Search

Turn-taking is a fundamental mechanism in human communication that ensures smooth and coherent verbal interactions. Recent advances in Large Language Models (LLMs) have motivated their use in improving the turn-taking capabilities of Spoken Dialogue Systems (SDS), such as their ability to respond at appropriate times. However, existing models often struggle to predict opportunities for speaking -- called Transition Relevance Places (TRPs) -- in natural, unscripted conversations, focusing only on turn-final TRPs and not within-turn TRPs. To address these limitations, we introduce a novel dataset of participant-labeled within-turn TRPs and use it to evaluate the performance of state-of-the-art LLMs in predicting opportunities for speaking. Our experiments reveal the current limitations of LLMs in modeling unscripted spoken interactions, highlighting areas for improvement and paving the way for more naturalistic dialogue systems., Comment: EMNLP 2024 (Findings)

Large Language Models (LLMs), despite achieving state-of-the-art results in a number of evaluation tasks, struggle to maintain their performance when logical reasoning is strictly required to correctly infer a prediction. In this work, we propose Argument Generation as a method of forcing models to utilize their reasoning capabilities when other approaches such as chain-of-thought reasoning prove insufficient. Our method involves the generation of arguments for each possible inference result, and asking the end model to rank the generated arguments. We show that Argument Generation can serve as an appropriate substitute for zero-shot prompting techniques without the requirement to add layers of complexity. Furthermore, we argue that knowledge-probing techniques such as chain-of-thought reasoning and Argument Generation are only useful when further reasoning is required to infer a prediction, making them auxiliary to more common zero-shot approaches. Finally, we demonstrate that our approach forces larger gains in smaller language models, showcasing a complex relationship between model size and prompting methods in foundation models., Comment: Accepted to Workshop on Customizable NLP: Progress and Challenges in Customizing NLP for a Domain, Application, Group, or Individual at EMNLP 2024

Liquid-droplet coalescence and the mergers of liquid lenses are problems of great practical and theoretical interest in fluid dynamics and the statistical mechanics of multi-phase flows. During such mergers, there is an interesting and intricate interplay between the shapes of the interfaces, separating two phases, and the background flow field. In experiments, it is easier to visualize concentration fields than to obtain the flow field. We demonstrate that two-dimensional (2D) encoder-decoder CNNs, 2D U-Nets, and three-dimensional (3D) U-Nets can be used to obtain flow fields from concentration fields here. To train these networks, we use concentration and flow fields, which we obtain from extensive direct numerical simulations (DNSs) of (a) the coalescence of two circular droplets in the two-component 2D Cahn-Hilliard-Navier-Stokes (CHNS) partial differential equations (PDEs), (b) liquid-lens mergers in the three-component 2D CHNS PDEs, and (c) spherical-droplet coalescence in the two-component 3D CHNS PDEs. We then show that, given test images of concentration fields, our trained models accurately predict the flow fields at both high and low Ohnesorge numbers $Oh$ (a dimensionless ratio of viscous stresses to the inertial and surface-tension forces). Using autoencoders and fully connected neural networks, we also investigate the mapping between the concentration and vorticity fields via low-dimensional latent variables for droplet mergers in the 2D CHNS system. We compare the accuracies of flow-field reconstruction based on the two approaches we employ. Finally, we use data from recent experiments on droplet coalescence to show how our method can be used to obtain the flow field from measurements of the concentration field.

