Reasoning over spans of tokens from different parts of the input is essential for natural language understanding (NLU) tasks such as fact-checking (FC), machine reading comprehension (MRC) or natural language inference (NLI). However, existing highlight-based explanations primarily focus on identifying individual important tokens or interactions only between adjacent tokens or tuples of tokens. Most notably, there is a lack of annotations capturing the human decision-making process w.r.t. the necessary interactions for informed decision-making in such tasks. To bridge this gap, we introduce SpanEx, a multi-annotator dataset of human span interaction explanations for two NLU tasks: NLI and FC. We then investigate the decision-making processes of multiple fine-tuned large language models in terms of the employed connections between spans in separate parts of the input and compare them to the human reasoning processes. Finally, we present a novel community detection based unsupervised method to extract such interaction explanations from a model's inner workings.

In the field of natural language processing, open-domain chatbots have emerged as an important research topic. However, a major limitation of existing open-domain chatbot research is its singular focus on short single-session dialogue, neglecting the potential need for understanding contextual information in multiple consecutive sessions that precede an ongoing dialogue. Among the elements that compose the context in multi-session conversation settings, the time intervals between sessions and the relationships between speakers would be particularly important. Despite their importance, current research efforts have not sufficiently addressed these dialogical components. In this paper, we introduce a new 1M multi-session dialogue dataset, called Conversation Chronicles, for implementing a long-term conversation setup in which time intervals and fine-grained speaker relationships are incorporated. Following recent works, we exploit a large language model to produce the data. The extensive human evaluation shows that dialogue episodes in Conversation Chronicles reflect those properties while maintaining coherent and consistent interactions across all the sessions. We also propose a dialogue model, called ReBot, which consists of chronological summarization and dialogue generation modules using only around 630M parameters. When trained on Conversation Chronicles, ReBot demonstrates long-term context understanding with a high human engagement score.

Prompting Large Language Models (LLMs) performs impressively in zero- and few-shot settings. Hence, small and medium-sized enterprises (SMEs) that cannot afford the cost of creating large task-specific training datasets, but also the cost of pretraining their own LLMs, are increasingly turning to third-party services that allow them to prompt LLMs. However, such services currently require a payment per call, which becomes a significant operating expense (OpEx). Furthermore, customer inputs are often very similar over time, hence SMEs end-up prompting LLMs with very similar instances. We propose a framework that allows reducing the calls to LLMs by caching previous LLM responses and using them to train a local inexpensive model on the SME side. The framework includes criteria for deciding when to trust the local model or call the LLM, and a methodology to tune the criteria and measure the tradeoff between performance and cost. For experimental purposes, we instantiate our framework with two LLMs, GPT-3.5 or GPT-4, and two inexpensive students, a k-NN classifier or a Multi-Layer Perceptron, using two common business tasks, intent recognition and sentiment analysis. Experimental results indicate that significant OpEx savings can be obtained with only slightly lower performance.

This paper presents a novel human-robot interaction setup for robot and human learning of symbolic language for identifying robot homeostatic needs. The robot and human learn to use and respond to the same language symbols that convey homeostatic needs and the stimuli that satisfy the homeostatic needs, respectively. We adopted a differential outcomes training (DOT) protocol whereby the robot provides feedback specific (differential) to its internal needs (e.g. `hunger') when satisfied by the correct stimulus (e.g. cookie). We found evidence that DOT can enhance the human's learning efficiency, which in turn enables more efficient robot language acquisition. The robot used in the study has a vocabulary similar to that of a human infant in the linguistic ``babbling'' phase. The robot software architecture is built upon a model for affect-grounded language acquisition where the robot associates vocabulary with internal needs (hunger, thirst, curiosity) through interactions with the human. The paper presents the results of an initial pilot study conducted with the interactive setup, which reveal that the robot's language acquisition achieves higher convergence rate in the DOT condition compared to the non-DOT control condition. Additionally, participants reported positive affective experiences, feeling of being in control, and an empathetic connection with the robot. This mutual learning (teacher-student learning) approach offers a potential contribution of facilitating cognitive interventions with DOT (e.g. for people with dementia) through increased therapy adherence as a result of engaging humans more in training tasks by taking an active teaching-learning role. The homeostatic motivational grounding of the robot's language acquisition has potential to contribute to more ecologically valid and social (collaborative/nurturing) interactions with robots.

The application of deep learning to nursing procedure activity understanding has the potential to greatly enhance the quality and safety of nurse-patient interactions. By utilizing the technique, we can facilitate training and education, improve quality control, and enable operational compliance monitoring. However, the development of automatic recognition systems in this field is currently hindered by the scarcity of appropriately labeled datasets. The existing video datasets pose several limitations: 1) these datasets are small-scale in size to support comprehensive investigations of nursing activity; 2) they primarily focus on single procedures, lacking expert-level annotations for various nursing procedures and action steps; and 3) they lack temporally localized annotations, which prevents the effective localization of targeted actions within longer video sequences. To mitigate these limitations, we propose NurViD, a large video dataset with expert-level annotation for nursing procedure activity understanding. NurViD consists of over 1.5k videos totaling 144 hours, making it approximately four times longer than the existing largest nursing activity datasets. Notably, it encompasses 51 distinct nursing procedures and 177 action steps, providing a much more comprehensive coverage compared to existing datasets that primarily focus on limited procedures. To evaluate the efficacy of current deep learning methods on nursing activity understanding, we establish three benchmarks on NurViD: procedure recognition on untrimmed videos, procedure and action recognition on trimmed videos, and action detection. Our benchmark and code will be available at \url{https://github.com/minghu0830/NurViD-benchmark}.

