Search

The remarkable success of foundation models has sparked growing interest in their application to single-cell biology. Models like Geneformer and scGPT promise to serve as versatile tools in this specialized field. However, representing a cell as a sequence of genes remains an open question since the order of genes is interchangeable. Injecting the gene network graph offers gene relative positions and compact data representation but poses a dilemma: limited receptive fields without in-layer message passing or parameter explosion with message passing in each layer. To pave the way forward, we propose GenoHoption, a new computational framework for single-cell sequencing data that effortlessly combines the strengths of these foundation models with explicit relationships in gene networks. We also introduce a constraint that lightens the model by focusing on learning the predefined graph structure while ensuring further hops are deducted to expand the receptive field. Empirical studies show that our model improves by an average of 1.27% on cell-type annotation and 3.86% on perturbation prediction. Furthermore, our method significantly decreases computational overhead and exhibits few-shot potential. GenoHoption can function as an efficient and expressive bridge, connecting existing single-cell foundation models to gene network graphs.

We present an energy-specific Bethe-Salpeter equation (BSE) implementation for efficient core and valence optical spectrum calculations. In energy-specific BSE, high-lying excitation energies are obtained by constructing trial vectors and expanding the subspace targeting excitation energies above the predefined energy threshold in the Davidson algorithm. To calculate optical spectra over a wide energy range, energy-specific BSE can be applied to multiple consecutive small energy windows, where trial vectors for each subsequent energy window are made orthogonal to the subspace of preceding windows to accelerate the convergence of the Davidson algorithm. For seven small molecules, energy-specific BSE combined with $G_0W_0$ provides small errors around 0.8 eV for absolute and relative $K$-edge excitation energies when starting from a hybrid PBEh solution with 45\% exact exchange. We further showcase the computational efficiency of this approach by simulating the N $1s$ $K$-edge excitation spectrum of the porphine molecule and the valence optical spectrum of silicon nanoclusters involving 5,000 excited states using $G_0W_0$-BSE. This work expands the applicability of the $GW$-BSE formalism for investigating high-energy excited states of large systems., Comment: 16 pages, 2 figures

Modeling cellular dynamics from single-cell RNA sequencing (scRNA-seq) data is critical for understanding cell development and underlying gene regulatory relationships. Many current methods rely on single-cell velocity to obtain pseudotime, which can lead to inconsistencies between pseudotime and velocity. It is challenging to simultaneously infer cell pseudotime and gene interaction networks, especially in multi-branch differentiation scenarios. We present single-cell Piecewise Network (scPN), a novel high-dimensional dynamical modeling approach that iteratively extracts temporal patterns and inter-gene relationships from scRNA-seq data. To tackle multi-branch differentiation challenges, scPN models gene regulatory dynamics using piecewise gene-gene interaction networks, offering an interpretable framework for deciphering complex gene regulation patterns over time. Results on synthetic data and multiple scRNA-seq datasets demonstrate the superior performance of scPN in reconstructing cellular dynamics and identifying key transcription factors involved in development compared to existing methods. To the best of our knowledge, scPN is the first attempt at modeling that can recover pseudotime, velocity fields, and gene interactions all at once on multi-branch datasets.

Researchers have long recognized the socio-technical gaps in personal tracking research, where machines can never fully model the complexity of human behavior, making it only able to produce basic rule-based outputs or "black-box" results that lack clear explanations. Real-world deployments rely on experts for this complex translation from sparse data to meaningful insights. In this study, we consider this translation process from data to insights by experts as "sensemaking" and explore how HCI researchers can support it through Vital Insight, an evidence-based 'sensemaking' system that combines direct representation and indirect inference through visualization and Large Language Models. We evaluate Vital Insight in user testing sessions with 14 experts in multi-modal tracking, synthesize design implications, and develop an expert sensemaking model where they iteratively move between direct data representations and AI-supported inferences to explore, retrieve, question, and validate insights.

In this paper, we focus on enhancing a diffusion-based text-to-video (T2V) model during the post-training phase by distilling a highly capable consistency model from a pretrained T2V model. Our proposed method, T2V-Turbo-v2, introduces a significant advancement by integrating various supervision signals, including high-quality training data, reward model feedback, and conditional guidance, into the consistency distillation process. Through comprehensive ablation studies, we highlight the crucial importance of tailoring datasets to specific learning objectives and the effectiveness of learning from diverse reward models for enhancing both the visual quality and text-video alignment. Additionally, we highlight the vast design space of conditional guidance strategies, which centers on designing an effective energy function to augment the teacher ODE solver. We demonstrate the potential of this approach by extracting motion guidance from the training datasets and incorporating it into the ODE solver, showcasing its effectiveness in improving the motion quality of the generated videos with the improved motion-related metrics from VBench and T2V-CompBench. Empirically, our T2V-Turbo-v2 establishes a new state-of-the-art result on VBench, with a Total score of 85.13, surpassing proprietary systems such as Gen-3 and Kling., Comment: Project Page: https://t2v-turbo-v2.github.io/

Offline imitation learning enables learning a policy solely from a set of expert demonstrations, without any environment interaction. To alleviate the issue of distribution shift arising due to the small amount of expert data, recent works incorporate large numbers of auxiliary demonstrations alongside the expert data. However, the performance of these approaches rely on assumptions about the quality and composition of the auxiliary data. However, they are rarely successful when those assumptions do not hold. To address this limitation, we propose Robust Offline Imitation from Diverse Auxiliary Data (ROIDA). ROIDA first identifies high-quality transitions from the entire auxiliary dataset using a learned reward function. These high-reward samples are combined with the expert demonstrations for weighted behavioral cloning. For lower-quality samples, ROIDA applies temporal difference learning to steer the policy towards high-reward states, improving long-term returns. This two-pronged approach enables our framework to effectively leverage both high and low-quality data without any assumptions. Extensive experiments validate that ROIDA achieves robust and consistent performance across multiple auxiliary datasets with diverse ratios of expert and non-expert demonstrations. ROIDA effectively leverages unlabeled auxiliary data, outperforming prior methods reliant on specific data assumptions.

