Accurately modeling multi-class cell topology is crucial in digital pathology, as it provides critical insights into tissue structure and pathology. The synthetic generation of cell topology enables realistic simulations of complex tissue environments, enhances downstream tasks by augmenting training data, aligns more closely with pathologists' domain knowledge, and offers new opportunities for controlling and generalizing the tumor microenvironment. In this paper, we propose a novel approach that integrates topological constraints into a diffusion model to improve the generation of realistic, contextually accurate cell topologies. Our method refines the simulation of cell distributions and interactions, increasing the precision and interpretability of results in downstream tasks such as cell detection and classification. To assess the topological fidelity of generated layouts, we introduce a new metric, Topological Fréchet Distance (TopoFD), which overcomes the limitations of traditional metrics like FID in evaluating topological structure. Experimental results demonstrate the effectiveness of our approach in generating multi-class cell layouts that capture intricate topological relationships. Code is available at https://github.com/Melon-Xu/TopoCellGen.

Diffusion models excel at creating visually impressive images but often struggle to generate images with a specified topology. The Betti number, which represents the number of structures in an image, is a fundamental measure in topology. Yet, diffusion models fail to satisfy even this basic constraint. This limitation restricts their utility in applications requiring exact control, like robotics and environmental modeling. To address this, we propose TopoDiffusionNet (TDN), a novel approach that enforces diffusion models to maintain the desired topology. We leverage tools from topological data analysis, particularly persistent homology, to extract the topological structures within an image. We then design a topology-based objective function to guide the denoising process, preserving intended structures while suppressing noisy ones. Our experiments across four datasets demonstrate significant improvements in topological accuracy. TDN is the first to integrate topology with diffusion models, opening new avenues of research in this area.

The emergence of Vision-Language Models (VLMs) represents a significant advancement in integrating computer vision with Large Language Models (LLMs) to generate detailed text descriptions from visual inputs. Despite their growing importance, the security of VLMs, particularly against backdoor attacks, is under explored. Moreover, prior works often assume attackers have access to the original training data, which is often unrealistic. In this paper, we address a more practical and challenging scenario where attackers must rely solely on Out-Of-Distribution (OOD) data. We introduce VLOOD (Backdoor Vision-Language Models using Out-of-Distribution Data), a novel approach with two key contributions: (1) demonstrating backdoor attacks on VLMs in complex image-to-text tasks while minimizing degradation of the original semantics under poisoned inputs, and (2) proposing innovative techniques for backdoor injection without requiring any access to the original training data. Our evaluation on image captioning and visual question answering (VQA) tasks confirms the effectiveness of VLOOD, revealing a critical security vulnerability in VLMs and laying the foundation for future research on securing multimodal models against sophisticated threats.

Today, short-videos dominate social media, yet short-video creators lack systematic tools to predict engagement and refine content before uploading. Existing approaches focus primarily on post-publication metrics, failing to address engagement prediction via video editing elements. To address this, we curate VidES, a novel dataset linking short-vdieo engagement to specific editing elements, referred to as Edit Signals (e.g.: narration, text overlays). VidES contains overall engagement scores and detailed human evaluations of Edit Signals. Building on VidES, we propose SmartEdit, a multi-stage framework that predicts short-video engagement and generates actionable improvement suggestions based on Edit Signals. SmartEdit decomposes engagement prediction into interpretable components and uses few-shot Multi-modal Large Language Models to provide specific feedback for content refinement. Experimental results validate SmartEdit’s effectiveness, demonstrating its superiority over zero-shot methods. This work bridges the gap between intuitive and data-driven short-video editing, offering the first dataset and tool tailored for Edit Signal-based short-video refinement.

How can we test AI performance? This question seems trivial, but it isn't. Standard benchmarks often have problems such as in-distribution and small-size test sets, oversimplified metrics, unfair comparisons, and short-term outcome pressure. As a consequence, good performance on standard benchmarks does not guarantee success in real-world scenarios. To address these problems, we present Touchstone, a large-scale collaborative segmentation benchmark of 9 types of abdominal organs. This benchmark is based on 5,195 training CT scans from 76 hospitals around the world and 5,903 testing CT scans from 11 additional hospitals. This diverse test set enhances the statistical significance of benchmark results and rigorously evaluates AI algorithms across various out-of-distribution scenarios. We invited 14 inventors of 19 AI algorithms to train their algorithms, while our team, as a third party, independently evaluated these algorithms on three test sets. In addition, we also evaluated pre-existing AI frameworks---which, differing from algorithms, are more flexible and can support different algorithms—including MONAI from NVIDIA, nnU-Net from DKFZ, and numerous other open-source frameworks. We are committed to expanding this benchmark to encourage more innovation of AI algorithms for the medical domain.

