

Story visualization is an underexplored task that falls at the intersection of many important research directions in both computer vision and natural language processing. In this task, given a series of natural language captions which compose a story, an agent must generate a sequence of images that correspond to the captions. Prior work has introduced recurrent generative models which outperform text-to-image synthesis models on this task. However, there is room for improvement of generated images in terms of visual quality, coherence and relevance. We present a number of improvements to prior modeling approaches, including (1) the addition of a dual learning framework that utilizes video captioning to reinforce the semantic alignment between the story and generated images, (2) a copy-transform mechanism for sequentially-consistent story visualization, and (3) MART-based transformers to model complex interactions between frames. We present ablation studies to demonstrate the effect of each of these techniques on the generative power of the model for both individual images as well as the entire narrative. Furthermore, due to the complexity and generative nature of the task, standard evaluation metrics do not accurately reflect performance. Therefore, we also provide an exploration of evaluation metrics for the model, focused on aspects of the generated frames such as the presence/quality of generated characters, the relevance to captions, and the diversity of the generated images. We also present correlation experiments of our proposed automated metrics with human evaluations.

While much research has been done in text-to-image synthesis, little work has been done to explore the usage of linguistic structure of the input text. Such information is even more important for story visualization since its inputs have an explicit narrative structure that needs to be translated into an image sequence (or visual story). Prior work in this domain has shown that there is ample room for improvement in the generated image sequence in terms of visual quality, consistency and relevance. In this paper, we first explore the use of constituency parse trees using a Transformer-based recurrent architecture for encoding structured input. Second, we augment the structured input with commonsense information and study the impact of this external knowledge on the generation of visual story. Third, we also incorporate visual structure via bounding boxes and dense captioning to provide feedback about the characters/objects in generated images within a dual learning setup. We show that off-the-shelf dense-captioning models trained on Visual Genome can improve the spatial structure of images from a different target domain without needing fine-tuning. We train the model end-to-end using intra-story contrastive loss (between words and image sub-regions) and show significant improvements in visual quality. Finally, we provide an analysis of the linguistic and visuo-spatial information.

Reading comprehension models often overfit to nuances of training datasets and fail at adversarial evaluation. Training with adversarially augmented dataset improves robustness against those adversarial attacks but hurts generalization of the models. In this work, we present several effective adversaries and automated data augmentation policy search methods with the goal of making reading comprehension models more robust to adversarial evaluation, but also improving generalization to the source domain as well as new domains and languages. We first propose three new methods for generating QA adversaries, that introduce multiple points of confusion within the context, show dependence on insertion location of the distractor, and reveal the compounding effect of mixing adversarial strategies with syntactic and semantic paraphrasing methods. Next, we find that augmenting the training datasets with uniformly sampled adversaries improves robustness to the adversarial attacks but leads to decline in performance on the original unaugmented dataset. We address this issue via RL and more efficient Bayesian policy search methods for automatically learning the best augmentation policy combinations of the transformation probability for each adversary in a large search space. Using these learned policies, we show that adversarial training can lead to significant improvements in in-domain, out-of-domain, and cross-lingual (German, Russian, Turkish) generalization.

Event detection from clinical notes has been traditionally solved with rule based and statistical natural language processing (NLP) approaches that require extensive domain knowledge and feature engineering. In this paper, we have explored the feasibility of approaching this task with recurrent neural networks, clinical word embeddings and introduced a hybrid architecture to improve detection for entities with smaller representation in the dataset. A comparative analysis is also done which reveals the complementary behavior of neural networks and conditional random fields in clinical entity detection.