Language models (LMs) possess a strong capability to comprehend natural language, making them effective in translating human instructions into detailed plans for simple robot tasks. Nevertheless, it remains a significant challenge to handle long-horizon tasks, especially in subtask identification and allocation for cooperative heterogeneous robot teams. To address this issue, we propose a Language Model-Driven Multi-Agent PDDL Planner (LaMMA-P), a novel multi-agent task planning framework that achieves state-of-the-art performance on long-horizon tasks. LaMMA-P integrates the strengths of the LMs' reasoning capability and the traditional heuristic search planner to achieve a high success rate and efficiency while demonstrating strong generalization across tasks. Additionally, we create MAT-THOR, a comprehensive benchmark that features household tasks with two different levels of complexity based on the AI2-THOR environment. The experimental results demonstrate that LaMMA-P achieves a 105% higher success rate and 36% higher efficiency than existing LM-based multi-agent planners. The experimental videos, code, and datasets of this work as well as the detailed prompts used in each module are available at https://lamma-p.github.io., Comment: Project website: https://lamma-p.github.io/

Recent studies show that text-to-image (T2I) models are vulnerable to adversarial attacks, especially with noun perturbations in text prompts. In this study, we investigate the impact of adversarial attacks on different POS tags within text prompts on the images generated by T2I models. We create a high-quality dataset for realistic POS tag token swapping and perform gradient-based attacks to find adversarial suffixes that mislead T2I models into generating images with altered tokens. Our empirical results show that the attack success rate (ASR) varies significantly among different POS tag categories, with nouns, proper nouns, and adjectives being the easiest to attack. We explore the mechanism behind the steering effect of adversarial suffixes, finding that the number of critical tokens and content fusion vary among POS tags, while features like suffix transferability are consistent across categories. We have made our implementation publicly available at - https://github.com/shahariar-shibli/Adversarial-Attack-on-POS-Tags., Comment: Findings of the EMNLP 2024

Cooperative perception systems play a vital role in enhancing the safety and efficiency of vehicular autonomy. Although recent studies have highlighted the efficacy of vehicle-to-everything (V2X) communication techniques in autonomous driving, a significant challenge persists: how to efficiently integrate multiple high-bandwidth features across an expanding network of connected agents such as vehicles and infrastructure. In this paper, we introduce CoMamba, a novel cooperative 3D detection framework designed to leverage state-space models for real-time onboard vehicle perception. Compared to prior state-of-the-art transformer-based models, CoMamba enjoys being a more scalable 3D model using bidirectional state space models, bypassing the quadratic complexity pain-point of attention mechanisms. Through extensive experimentation on V2X/V2V datasets, CoMamba achieves superior performance compared to existing methods while maintaining real-time processing capabilities. The proposed framework not only enhances object detection accuracy but also significantly reduces processing time, making it a promising solution for next-generation cooperative perception systems in intelligent transportation networks., Comment: Project Page: this https URL https://taco-group.github.io/CoMamba/

Recent progress in large language model (LLM) technology has significantly enhanced the interaction experience between humans and voice assistants (VAs). This project aims to explore a user's continuous interaction with LLM-based VA (LLM-VA) during a complex task. We recruited 12 participants to interact with an LLM-VA during a cooking task, selected for its complexity and the requirement for continuous interaction. We observed that users show both verbal and nonverbal behaviors, though they know that the LLM-VA can not capture those nonverbal signals. Despite the prevalence of nonverbal behavior in human-human communication, there is no established analytical methodology or framework for exploring it in human-VA interactions. After analyzing 3 hours and 39 minutes of video recordings, we developed an analytical framework with three dimensions: 1) behavior characteristics, including both verbal and nonverbal behaviors, 2) interaction stages--exploration, conflict, and integration--that illustrate the progression of user interactions, and 3) stage transition throughout the task. This analytical framework identifies key verbal and nonverbal behaviors that provide a foundation for future research and practical applications in optimizing human and LLM-VA interactions.

Mental health has become a growing concern among university students. While medication is a common treatment, understanding how university students manage their medication for mental health symptoms in real-world practice has not been fully explored. In this study, we conducted semi-structured interviews with university students to understand the unique challenges in the mental health medication management process and their coping strategies, particularly examining the role of various technologies in this process. We discovered that due to struggles with self-acceptance and the interdependent relationship between medication, symptoms, schedules, and life changes, the medication management process for students was a highly dynamic journey involving frequent dosage changes. Thus, students adopted flexible strategies of using minimal technology to manage their medication in different situations while maintaining a high degree of autonomy. Based on our findings, we propose design implications for future technologies to seamlessly integrate into their daily lives and assist students in managing their mental health medications.

Environment prediction frameworks are critical for the safe navigation of autonomous vehicles (AVs) in dynamic settings. LiDAR-generated occupancy grid maps (L-OGMs) offer a robust bird's-eye view for the scene representation, enabling self-supervised joint scene predictions while exhibiting resilience to partial observability and perception detection failures. Prior approaches have focused on deterministic L-OGM prediction architectures within the grid cell space. While these methods have seen some success, they frequently produce unrealistic predictions and fail to capture the stochastic nature of the environment. Additionally, they do not effectively integrate additional sensor modalities present in AVs. Our proposed framework performs stochastic L-OGM prediction in the latent space of a generative architecture and allows for conditioning on RGB cameras, maps, and planned trajectories. We decode predictions using either a single-step decoder, which provides high-quality predictions in real-time, or a diffusion-based batch decoder, which can further refine the decoded frames to address temporal consistency issues and reduce compression losses. Our experiments on the nuScenes and Waymo Open datasets show that all variants of our approach qualitatively and quantitatively outperform prior approaches.