Many real-world problems, such as controlling swarms of drones and urban traffic, naturally lend themselves to modeling as multi-agent reinforcement learning (RL) problems. However, existing multi-agent RL methods often suffer from scalability challenges, primarily due to the introduction of communication among agents. Consequently, a key challenge lies in adapting the success of deep learning in single-agent RL to the multi-agent setting. In response to this challenge, we propose an approach that fundamentally reimagines multi-agent environments. Unlike conventional methods that model each agent individually with separate networks, our approach, the Bottom Up Network (BUN), adopts a unique perspective. BUN treats the collective of multi-agents as a unified entity while employing a specialized weight initialization strategy that promotes independent learning. Furthermore, we dynamically establish connections among agents using gradient information, enabling coordination when necessary while maintaining these connections as limited and sparse to effectively manage the computational budget. Our extensive empirical evaluations across a variety of cooperative multi-agent scenarios, including tasks such as cooperative navigation and traffic control, consistently demonstrate BUN's superiority over baseline methods with substantially reduced computational costs., Comment: 13 pages, 24 figures

Deep reinforcement learning (DRL) has found application in numerous use-cases pertaining to flow control. Multi-agent RL (MARL), a variant of DRL, has shown to be more effective than single-agent RL in controlling flows exhibiting locality and translational invariance. We present, for the first time, an implementation of MARL-based control of three-dimensional Rayleigh-B\'enard convection (RBC). Control is executed by modifying the temperature distribution along the bottom wall divided into multiple control segments, each of which acts as an independent agent. Two regimes of RBC are considered at Rayleigh numbers $\mathrm{Ra}=500$ and $750$. Evaluation of the learned control policy reveals a reduction in convection intensity by $23.5\%$ and $8.7\%$ at $\mathrm{Ra}=500$ and $750$, respectively. The MARL controller converts irregularly shaped convective patterns to regular straight rolls with lower convection that resemble flow in a relatively more stable regime. We draw comparisons with proportional control at both $\mathrm{Ra}$ and show that MARL is able to outperform the proportional controller. The learned control strategy is complex, featuring different non-linear segment-wise actuator delays and actuation magnitudes. We also perform successful evaluations on a larger domain than used for training, demonstrating that the invariant property of MARL allows direct transfer of the learnt policy., Comment: Submitted to the special issue titled 'Machine Learning for Fluid Dynamics' in the journal Flow, Turbulence and Combusion. 39 pages and 20 figures

Flow control is key to maximize energy efficiency in a wide range of applications. However, traditional flow-control methods face significant challenges in addressing non-linear systems and high-dimensional data, limiting their application in realistic energy systems. This study advances deep-reinforcement-learning (DRL) methods for flow control, particularly focusing on integrating group-invariant networks and positional encoding into DRL architectures. Our methods leverage multi-agent reinforcement learning (MARL) to exploit policy invariance in space, in combination with group-invariant networks to ensure local symmetry invariance. Additionally, a positional encoding inspired by the transformer architecture is incorporated to provide location information to the agents, mitigating action constraints from strict invariance. The proposed methods are verified using a case study of Rayleigh-B\'enard convection, where the goal is to minimize the Nusselt number Nu. The group-invariant neural networks (GI-NNs) show faster convergence compared to the base MARL, achieving better average policy performance. The GI-NNs not only cut DRL training time in half but also notably enhance learning reproducibility. Positional encoding further enhances these results, effectively reducing the minimum Nu and stabilizing convergence. Interestingly, group invariant networks specialize in improving learning speed and positional encoding specializes in improving learning quality. These results demonstrate that choosing a suitable feature-representation method according to the purpose as well as the characteristics of each control problem is essential. We believe that the results of this study will not only inspire novel DRL methods with invariant and unique representations, but also provide useful insights for industrial applications.

We release the Nemotron-4 340B model family, including Nemotron-4-340B-Base, Nemotron-4-340B-Instruct, and Nemotron-4-340B-Reward. Our models are open access under the NVIDIA Open Model License Agreement, a permissive model license that allows distribution, modification, and use of the models and its outputs. These models perform competitively to open access models on a wide range of evaluation benchmarks, and were sized to fit on a single DGX H100 with 8 GPUs when deployed in FP8 precision. We believe that the community can benefit from these models in various research studies and commercial applications, especially for generating synthetic data to train smaller language models. Notably, over 98% of data used in our model alignment process is synthetically generated, showcasing the effectiveness of these models in generating synthetic data. To further support open research and facilitate model development, we are also open-sourcing the synthetic data generation pipeline used in our model alignment process.