Segmentation of curvilinear structures such as vasculature and road networks is challenging due to relatively weak signals and complex geometry/topology. To facilitate and accelerate large scale annotation, one has to adopt semi-automatic approaches such as proofreading by experts. In this work, we focus on uncertainty estimation for such tasks, so that highly uncertain, and thus error-prone structures can be identified for human annotators to verify. Unlike most existing works, which provide pixel-wise uncertainty maps, we stipulate it is crucial to estimate uncertainty in the units of topological structures, e.g., small pieces of connections and branches. To achieve this, we leverage tools from topological data analysis, specifically discrete Morse theory (DMT), to first capture the structures, and then reason about their uncertainties. To model the uncertainty, we (1) propose a joint prediction model that estimates the uncertainty of a structure while taking the neighboring structures into consideration (inter-structural uncertainty); (2) propose a novel Probabilistic DMT to model the inherent uncertainty within each structure (intra-structural uncertainty) by sampling its representations via a perturb-and-walk scheme. On various 2D and 3D datasets, our method produces better structure-wise uncertainty maps compared to existing works. Code available at: https://github.com/Saumya-Gupta-26/struct-uncertainty

MOTIVATION: Analysis of allele-specific expression is strongly affected by the technical noise present in RNA-seq experiments. Previously, we showed that technical replicates can be used for precise estimates of this noise, and we provided a tool for correction of technical noise in allele-specific expression analysis. This approach is very accurate but costly due to the need for two or more replicates of each library. Here, we develop a spike-in approach which is highly accurate at only a small fraction of the cost. RESULTS: We show that a distinct RNA added as a spike-in before library preparation reflects technical noise of the whole library and can be used in large batches of samples. We experimentally demonstrate the effectiveness of this approach using combinations of RNA from species distinguishable by alignment, namely, mouse, human, and Caenorhabditis elegans. Our new approach, controlFreq, enables highly accurate and computationally efficient analysis of allele-specific expression in (and between) arbitrarily large studies at an overall cost increase of ∼5%. AVAILABILITY AND IMPLEMENTATION: Analysis pipeline for this approach is available at GitHub as R package controlFreq (github.com/gimelbrantlab/controlFreq).

Much work has been done on the design of both interactive narratives and food-based interaction systems, however their intersection has received little attention within the design research community. We took the relative dearth of food-based interactive storytelling systems as an opportunity to explore a compelling design space for multisensory interactive narratives, given these few systems often position people only as passive observers. We report the results of a design workshop conducted among the authors who are design researchers in interactive narratives and food-based play. In the workshop, we discussed existing works, asynchronously brainstormed new food-based interactive storytelling systems, and then reconvened to discuss our ideas. We present the portfolio of designs that arose from this workshop, and the design themes that we synthesized from the portfolio. We recommend how to position people not just as observers, but as active participants in the interactive food-based storytelling system.

In this paper we describe the design and evaluation of The Next Fairy Tale (TNFT) VR, a theatrical interactive storytelling system created in virtual reality and informed by performing arts theories. TNFT was designed to produce opportunities for interactors to experience role-taking and character identification using design principles drawn from actor training and theatrical performance. We report the results of a pilot qualitative study of interactors using TNFT to explore the elements of the design that supported or hindered roleplaying behavior. We identify four design patterns that supported roleplaying in the system: (1) using explicit roles to set player expectations, (2) embracing the "mask and the mirror" effect, (3) attending to visual and interactional details, and (4) easing the player gently into the roleplaying experience. These patterns speak to a broader need to support roleplay through explicit scaffolding of desired player behaviors in digital narrative experiences.

The experience of agency has been centralized in interactive digital storytelling research. Current approaches to agency place the actions of a reader into direct conflict with the authorial intentions of an interactive storyteller by privileging the authorial pleasures that come by interacting with the story's plot. In this paper we present the design and results of an empirical study of Shiva's Rangoli, a tangible interactive narrative system that supports meaningful interactions with the narratives that are decoupled from character actions and plot outcomes. Shiva's Rangoli allows readers to change the emotional, presentational, and aesthetic context of the story by giving them control over the ambience of the installation space where it is experienced, through a tangible interface. From our study, we identify three interaction styles or strategies undertaken by readers that reframe readers as Story Supporters, Meaning Makers, and Story Controllers to explore the pleasures of bounded agency within interactive storytelling.

This paper describes the underlying motivation, creation process, and evaluation outcomes of Shiva's Rangoli, a tangible storytelling installation that allows readers to impact the emotional tone of a narrative by sculpting the ambience of their space. Readers interact with a tangible interface that acts as a boundary object between the reader and the fictional world. We discuss how these kinds of interfaces can engage readers to feel like they are a part of the story, endow them with responsibility, and blur the line between real and fictional worlds.

Social communication frequently includes nuanced nonverbal communication cues, including eye contact, gestures, facial expressions, body language, and tone of voice. This type of communication is central to face-to-face interaction, but can be challenging for children and adults with autism. Innovative technologies can provide support by augmenting human-delivered cuing and automated prompting. Specifically, immersive virtual reality (VR) offers an option to generalize social skill interventions by concretizing nonverbal information in real-time social interactions. In this work, we explore the design and evaluation of three nonverbal communication applications in immersive VR. The results of this work indicate that delivering real-time visualizations of proximity, speaker volume, and duration of one's speech is feasible in immersive VR and effective for real-time support for proximity regulation for children with autism. We conclude with design considerations for therapeutic VR systems.

This paper presents a mathematical framework for dynamic congestion pricing. The objective is to calculate an optimal toll using the optimal control theory. The problem consists of tolled lanes or routes and alternate non-tolled lanes or routes. The model is developed using a traffic conservation law, the queuing theory, and fundamental macroscopic relationships. A logit model is used for establishing the relationship between the price and the driver's choice behavior. We design a cost function and then use Hamilton-Jacobi-Bellman equation to derive an optimal control law that uses real-time information to determine an optimal tolling price. Simulations are performed to demonstrate the performance of this optimal control congestion-pricing algorithm.