Quantum many-body methods provide a systematic route to computing electronic properties of molecules and materials, but high computational costs restrict their use in large-scale applications. Due to the complexity in many-electron wavefunctions, machine learning models capable of capturing fundamental many-body physics remain limited. Here, we present a deep learning framework targeting the many-body Green's function, which unifies predictions of electronic properties in ground and excited states, while offering deep physical insights into electron correlation effects. By learning the $GW$ or coupled-cluster self-energy from mean-field features, our graph neural network achieves competitive performance in predicting one- and two-particle excitations and quantities derivable from one-particle density matrix. We demonstrate its high data efficiency and good transferability across chemical species, system sizes, molecular conformations, and correlation strengths in bond breaking, through multiple molecular and nanomaterial benchmarks. This work opens up new opportunities for utilizing machine learning to solve many-electron problems., Comment: 16 pages, 5 figures

Reinforcement Learning (RL) has enabled social robots to generate trajectories without human-designed rules or interventions, which makes it more effective than hard-coded systems for generalizing to complex real-world scenarios. However, social navigation is a safety-critical task that requires robots to avoid collisions with pedestrians while previous RL-based solutions fall short in safety performance in complex environments. To enhance the safety of RL policies, to the best of our knowledge, we propose the first algorithm, SoNIC, that integrates adaptive conformal inference (ACI) with constrained reinforcement learning (CRL) to learn safe policies for social navigation. More specifically, our method augments RL observations with ACI-generated nonconformity scores and provides explicit guidance for agents to leverage the uncertainty metrics to avoid safety-critical areas by incorporating safety constraints with spatial relaxation. Our method outperforms state-of-the-art baselines in terms of both safety and adherence to social norms by a large margin and demonstrates much stronger robustness to out-of-distribution scenarios. Our code and video demos are available on our project website: https://sonic-social-nav.github.io/., Comment: Project website: https://sonic-social-nav.github.io/

Hybrid Quantum Neural Networks (HQNNs) represent a promising advancement in Quantum Machine Learning (QML), yet their security has been rarely explored. In this paper, we present the first systematic study of backdoor attacks on HQNNs. We begin by proposing an attack framework and providing a theoretical analysis of the generalization bounds and minimum perturbation requirements for backdoor attacks on HQNNs. Next, we employ two classic backdoor attack methods on HQNNs and Convolutional Neural Networks (CNNs) to further investigate the robustness of HQNNs. Our experimental results demonstrate that HQNNs are more robust than CNNs, requiring more significant image modifications for successful attacks. Additionally, we introduce the Qcolor backdoor, which utilizes color shifts as triggers and employs the Non-dominated Sorting Genetic Algorithm II (NSGA-II) to optimize hyperparameters. Through extensive experiments, we demonstrate the effectiveness, stealthiness, and robustness of the Qcolor backdoor.

Training intelligent agents to navigate highly interactive environments presents significant challenges. While guided meta reinforcement learning (RL) approach that first trains a guiding policy to train the ego agent has proven effective in improving generalizability across scenarios with various levels of interaction, the state-of-the-art method tends to be overly sensitive to extreme cases, impairing the agents' performance in the more common scenarios. This study introduces a novel training framework that integrates guided meta RL with importance sampling (IS) to optimize training distributions iteratively for navigating highly interactive driving scenarios, such as T-intersections or roundabouts. Unlike traditional methods that may underrepresent critical interactions or overemphasize extreme cases during training, our approach strategically adjusts the training distribution towards more challenging driving behaviors using IS proposal distributions and applies the importance ratio to de-bias the result. By estimating a naturalistic distribution from real-world datasets and employing a mixture model for iterative training refinements, the framework ensures a balanced focus across common and extreme driving scenarios. Experiments conducted with both synthetic and naturalistic datasets demonstrate both accelerated training and performance improvements under highly interactive driving tasks.

Deep learning-based trajectory prediction models for autonomous driving often struggle with generalization to out-of-distribution (OOD) scenarios, sometimes performing worse than simple rule-based models. To address this limitation, we propose a novel framework, Adaptive Prediction Ensemble (APE), which integrates deep learning and rule-based prediction experts. A learned routing function, trained concurrently with the deep learning model, dynamically selects the most reliable prediction based on the input scenario. Our experiments on large-scale datasets, including Waymo Open Motion Dataset (WOMD) and Argoverse, demonstrate improvement in zero-shot generalization across datasets. We show that our method outperforms individual prediction models and other variants, particularly in long-horizon prediction and scenarios with a high proportion of OOD data. This work highlights the potential of hybrid approaches for robust and generalizable motion prediction in autonomous driving.

The particle-particle random phase approximation (ppRPA) within the hole-hole channel was recently proposed as an efficient tool for computing excitation energies of point defects in solids [J. Phys. Chem. Lett. 2024, 15, 2757-2764]. In this work, we investigate the application of ppRPA within the particle-particle channel for predicting correlated excited states of point defects, including the carbon-vacancy (VC) in diamond, the oxygen-vacancy (VO) in magnesium oxide (MgO), and the carbon dimer defect (C$_{\text{B}}$C$_{\text{N}}$) in two-dimensional hexagonal boron nitride (h-BN). Starting from a density functional theory calculation of the ($N-2$)-electron ground state, vertical excitation energies of the $N$-electron system are obtained as the differences between the two-electron addition energies. We show that active-space ppRPA with the B3LYP functional yields accurate excitation energies, with errors mostly smaller than 0.1 eV for tested systems compared to available experimental values. We further develop a natural transition orbital scheme within ppRPA, which provides insights into the multireference character of defect states. This study, together with our previous work, establishes ppRPA as a low-cost and accurate method for investigating excited-state properties of point defect systems.

