This paper presents a novel approach named Contextually Relevant Imputation leveraging pre-trained Language Models (CRILM) for handling missing data in tabular datasets. Instead of relying on traditional numerical estimations, CRILM uses pre-trained language models (LMs) to create contextually relevant descriptors for missing values. This method aligns datasets with LMs’ strengths, allowing large LMs to generate these descriptors and small LMs to be fine-tuned on the enriched datasets for enhanced downstream task performance. Our evaluations demonstrate CRILM’s superior performance and robustness across MCAR, MAR, and challenging MNAR scenarios, with up to a 10% improvement over the best-performing baselines. By mitigating biases, particularly in MNAR settings, CRILM improves downstream task performance and offers a cost-effective solution for resource-constrained environments.

Vision-language models (VLMs) excel in image understanding but often hallucinate, generating descriptions of non-existent objects. While prior work has addressed static object hallucination, we identify a distinct, understudied failure mode: hypothetical object hallucination (HOH), where VLMs misjudge whether an object could plausibly appear under altered temporal or contextual conditions. This phenomenon, related to counterfactual reasoning, exposes gaps in dynamic multimodal reasoning. To mitigate HOH, we propose a training-free, prompt-based meta-reasoning approach that leverages contextual cues to guide VLMs toward deliberate reasoning. Evaluations across 12 diverse VLMs reveal variable performance, sensitivity to query phrasing, and inconsistent effectiveness of meta-information. While our method enhances reasoning accuracy in several models, its variable impact underscores the fragility of prompt-based interventions. By exposing these complexities, our work deepens the understanding of VLM failure modes and emphasizes the challenges in achieving robust multimodal reasoning.

Despite the effort toward benchmarking self-supervised learning (SSL) methods for image recognition transfer learning tasks, our understanding is still limited about their performance on noisy real-world datasets. This paper presents an extensive analysis of various types of SSL methods on real-world datasets containing noisy images of wildlife animals. These uncurated images are auto-captured by motion-activated cameras or camera traps installed in the wild. The camera-trap datasets exhibit different types of biases typically present in practical tasks. Using a set of variably-size biased datasets, we compare the supervised learning (SL) method to two types of SSL methods, i.e., instance discrimination and cluster discrimination. Our results reveal nuances in SSL’s performance. For example, we show that though SSL methods are often more generalizable than the SL method, the performance gain of some SSL methods diminishes with the reduction in the size of the target dataset. Also, there exists significant variability in the effectiveness of the two types of SSL methods. In addition to this, we show that, unlike SL, both types of SSL gain from increased model capacity.

Irrespective of the success of the deep learningbased mixed-domain transfer learning approach for solving various Natural Language Processing tasks, it does not lend a generalizable solution for detecting misinformation from COVID-19 social media data.Due to the inherent complexity of this type of data, caused by its dynamic (context evolves rapidly), nuanced (misinformation types are often ambiguous), and diverse (skewed, finegrained, and overlapping categories) nature, it is imperative for an effective model to capture both the local and global context of the target domain.By conducting a systematic investigation, we show that: (i) the deep Transformerbased pre-trained models, utilized via the mixed-domain transfer learning, are only good at capturing the local context, thus exhibits poor generalization, and (ii) a combination of shallow network-based domain-specific models and convolutional neural networks can efficiently extract local as well as global context directly from the target data in a hierarchical fashion, enabling it to offer a more generalizable solution.

In this paper, we design a distributed mechanism that is able to create a social convention within a large convention space for multiagent systems (MAS) operating on various topologies. Specifically, we investigate a language coordination problem in which agents in a dynamic MAS construct a common lexicon in a decentralized fashion. Agent interactions are modeled using a language game where every agent repeatedly plays with its neighbors. Each agent stochastically updates its lexicons based on the utility values of the received lexicons from its immediate neighbors. We present a novel topology-aware utility computation mechanism and equip the agents with the ability to reorganize their neighborhood based on this utility estimate to expedite the convention formation process. Extensive simulation results indicate that our proposed mechanism is both effective (able to converge into a large majority convention state with more than 90\% agents sharing a high-quality lexicon) and efficient (faster) as compared to state-of-the-art approaches for social conventions in large convention spaces.

Emergence of a single coalition among self-interested agents operating on large scale-free networks is a challenging task. Many existing approaches assume a given static network platform and do not use the network dynamics to facilitate the dynamics of agent interactions. In this paper, we present a decentralized game-theoretic approach to this single coalition emergence problem in which agent communications are limited only to their immediate neighbors. Our coalition emergence algorithm is based on the heuristic that agents benefit by forming coalitions with wealthy (higher payoff) and influential (higher accumulated coupling strength) neighbors. Simulation results show that the emergence phenomenon is significantly enhanced when the topological insights, such as increasing degree-heterogeneity and clustering, are embedded into the agent partner selection strategy.

Diaspora is a decentralized online social networking platform where user profiles are hosted in multiple Diaspora nodes (pods) and the social connections can exist across different pods. User profile migration is a promising feature that would enable users to seamlessly migrate their profile data between different pods. However, to the best of our knowledge, there has been no research done on how this data portability may affect the user distribution and the performance of the pods. In this paper, our goal is to design an approach that facilitates the users to choose appropriate pods that would ensure better service quality. We propose a decentralized game-theoretic approach that is based on user's local neighborhood information and the quality of the pods. We have analytically determined, and experimentally substantiated, that through the proposed profile migration approach the users of Diaspora reach a stable and balanced distribution that improves their overall experience in respective pods.

In Vehicular Ad Hoc Networks (VANET) broadcasting is the key means for dissemination of safety messages. Existing broadcasting schemes, however, result in broadcast storm problem. On the other hand, setting up maximum broadcast range is another critical issue for successful communication. Based on the neighbor vehicle density, we intend to mitigate the broadcast storm problem. We also try to reduce the propagation delay for faster dissemination of safety messages across the VANET. Simulation results show that our proposed scheme reduces message travel time for both light and intense traffic scenario compared to some of the existing solutions.

With the increasing demand of voice, video and data, the backhaul wireless mesh networks (WMNs) face the challenge to meet end user requirements. Due to the heavy flow of aggregated backhaul traffic, the wired gateways affect the network parameters adversely if their proper placement is not considered. In this paper, we propose a link aware clustering algorithm for the optimal placement of the gateways and integrating these gateways we design an adaptive overlay for the contention free flow of delay sensitive data. The observations taken from a test bed setup facilitated to determine a feasible network diameter for our system model. Simulation results exhibit that our proposed scheme improves the performance of the backhaul wireless mesh network significantly.