Physically-based renderings contain Monte-Carlo noise, with variance that increases as the number of rays per pixel decreases. This noise, while zero-mean for good modern renderers, can have heavy tails (most notably, for scenes containing specular or refractive objects). Learned methods for restoring low fidelity renders are highly developed, because suppressing render noise means one can save compute and use fast renders with few rays per pixel. We demonstrate that a diffusion model can denoise low fidelity renders successfully. Furthermore, our method can be conditioned on a variety of natural render information, and this conditioning helps performance. Quantitative experiments show that our method is competitive with SOTA across a range of sampling rates. Qualitative examination of the reconstructions suggests that the image prior applied by a diffusion method strongly favors reconstructions that are like real images -- so have straight shadow boundaries, curved specularities and no fireflies., Comment: 25 pages, 18 figures, 2 tables

This paper presents a detailed study of the type II solar radio burst that occurred on 06 March 2014 using combined data analysis. It is a classical radio event consisting of type III radio burst and a following type II radio burst in the dynamic spectrum. The type II radio burst is observed between 235 - 130 MHz (120 - 60 MHz) in harmonic (fundamental) bands with the life time of 5 minutes between 09:26 UT - 09:31 UT. The estimated speed of type II burst by applying twofold Saito model is 650 km/s. An extreme ultraviolet (EUV) wave is observed with Atmospheric Imaging Assembly (AIA) onboard the Solar Dynamics Observatory (SDO) pass bands. The very close temporal onset association of EUV wave and flare energy release indicates that the EUV wave is likely produced by a flare pressure pulse. The eruption is also accompanied by a weak coronal mass ejection (CME) observed with the coronagraphs onboard Solar and Heliospheric Observatory (SOHO) and the twin Solar Terrestrial Relations Observatory (STEREO). The plane of sky speed of the CME was 252 km/s at SOHO/LASCO-C2 and 280 km/s at STEREO-B/SECCHI-COR1 FOV. The EUV wave has two wave fronts, one expanding radially outward, and the other one moving along the arcade. The source position of the type II burst imaged by the Nancay Radio Heliograph (NRH) shows that it was associated with the outward moving EUV wave. The CME is independent of the shock wave as confirmed by the location of NRH radio sources below the CME's leading edge. Therefore the type II radio burst is probably ignited by the flare. This study shows the possibility of EUV wave and coronal shock triggered by flare pressure pulse, generating the observed type II radio burst., Comment: Accepted for publication in Solar Physics

Frequent disruptions like highway constructions are common now-a-days, often impacting environmental justice communities (communities with low socio-economic status with disproportionately high and adverse human health and environmental effects) that live nearby. Based on our interactions via focus groups with the members of four environmental justice communities impacted by a major highway construction, a common concern is a sense of uncertainty about project activities and loss of social connectedness, leading to increased stress, depression, anxiety and diminished well-being. This paper addresses this concern by developing a localized social media system called PureConnect with a goal to raise the level of awareness about the project and increase social connectedness among the community members. PureConnect has been designed using active engagement with four environmental justice communities affected by a major highway construction. It has been deployed in the real world among the members of the four environmental justice communities, and a detailed analysis of the data collected from this deployment as well as surveys show that PureConnect is potentially useful in improving community members' well-being and the members appreciate the functionalities it provides., Comment: Submitted in COMPSAC 2024