Counterfactual explanations of Graph Neural Networks (GNNs) offer a powerful way to understand data that can naturally be represented by a graph structure. Furthermore, in many domains, it is highly desirable to derive data-driven global explanations or rules that can better explain the high-level properties of the models and data in question. However, evaluating global counterfactual explanations is hard in real-world datasets due to a lack of human-annotated ground truth, which limits their use in areas like molecular sciences. Additionally, the increasing scale of these datasets provides a challenge for random search-based methods. In this paper, we develop a novel global explanation model RLHEX for molecular property prediction. It aligns the counterfactual explanations with human-defined principles, making the explanations more interpretable and easy for experts to evaluate. RLHEX includes a VAE-based graph generator to generate global explanations and an adapter to adjust the latent representation space to human-defined principles. Optimized by Proximal Policy Optimization (PPO), the global explanations produced by RLHEX cover 4.12% more input graphs and reduce the distance between the counterfactual explanation set and the input set by 0.47% on average across three molecular datasets. RLHEX provides a flexible framework to incorporate different human-designed principles into the counterfactual explanation generation process, aligning these explanations with domain expertise. The code and data are released at https://github.com/dqwang122/RLHEX., Comment: Accepted by KDD 2024

Direct alignment from preferences (DAP) has emerged as a promising paradigm for aligning large language models (LLMs) to human desiderata from pre-collected, offline preference datasets. While recent studies indicate that existing offline DAP methods can directly benefit from online training samples, we highlight the need to develop specific online DAP algorithms to fully harness the power of online training. Specifically, we identify that the learned LLM should adhere to the proximity of the behavior LLM, which collects the training samples. To this end, we propose online Preference Optimization in proximity to the Behavior LLM (BPO), emphasizing the importance of constructing a proper trust region for LLM alignment. We conduct extensive experiments to validate the effectiveness and applicability of our approach by integrating it with various DAP methods, resulting in significant performance improvements across a wide range of tasks when training with the same amount of preference data. Even when only introducing one additional data collection phase, our online BPO improves its offline DAP baseline from 72.0% to 80.2% on TL;DR and from 82.2% to 89.1% on Anthropic Helpfulness in terms of win rate against human reference text., Comment: Wenda Xu and Jiachen Li contributed equally. Accepted by EMNLP 2024

Video generation has many unique challenges beyond those of image generation. The temporal dimension introduces extensive possible variations across frames, over which consistency and continuity may be violated. In this study, we move beyond evaluating simple actions and argue that generated videos should incorporate the emergence of new concepts and their relation transitions like in real-world videos as time progresses. To assess the Temporal Compositionality of video generation models, we propose TC-Bench, a benchmark of meticulously crafted text prompts, corresponding ground truth videos, and robust evaluation metrics. The prompts articulate the initial and final states of scenes, effectively reducing ambiguities for frame development and simplifying the assessment of transition completion. In addition, by collecting aligned real-world videos corresponding to the prompts, we expand TC-Bench's applicability from text-conditional models to image-conditional ones that can perform generative frame interpolation. We also develop new metrics to measure the completeness of component transitions in generated videos, which demonstrate significantly higher correlations with human judgments than existing metrics. Our comprehensive experimental results reveal that most video generators achieve less than 20% of the compositional changes, highlighting enormous space for future improvement. Our analysis indicates that current video generation models struggle to interpret descriptions of compositional changes and synthesize various components across different time steps.

Multimodal Language Language Models (MLLMs) demonstrate the emerging abilities of "world models" -- interpreting and reasoning about complex real-world dynamics. To assess these abilities, we posit videos are the ideal medium, as they encapsulate rich representations of real-world dynamics and causalities. To this end, we introduce MMWorld, a new benchmark for multi-discipline, multi-faceted multimodal video understanding. MMWorld distinguishes itself from previous video understanding benchmarks with two unique advantages: (1) multi-discipline, covering various disciplines that often require domain expertise for comprehensive understanding; (2) multi-faceted reasoning, including explanation, counterfactual thinking, future prediction, etc. MMWorld consists of a human-annotated dataset to evaluate MLLMs with questions about the whole videos and a synthetic dataset to analyze MLLMs within a single modality of perception. Together, MMWorld encompasses 1,910 videos across seven broad disciplines and 69 subdisciplines, complete with 6,627 question-answer pairs and associated captions. The evaluation includes 2 proprietary and 10 open-source MLLMs, which struggle on MMWorld (e.g., GPT-4V performs the best with only 52.3\% accuracy), showing large room for improvement. Further ablation studies reveal other interesting findings such as models' different skill sets from humans. We hope MMWorld can serve as an essential step towards world model evaluation in videos.

Quantitative simulation of electronic structure of solids requires treating local and non-local electron correlations on an equal footing. We present a new ab initio formulation of Green's function embedding which, unlike dynamical mean-field theory that uses non-interacting bath, derives bath representation with general two-particle interactions in a systematically improvable manner. The resulting interacting-bath dynamical embedding theory (ibDET) utilizes an efficient real-axis coupled-cluster solver to compute the self-energy, approaching the full system limit at much reduced cost. When combined with the GW theory, GW+ibDET achieves good agreement with experimental spectral properties across a range of semiconducting, insulating and metallic materials. Our approach also enables quantifying the role of non-local electron correlation in determining material properties and addressing the long-standing debate on the bandwidth narrowing of metallic sodium., Comment: 4 pages, 4 figures

Diffusion-based text-to-video (T2V) models have achieved significant success but continue to be hampered by the slow sampling speed of their iterative sampling processes. To address the challenge, consistency models have been proposed to facilitate fast inference, albeit at the cost of sample quality. In this work, we aim to break the quality bottleneck of a video consistency model (VCM) to achieve $\textbf{both fast and high-quality video generation}$. We introduce T2V-Turbo, which integrates feedback from a mixture of differentiable reward models into the consistency distillation (CD) process of a pre-trained T2V model. Notably, we directly optimize rewards associated with single-step generations that arise naturally from computing the CD loss, effectively bypassing the memory constraints imposed by backpropagating gradients through an iterative sampling process. Remarkably, the 4-step generations from our T2V-Turbo achieve the highest total score on VBench, even surpassing Gen-2 and Pika. We further conduct human evaluations to corroborate the results, validating that the 4-step generations from our T2V-Turbo are preferred over the 50-step DDIM samples from their teacher models, representing more than a tenfold acceleration while improving video generation quality., Comment: Project page: https://t2v-turbo.github.io/