With the development of large language models (LLMs) like the GPT series, their widespread use across various application scenarios presents a myriad of challenges. This review initially explores the issue of domain specificity, where LLMs may struggle to provide precise answers to specialized questions within niche fields. The problem of knowledge forgetting arises as these LLMs might find it hard to balance old and new information. The knowledge repetition phenomenon reveals that sometimes LLMs might deliver overly mechanized responses, lacking depth and originality. Furthermore, knowledge illusion describes situations where LLMs might provide answers that seem insightful but are actually superficial, while knowledge toxicity focuses on harmful or biased information outputs. These challenges underscore problems in the training data and algorithmic design of LLMs. To address these issues, it's suggested to diversify training data, fine-tune models, enhance transparency and interpretability, and incorporate ethics and fairness training. Future technological trends might lean towards iterative methodologies, multimodal learning, model personalization and customization, and real-time learning and feedback mechanisms. In conclusion, future LLMs should prioritize fairness, transparency, and ethics, ensuring they uphold high moral and ethical standards when serving humanity.

Question answering over hybrid contexts is a complex task, which requires the combination of information extracted from unstructured texts and structured tables in various ways. Recently, In-Context Learning demonstrated significant performance advances for reasoning tasks. In this paradigm, a large language model performs predictions based on a small set of supporting exemplars. The performance of In-Context Learning depends heavily on the selection procedure of the supporting exemplars, particularly in the case of HybridQA, where considering the diversity of reasoning chains and the large size of the hybrid contexts becomes crucial. In this work, we present Selection of ExEmplars for hybrid Reasoning (SEER), a novel method for selecting a set of exemplars that is both representative and diverse. The key novelty of SEER is that it formulates exemplar selection as a Knapsack Integer Linear Program. The Knapsack framework provides the flexibility to incorporate diversity constraints that prioritize exemplars with desirable attributes, and capacity constraints that ensure that the prompt size respects the provided capacity budgets. The effectiveness of SEER is demonstrated on FinQA and TAT-QA, two real-world benchmarks for HybridQA, where it outperforms previous exemplar selection methods.

Machine learning (ML) has become a popular tool in the industrial sector as it helps to improve operations, increase efficiency, and reduce costs. However, deploying and managing ML models in production environments can be complex. This is where Machine Learning Operations (MLOps) comes in. MLOps aims to streamline this deployment and management process. One of the remaining MLOps challenges is the need for explanations. These explanations are essential for understanding how ML models reason, which is key to trust and acceptance. Better identification of errors and improved model accuracy are only two resulting advantages. An often neglected fact is that deployed models are bypassed in practice when accuracy and especially explainability do not meet user expectations. We developed a novel MLOps software architecture to address the challenge of integrating explanations and feedback capabilities into the ML development and deployment processes. In the project EXPLAIN, our architecture is implemented in a series of industrial use cases. The proposed MLOps software architecture has several advantages. It provides an efficient way to manage ML models in production environments. Further, it allows for integrating explanations into the development and deployment processes.

We introduce a new neural architecture for solving visual abstract reasoning tasks inspired by human cognition, specifically by observations that human abstract reasoning often interleaves perceptual and conceptual processing as part of a flexible, iterative, and dynamic cognitive process. Inspired by this principle, our architecture models visual abstract reasoning as an iterative, self-contrasting learning process that pursues consistency between perceptual and conceptual processing of visual stimuli. We explain how this new Contrastive Perceptual-Conceptual Network (CPCNet) works using matrix reasoning problems in the style of the well-known Raven's Progressive Matrices intelligence test. Experiments on the machine learning dataset RAVEN show that CPCNet achieves higher accuracy than all previously published models while also using the weakest inductive bias. We also point out a substantial and previously unremarked class imbalance in the original RAVEN dataset, and we propose a new variant of RAVEN -- AB-RAVEN -- that is more balanced in terms of abstract concepts.

Imitation learning has been applied to a range of robotic tasks, but can struggle when robots encounter edge cases that are not represented in the training data (i.e., distribution shift). Interactive fleet learning (IFL) mitigates distribution shift by allowing robots to access remote human supervisors during task execution and learn from them over time, but different supervisors may demonstrate the task in different ways. Recent work proposes Implicit Behavior Cloning (IBC), which is able to represent multimodal demonstrations using energy-based models (EBMs). In this work, we propose Implicit Interactive Fleet Learning (IIFL), an algorithm that builds on IBC for interactive imitation learning from multiple heterogeneous human supervisors. A key insight in IIFL is a novel approach for uncertainty quantification in EBMs using Jeffreys divergence. While IIFL is more computationally expensive than explicit methods, results suggest that IIFL achieves a 2.8x higher success rate in simulation experiments and a 4.5x higher return on human effort in a physical block pushing task over (Explicit) IFL, IBC, and other baselines.