Recent advancements in reasoning abilities of Large Language Models (LLM) has promoted their usage in problems that require high-level planning for robots and artificial agents. However, current techniques that utilize LLMs for such planning tasks make certain key assumptions such as, access to datasets that permit finetuning, meticulously engineered prompts that only provide relevant and essential information to the LLM, and most importantly, a deterministic approach to allow execution of the LLM responses either in the form of existing policies or plan operators. In this work, we propose LgTS (LLM-guided Teacher-Student learning), a novel approach that explores the planning abilities of LLMs to provide a graphical representation of the sub-goals to a reinforcement learning (RL) agent that does not have access to the transition dynamics of the environment. The RL agent uses Teacher-Student learning algorithm to learn a set of successful policies for reaching the goal state from the start state while simultaneously minimizing the number of environmental interactions. Unlike previous methods that utilize LLMs, our approach does not assume access to a propreitary or a fine-tuned LLM, nor does it require pre-trained policies that achieve the sub-goals proposed by the LLM. Through experiments on a gridworld based DoorKey domain and a search-and-rescue inspired domain, we show that generating a graphical structure of sub-goals helps in learning policies for the LLM proposed sub-goals and the Teacher-Student learning algorithm minimizes the number of environment interactions when the transition dynamics are unknown.

We present Blocks2World, a novel method for 3D scene rendering and editing that leverages a two-step process: convex decomposition of images and conditioned synthesis. Our technique begins by extracting 3D parallelepipeds from various objects in a given scene using convex decomposition, thus obtaining a primitive representation of the scene. These primitives are then utilized to generate paired data through simple ray-traced depth maps. The next stage involves training a conditioned model that learns to generate images from the 2D-rendered convex primitives. This step establishes a direct mapping between the 3D model and its 2D representation, effectively learning the transition from a 3D model to an image. Once the model is fully trained, it offers remarkable control over the synthesis of novel and edited scenes. This is achieved by manipulating the primitives at test time, including translating or adding them, thereby enabling a highly customizable scene rendering process. Our method provides a fresh perspective on 3D scene rendering and editing, offering control and flexibility. It opens up new avenues for research and applications in the field, including authoring and data augmentation., Comment: 16 pages, 15 figures

We examine the problem of two-point boundary optimal control of nonlinear systems over finite-horizon time periods with unknown model dynamics by employing reinforcement learning. We use techniques from singular perturbation theory to decompose the control problem over the finite horizon into two sub-problems, each solved over an infinite horizon. In the process, we avoid the need to solve the time-varying Hamilton-Jacobi-Bellman equation. Using a policy iteration method, which is made feasible as a result of this decomposition, it is now possible to learn the controller gains of both sub-problems. The overall control is then formed by piecing together the solutions to the two sub-problems. We show that the performance of the proposed closed-loop system approaches that of the model-based optimal performance as the time horizon gets long. Finally, we provide three simulation scenarios to support the paper's claims.

Rayleigh-B\'enard convection (RBC) is a recurrent phenomenon in several industrial and geoscience flows and a well-studied system from a fundamental fluid-mechanics viewpoint. However, controlling RBC, for example by modulating the spatial distribution of the bottom-plate heating in the canonical RBC configuration, remains a challenging topic for classical control-theory methods. In the present work, we apply deep reinforcement learning (DRL) for controlling RBC. We show that effective RBC control can be obtained by leveraging invariant multi-agent reinforcement learning (MARL), which takes advantage of the locality and translational invariance inherent to RBC flows inside wide channels. The MARL framework applied to RBC allows for an increase in the number of control segments without encountering the curse of dimensionality that would result from a naive increase in the DRL action-size dimension. This is made possible by the MARL ability for re-using the knowledge generated in different parts of the RBC domain. We show in a case study that MARL DRL is able to discover an advanced control strategy that destabilizes the spontaneous RBC double-cell pattern, changes the topology of RBC by coalescing adjacent convection cells, and actively controls the resulting coalesced cell to bring it to a new stable configuration. This modified flow configuration results in reduced convective heat transfer, which is beneficial in several industrial processes. Therefore, our work both shows the potential of MARL DRL for controlling large RBC systems, as well as demonstrates the possibility for DRL to discover strategies that move the RBC configuration between different topological configurations, yielding desirable heat-transfer characteristics. These results are useful for both gaining further understanding of the intrinsic properties of RBC, as well as for developing industrial applications., Comment: 34 pages, 11 figures submitted to Physics of Fluids