Recent studies reveal that Large Language Models (LLMs) face challenges in balancing safety with utility, particularly when processing long texts for NLP tasks like summarization and translation. Despite defenses against malicious short questions, the ability of LLMs to safely handle dangerous long content, such as manuals teaching illicit activities, remains unclear. Our work aims to develop robust defenses for LLMs in processing malicious documents alongside benign NLP task queries. We introduce a defense dataset comprised of safety-related examples and propose single-task and mixed-task losses for instruction tuning. Our empirical results demonstrate that LLMs can significantly enhance their capacity to safely manage dangerous content with appropriate instruction tuning. Additionally, strengthening the defenses of tasks most susceptible to misuse is effective in protecting LLMs against processing harmful information. We also observe that trade-offs between utility and safety exist in defense strategies, where Llama2, utilizing our proposed approach, displays a significantly better balance compared to Llama1., Comment: accepted to NAACL2024 TrustNLP workshop

Recent advancements in Multimodal Large Language Models (LLMs) have focused primarily on scaling by increasing text-image pair data and enhancing LLMs to improve performance on multimodal tasks. However, these scaling approaches are computationally expensive and overlook the significance of improving model capabilities from the vision side. Inspired by the successful applications of Mixture-of-Experts (MoE) in LLMs, which improves model scalability during training while keeping inference costs similar to those of smaller models, we propose CuMo. CuMo incorporates Co-upcycled Top-K sparsely-gated Mixture-of-experts blocks into both the vision encoder and the MLP connector, thereby enhancing the multimodal LLMs with minimal additional activated parameters during inference. CuMo first pre-trains the MLP blocks and then initializes each expert in the MoE block from the pre-trained MLP block during the visual instruction tuning stage. Auxiliary losses are used to ensure a balanced loading of experts. CuMo outperforms state-of-the-art multimodal LLMs across various VQA and visual-instruction-following benchmarks using models within each model size group, all while training exclusively on open-sourced datasets. The code and model weights for CuMo are open-sourced at https://github.com/SHI-Labs/CuMo.

We numerically study the spectral statistics of open quantum many-body systems (OQMBS) as signatures of quantum chaos (or the lack thereof), using the dissipative spectral form factor (DSFF), a generalization of the spectral form factor to complex spectra. We show that the DSFF of chaotic OQMBS generically displays the $\textit{quadratic}$ ramp-plateau behaviour of the Ginibre ensemble from random matrix theory, in contrast to the linear ramp-plateau behaviour of the Gaussian ensemble in closed quantum systems. Furthermore, in the presence of many-body interactions, such RMT behaviour emerges only after a time scale $\tau_{\mathrm{dev}}$, which generally increases with system size for sufficiently large system size, and can be identified as the non-Hermitian analogue of the $\textit{many-body Thouless time}$. The universality of the random matrix theory behavior is demonstrated by surveying twelve models of OQMBS, including random Kraus circuits (quantum channels) and random Lindbladians (Liouvillians) in several symmetry classes, as well as Lindbladians of paradigmatic models such as the Sachdev-Ye-Kitaev (SYK), XXZ, and the transverse field Ising models. We devise an unfolding and filtering procedure to remove variations of the averaged density of states which would otherwise hide the universal RMT-like signatures in the DSFF for chaotic OQMBS. Beyond chaotic OQMBS, we study the spectral statistics of non-chaotic OQMBS, specifically the integrable XX model and a system in the many-body localized (MBL) regime in the presence of dissipation, which exhibit DSFF behaviours distinct from the ramp-plateau behaviour of random matrix theory. Lastly, we study the DSFF of Lindbladians with the Hamiltonian term set to zero, i.e. only the jump operators are present, and demonstrate that the results of RMT universality and scaling of many-body Thouless time survive even without coherent evolution.

Patient-provider communication has been crucial to cancer patients' survival after their cancer treatments. However, the research community and patients themselves often overlook the communication challenges after cancer treatments as they are overshadowed by the severity of the patient's illness and the variety and rarity of the cancer disease itself. Meanwhile, the recent technical advances in AI, especially in Large Language Models (LLMs) with versatile natural language interpretation and generation ability, demonstrate great potential to support communication in complex real-world medical situations. By interviewing six healthcare providers and eight cancer patients, our goal is to explore the providers' and patients' communication barriers in the post-cancer treatment recovery period, their expectations for future communication technologies, and the potential of AI technologies in this context. Our findings reveal several challenges in current patient-provider communication, including the knowledge and timing gaps between cancer patients and providers, their collaboration obstacles, and resource limitations. Moreover, based on providers' and patients' needs and expectations, we summarize a set of design implications for intelligent communication systems, especially with the power of LLMs. Our work sheds light on the design of future AI-powered systems for patient-provider communication under high-stake and high-uncertainty situations., Comment: 18 pages, 2 figures, submission to CSCW'24

The confluence of the advancement of Autonomous Vehicles (AVs) and the maturity of Vehicle-to-Everything (V2X) communication has enabled the capability of cooperative connected and automated vehicles (CAVs). Building on top of cooperative perception, this paper explores the feasibility and effectiveness of cooperative motion prediction. Our method, CMP, takes LiDAR signals as model input to enhance tracking and prediction capabilities. Unlike previous work that focuses separately on either cooperative perception or motion prediction, our framework, to the best of our knowledge, is the first to address the unified problem where CAVs share information in both perception and prediction modules. Incorporated into our design is the unique capability to tolerate realistic V2X bandwidth limitations and transmission delays, while dealing with bulky perception representations. We also propose a prediction aggregation module, which unifies the predictions obtained by different CAVs and generates the final prediction. Through extensive experiments and ablation studies on the OPV2V and V2V4Real datasets, we demonstrate the effectiveness of our method in cooperative perception, tracking, and motion prediction. In particular, CMP reduces the average prediction error by 16.4\% with fewer missing detections compared with the no cooperation setting and by 12.3\% compared with the strongest baseline. Our work marks a significant step forward in the cooperative capabilities of CAVs, showcasing enhanced performance in complex scenarios. The code can be found on the project website: https://cmp-cooperative-prediction.github.io/., Comment: Project website: https://cmp-cooperative-prediction.github.io/

Latent Consistency Distillation (LCD) has emerged as a promising paradigm for efficient text-to-image synthesis. By distilling a latent consistency model (LCM) from a pre-trained teacher latent diffusion model (LDM), LCD facilitates the generation of high-fidelity images within merely 2 to 4 inference steps. However, the LCM's efficient inference is obtained at the cost of the sample quality. In this paper, we propose compensating the quality loss by aligning LCM's output with human preference during training. Specifically, we introduce Reward Guided LCD (RG-LCD), which integrates feedback from a reward model (RM) into the LCD process by augmenting the original LCD loss with the objective of maximizing the reward associated with LCM's single-step generation. As validated through human evaluation, when trained with the feedback of a good RM, the 2-step generations from our RG-LCM are favored by humans over the 50-step DDIM samples from the teacher LDM, representing a 25-time inference acceleration without quality loss. As directly optimizing towards differentiable RMs can suffer from over-optimization, we take the initial step to overcome this difficulty by proposing the use of a latent proxy RM (LRM). This novel component serves as an intermediary, connecting our LCM with the RM. Empirically, we demonstrate that incorporating the LRM into our RG-LCD successfully avoids high-frequency noise in the generated images, contributing to both improved Fr\'echet Inception Distance (FID) on MS-COCO and a higher HPSv2.1 score on HPSv2's test set, surpassing those achieved by the baseline LCM., Comment: Accepted by TMLR. Project page: https://rg-lcd.github.io/

We develop a functional derivative approach to calculate the chemical potentials of the second-order perturbation theory (MP2). In the functional derivative approach, the correlation part of the MP2 chemical potential, which is the derivative of the MP2 correlation energy with respect to the occupation number of frontier orbitals, is obtained from the chain rule via the non-interacting Green's function. First, the MP2 correlation energy is expressed in terms of the non-interacting Green's function and its functional derivative to the non-interacting Green's function is the second-order self-energy. Then the derivative of the non-interacting Green's function to the occupation number is obtained by including the orbital relaxation effect. We show that the MP2 chemical potentials obtained from the functional derivative approach agrees with that obtained from the finite difference approach. The one-electron Hamiltonian, defined as the derivative of the MP2 energy with respect to the one particle density matrix, is also derived using the functional derivative approach, which can be used in the self-consistent calculations of MP2 and double-hybrid density functionals. The developed functional derivative approach is promising for calculating the chemical potentials and the one-electron Hamiltonian of approximate functionals and many-body perturbation approaches dependent explicitly on the non-interacting Green's function.

Data-driven methods have great advantages in modeling complicated human behavioral dynamics and dealing with many human-robot interaction applications. However, collecting massive and annotated real-world human datasets has been a laborious task, especially for highly interactive scenarios. On the other hand, algorithmic data generation methods are usually limited by their model capacities, making them unable to offer realistic and diverse data needed by various application users. In this work, we study trajectory-level data generation for multi-human or human-robot interaction scenarios and propose a learning-based automatic trajectory generation model, which we call Multi-Agent TRajectory generation with dIverse conteXts (MATRIX). MATRIX is capable of generating interactive human behaviors in realistic diverse contexts. We achieve this goal by modeling the explicit and interpretable objectives so that MATRIX can generate human motions based on diverse destinations and heterogeneous behaviors. We carried out extensive comparison and ablation studies to illustrate the effectiveness of our approach across various metrics. We also presented experiments that demonstrate the capability of MATRIX to serve as data augmentation for imitation-based motion planning., Comment: IEEE International Conference on Robotics and Automation (ICRA 2024)

Offline learning has become widely used due to its ability to derive effective policies from offline datasets gathered by expert demonstrators without interacting with the environment directly. Recent research has explored various ways to enhance offline learning efficiency by considering the characteristics (e.g., expertise level or multiple demonstrators) of the dataset. However, a different approach is necessary in the context of zero-sum games, where outcomes vary significantly based on the strategy of the opponent. In this study, we introduce a novel approach that uses unsupervised learning techniques to estimate the exploited level of each trajectory from the offline dataset of zero-sum games made by diverse demonstrators. Subsequently, we incorporate the estimated exploited level into the offline learning to maximize the influence of the dominant strategy. Our method enables interpretable exploited level estimation in multiple zero-sum games and effectively identifies dominant strategy data. Also, our exploited level augmented offline learning significantly enhances the original offline learning algorithms including imitation learning and offline reinforcement learning for zero-sum games., Comment: 12 pages, 8 figures

Social robot navigation can be helpful in various contexts of daily life but requires safe human-robot interactions and efficient trajectory planning. While modeling pairwise relations has been widely studied in multi-agent interacting systems, the ability to capture larger-scale group-wise activities is limited. In this paper, we propose a systematic relational reasoning approach with explicit inference of the underlying dynamically evolving relational structures, and we demonstrate its effectiveness for multi-agent trajectory prediction and social robot navigation. In addition to the edges between pairs of nodes (i.e., agents), we propose to infer hyperedges that adaptively connect multiple nodes to enable group-wise reasoning in an unsupervised manner. Our approach infers dynamically evolving relation graphs and hypergraphs to capture the evolution of relations, which the trajectory predictor employs to generate future states. Meanwhile, we propose to regularize the sharpness and sparsity of the learned relations and the smoothness of the relation evolution, which proves to enhance training stability and model performance. The proposed approach is validated on synthetic crowd simulations and real-world benchmark datasets. Experiments demonstrate that the approach infers reasonable relations and achieves state-of-the-art prediction performance. In addition, we present a deep reinforcement learning (DRL) framework for social robot navigation, which incorporates relational reasoning and trajectory prediction systematically. In a group-based crowd simulation, our method outperforms the strongest baseline by a significant margin in terms of safety, efficiency, and social compliance in dense, interactive scenarios. We also demonstrate the practical applicability of our method with real-world robot experiments. The code and videos can be found at https://relational-reasoning-nav.github.io/., Comment: Project website: https://relational-reasoning-nav.github.io/; 20 pages, 9 figures, 6 tables

We present an efficient particle-particle random phase approximation (ppRPA) approach that predicts accurate excitation energies of point defects, including the nitrogen-vacancy (NV$^-$) and the silicon-vacancy (SiV$^0$) centers in diamond and the divacancy center (VV$^0$) in 4H silicon carbide, with errors within 0.2 eV compared with experimental values. Starting from the ($N+2$)-electron ground state calculated with the density functional theory (DFT), the ppRPA excitation energies of the $N$-electron system are calculated as the differences between the two-electron removal energies of the ($N+2$)-electron system. We demonstrate that the ppRPA excitation energies converge rapidly with a few hundred of canonical active-space orbitals. We also show that active-space ppRPA has weak DFT starting-point dependence and is significantly cheaper than the corresponding ground-state DFT calculation. This work establishes ppRPA as an accurate and low-cost tool for investigating excited-state properties of point defects and opens up new opportunities for applications of ppRPA to periodic bulk materials.

Graph-structured data is ubiquitous throughout natural and social sciences, and Graph Neural Networks (GNNs) have recently been shown to be effective at solving prediction and inference problems on graph data. In this paper, we propose and demonstrate that GNNs can be applied to solve Combinatorial Optimization (CO) problems. CO concerns optimizing a function over a discrete solution space that is often intractably large. To learn to solve CO problems, we formulate the optimization process as a sequential decision making problem, where the return is related to how close the candidate solution is to optimality. We use a GNN to learn a policy to iteratively build increasingly promising candidate solutions. We present preliminary evidence that GNNs trained through Q-Learning can solve CO problems with performance approaching state-of-the-art heuristic-based solvers, using only a fraction of the parameters and training time.

Effective interaction modeling and behavior prediction of dynamic agents play a significant role in interactive motion planning for autonomous robots. Although existing methods have improved prediction accuracy, few research efforts have been devoted to enhancing prediction model interpretability and out-of-distribution (OOD) generalizability. This work addresses these two challenging aspects by designing a variational auto-encoder framework that integrates graph-based representations and time-sequence models to efficiently capture spatio-temporal relations between interactive agents and predict their dynamics. Our model infers dynamic interaction graphs in a latent space augmented with interpretable edge features that characterize the interactions. Moreover, we aim to enhance model interpretability and performance in OOD scenarios by disentangling the latent space of edge features, thereby strengthening model versatility and robustness. We validate our approach through extensive experiments on both simulated and real-world datasets. The results show superior performance compared to existing methods in modeling spatio-temporal relations, motion prediction, and identifying time-invariant latent features.

Drug combination refers to the use of two or more drugs to treat a specific disease at the same time. It is currently the mainstream way to treat complex diseases. Compared with single drugs, drug combinations have better efficacy and can better inhibit toxicity and drug resistance. The computational model based on deep learning concatenates the representation of multiple drugs and the corresponding cell line feature as input, and the output is whether the drug combination can have an inhibitory effect on the cell line. However, this strategy of concatenating multiple representations has the following defects: the alignment of drug representation and cell line representation is ignored, resulting in the synergistic relationship not being reflected positionally in the embedding space. Moreover, the alignment measurement function in deep learning cannot be suitable for drug synergy prediction tasks due to differences in input types. Therefore, in this work, we propose ALNSynergy, a graph convolutional network with multi-representation alignment for predicting drug synergy. In the ALNSynergy model, we designed a multi-representation alignment function suitable for the drug synergy prediction task so that the positional relationship between drug representations and cell line representation is reflected in the embedding space. In addition, the vector modulus of drug representations and cell line representation is considered to improve the accuracy of calculation results and accelerate model convergence. Finally, many relevant experiments were run on multiple drug synergy datasets to verify the effectiveness of the above innovative elements and the excellence of the ALNSynergy model., Comment: 9 pages

While most existing works on LLM prompting techniques focus only on how to select a better set of data samples inside one single prompt input (In-Context Learning or ICL), why can not we design and leverage multiple prompts together to further improve the LLM's performance? In this work, we propose In-Context Sampling (ICS), a low-resource LLM prompting technique to produce confident predictions by optimizing the construction of multiple ICL prompt inputs. Extensive experiments with three open-source LLMs (FlanT5-XL, Mistral-7B, and Mixtral-8x7B) on four NLI datasets (e-SNLI, Multi-NLI, ANLI, and Contract-NLI) and one QA dataset (CommonsenseQA) illustrate that ICS can consistently enhance LLMs' performance. An in-depth evaluation with three data similarity-based ICS strategies suggests that these strategies can further elevate LLM's performance, which sheds light on a new yet promising future research direction., Comment: Accepted at NAACL 2024 Findings

Conventional video matting outputs one alpha matte for all instances appearing in a video frame so that individual instances are not distinguished. While video instance segmentation provides time-consistent instance masks, results are unsatisfactory for matting applications, especially due to applied binarization. To remedy this deficiency, we propose Video Instance Matting~(VIM), that is, estimating alpha mattes of each instance at each frame of a video sequence. To tackle this challenging problem, we present MSG-VIM, a Mask Sequence Guided Video Instance Matting neural network, as a novel baseline model for VIM. MSG-VIM leverages a mixture of mask augmentations to make predictions robust to inaccurate and inconsistent mask guidance. It incorporates temporal mask and temporal feature guidance to improve the temporal consistency of alpha matte predictions. Furthermore, we build a new benchmark for VIM, called VIM50, which comprises 50 video clips with multiple human instances as foreground objects. To evaluate performances on the VIM task, we introduce a suitable metric called Video Instance-aware Matting Quality~(VIMQ). Our proposed model MSG-VIM sets a strong baseline on the VIM50 benchmark and outperforms existing methods by a large margin. The project is open-sourced at https://github.com/SHI-Labs/VIM.

This work aims to adapt large-scale pre-trained vision-language models, such as contrastive language-image pretraining (CLIP), to enhance the performance of object reidentification (Re-ID) across various supervision settings. Although prompt learning has enabled a recent work named CLIP-ReID to achieve promising performance, the underlying mechanisms and the necessity of prompt learning remain unclear due to the absence of semantic labels in ReID tasks. In this work, we first analyze the role prompt learning in CLIP-ReID and identify its limitations. Based on our investigations, we propose a simple yet effective approach to adapt CLIP for supervised object Re-ID. Our approach directly fine-tunes the image encoder of CLIP using a prototypical contrastive learning (PCL) loss, eliminating the need for prompt learning. Experimental results on both person and vehicle Re-ID datasets demonstrate the competitiveness of our method compared to CLIP-ReID. Furthermore, we extend our PCL-based CLIP fine-tuning approach to unsupervised scenarios, where we achieve state-of-the art performance.

We present a machine learning (ML) framework for predicting Green's functions of molecular systems, from which photoemission spectra and quasiparticle energies at quantum many-body level can be obtained. Kernel ridge regression is adopted to predict self-energy matrix elements on compact imaginary frequency grids from static and dynamical mean-field electronic features, which gives direct access to real-frequency many-body Green's functions through analytic continuation and Dyson's equation. Feature and self-energy matrices are represented in a symmetry-adapted intrinsic atomic orbital plus projected atomic orbital basis to enforce rotational invariance. We demonstrate good transferability and high data efficiency of proposed ML method across molecular sizes and chemical species by showing accurate predictions of density of states (DOS) and quasiparticle energies at the level of many-body perturbation theory (GW) or full configuration interaction. For the ML model trained on 48 out of 1995 molecules randomly sampled from the QM7 and QM9 datasets, we report the mean absolute errors of ML-predicted HOMO and LUMO energies to be 0.13 eV and 0.10 eV compared to GW@PBE0. We further showcase the capability of this method by applying the same ML model to predict DOS for significantly larger organic molecules with up to 44 heavy atoms., Comment: 9 pages, 5 figures

Prompt-based learning has been demonstrated as a compelling paradigm contributing to large language models' tremendous success (LLMs). Inspired by their success in language tasks, existing research has leveraged LLMs in embodied instruction following and task planning. In this work, we tackle the problem of training a robot to understand multimodal prompts, interleaving vision signals with text descriptions. This type of task poses a major challenge to robots' capability to understand the interconnection and complementarity between vision and language signals. In this work, we introduce an effective framework that learns a policy to perform robot manipulation with multimodal prompts from multi-task expert trajectories. Our methods consist of a two-stage training pipeline that performs inverse dynamics pretraining and multi-task finetuning. To facilitate multimodal understanding, we design our multimodal prompt encoder by augmenting a pretrained LM with a residual connection to the visual input and model the dependencies among action dimensions. Empirically, we evaluate the efficacy of our method on the VIMA-BENCH and establish a new state-of-the-art (10% improvement in success rate). Moreover, we demonstrate that our model exhibits remarkable in-context learning ability. Project page: \url{https://midas-icml.github.io/}., Comment: Accepted by ICML 2024. Project page: https://midas-icml.github.io

The rapid advancement of the Large Language Model (LLM) has created numerous potentials for integration with conversational assistants (CAs) assisting people in their daily tasks, particularly due to their extensive flexibility. However, users' real-world experiences interacting with these assistants remain unexplored. In this research, we chose cooking, a complex daily task, as a scenario to investigate people's successful and unsatisfactory experiences while receiving assistance from an LLM-based CA, Mango Mango. We discovered that participants value the system's ability to provide extensive information beyond the recipe, offer customized instructions based on context, and assist them in dynamically planning the task. However, they expect the system to be more adaptive to oral conversation and provide more suggestive responses to keep users actively involved. Recognizing that users began treating our LLM-CA as a personal assistant or even a partner rather than just a recipe-reading tool, we propose several design considerations for future development., Comment: Under submission to CHI2024

For autonomous vehicles to proactively plan safe trajectories and make informed decisions, they must be able to predict the future occupancy states of the local environment. However, common issues with occupancy prediction include predictions where moving objects vanish or become blurred, particularly at longer time horizons. We propose an environment prediction framework that incorporates environment semantics for future occupancy prediction. Our method first semantically segments the environment and uses this information along with the occupancy information to predict the spatiotemporal evolution of the environment. We validate our approach on the real-world Waymo Open Dataset. Compared to baseline methods, our model has higher prediction accuracy and is capable of maintaining moving object appearances in the predictions for longer prediction time horizons., Comment: Accepted at the 35th IEEE Intelligent Vehicles Symposium 2024

Navigating complex and dynamic environments requires autonomous vehicles (AVs) to reason about both visible and occluded regions. This involves predicting the future motion of observed agents, inferring occluded ones, and modeling their interactions based on vectorized scene representations of the partially observable environment. However, prior work on occlusion inference and trajectory prediction have developed in isolation, with the former based on simplified rasterized methods and the latter assuming full environment observability. We introduce the Scene Informer, a unified approach for predicting both observed agent trajectories and inferring occlusions in a partially observable setting. It uses a transformer to aggregate various input modalities and facilitate selective queries on occlusions that might intersect with the AV's planned path. The framework estimates occupancy probabilities and likely trajectories for occlusions, as well as forecast motion for observed agents. We explore common observability assumptions in both domains and their performance impact. Our approach outperforms existing methods in both occupancy prediction and trajectory prediction in partially observable setting on the Waymo Open Motion Dataset., Comment: Accepted to 2024 IEEE International Conference on Robotics and Automation (ICRA)