Search

Installing probabilistic world models into artificial agents opens an efficient channel for humans to communicate with and control these agents. In addition to updating agent policies, humans can modify their internal world models in order to influence their decisions. The challenge, however, is that currently existing world models are difficult for humans to adapt because they lack a natural communication interface. Aimed at addressing this shortcoming, we develop Language-Guided World Models (LWMs), which can capture environment dynamics by reading language descriptions. These models enhance agent communication efficiency, allowing humans to simultaneously alter their behavior on multiple tasks with concise language feedback. They also enable agents to self-learn from texts originally written to instruct humans. To facilitate the development of LWMs, we design a challenging benchmark based on the game of MESSENGER (Hanjie et al., 2021), requiring compositional generalization to new language descriptions and environment dynamics. Our experiments reveal that the current state-of-the-art Transformer architecture performs poorly on this benchmark, motivating us to design a more robust architecture. To showcase the practicality of our proposed LWMs, we simulate a scenario where these models augment the interpretability and safety of an agent by enabling it to generate and discuss plans with a human before execution. By effectively incorporating language feedback on the plan, the models boost the agent performance in the real environment by up to three times without collecting any interactive experiences in this environment.

Biosensors have become a prominent field of research in a world where detection is one of the key methods to prevent spread of disease and obtain medical information in a patient. One of the most cost-effective ways to currently do this involves paper-based analytical devices which provide quick feedback on an analyte of interest. The commonly used method, a lateral flow assay with a colorimetric readout, has currently been one of the key proponents of the field with constant research being done on this subject. However, significant limitations to this method are the sensitivity of the test as well as an immobilization requirement for the reactant, requiring extra steps. We suggest combining paper-based devices with a previously established readout method, transient induced molecular electronic signal (TIMES), which uses electrochemical signals to determine if a reaction has occurred. By reading the induced charge that comes from the chemical reaction or analyte being observed, immobilization of the analyte can be bypassed. Furthermore, a direct signal can be generated and analyzed to confirm results rather than relying on a weak colorimetric readout. This method will not only lower the amount, and therefore, cost of the material being used, but also simplify the overall process by skipping the immobilization step. In this paper, we use this method to study protein and antibody binding interactions to confirm the validity of this technique. By showing how the results fit to kinetic standards, we can open up the opportunity for further studies on this method.

Biosensors have become a prominent field of research in a world where detection is one of the key methods to prevent spread of disease and obtain medical information in a patient. One of the most cost-effective ways to currently do this involves paper-based analytical devices which provide quick feedback on an analyte of interest. The commonly used method, a lateral flow assay with a colorimetric readout, has currently been one of the key proponents of the field with constant research being done on this subject. However, significant limitations to this method are the sensitivity of the test as well as an immobilization requirement for the reactant, requiring extra steps. We suggest combining paper-based devices with a previously established readout method, transient induced molecular electronic signal (TIMES), which uses electrochemical signals to determine if a reaction has occurred. By reading the induced charge that comes from the chemical reaction or analyte being observed, immobilization of the analyte can be bypassed. Furthermore, a direct signal can be generated and analyzed to confirm results rather than relying on a weak colorimetric readout. This method will not only lower the amount, and therefore, cost of the material being used, but also simplify the overall process by skipping the immobilization step. In this paper, we use this method to study protein and antibody binding interactions to confirm the validity of this technique. By showing how the results fit to kinetic standards, we can open up the opportunity for further studies on this method.

In the contemporary digital landscape, access to timely and relevant information during crisis events is crucial for effective decision-making and response coordination. This project addresses the need for a specialized web application equipped with a sophisticated crawler system to streamline the process of collecting pertinent information related to a user-specified crisis event. The inherent challenge lies in the vast and dynamic nature of online content, where identifying and extracting valuable data from a multitude of sources can be overwhelming. This project aims to empower users by allowing them to input a list of newline-delimited URLs associated with the crisis at hand. The embedded crawler software then systematically traverses these URLs, extracting additional outgoing links for further exploration. Afterwards, the contents of each outgoing URL is then run through a predict function, which evaluates the relevance of each URL based on a scoring system ranging from 0 to 1. This scoring mechanism serves as a critical filter, ensuring that the collected web pages are not only related to the specified crisis event but also possess a significant degree of pertinence. We allow the user to set these thresholds, which enhances the efficiency of information retrieval by prioritizing content most likely to be valuable to the user's needs. Throughout the crawling process, our system tracks a range of statistics, including individual website domains, the origin of each child URL, and the average score assigned to each domain. To provide users with a comprehensive and visually intuitive experience, our user interface leverages React and D3 to display these statistics effectively. Moreover, to enhance user engagement and customization, our platform allows users to create individual accounts. This feature not only provides a personalized experience but also grants users access to a historical record of every crawl they have executed. Users are further empowered with the

Transaction fees represent a major incentive in many blockchain systems as a way to incentivize processing transactions. Unfortunately, they also introduce an enormous amount of incentive asymmetry compared to alternatives like fixed block rewards. We analyze some of the incentive compatibility issues that arise from transaction fees, which relate to the bids that users submit, the allocation rules that miners use to choose which transactions to include, and where they choose to mine in the context of longest-chain consensus. We start by surveying a variety of mining attacks including undercutting, fee sniping, and fee-optimized selfish mining. Then, we move to analyzing mechanistic notions of user incentive compatibility, myopic miner incentive compatibility, and off-chain-agreement-proofness, as well as why they are provably incompatible in their full form. Then, we discuss weaker notions of nearly and $\gamma$-weak incentive compatibility, and how all of these forms of incentive compatibility hold or fail in the trustless auctioneer setup of blockchains, examining classical mechanisms as well as more recent ones such as Ethereum's EIP-1559 mechanism and \cite{chung}'s burning second-price auction. Throughout, we generalize and interrelate existing notions, provide new unifying perspectives and intuitions on analysis, and discuss both specific and overarching open problems for future work.

Existing benchmarks for evaluating long video understanding falls short on two critical aspects, either lacking in scale or quality of annotations. These limitations arise from the difficulty in collecting dense annotations for long videos, which often require manually labeling each frame. In this work, we introduce an automated Annotation and Video Stream Alignment Pipeline (abbreviated ASAP). We demonstrate the generality of ASAP by aligning unlabeled videos of four different sports with corresponding freely available dense web annotations (i.e. commentary). We then leverage ASAP scalability to create LCric, a large-scale long video understanding benchmark, with over 1000 hours of densely annotated long Cricket videos (with an average sample length of ~50 mins) collected at virtually zero annotation cost. We benchmark and analyze state-of-the-art video understanding models on LCric through a large set of compositional multi-choice and regression queries. We establish a human baseline that indicates significant room for new research to explore. Our human studies indicate that ASAP can align videos and annotations with high fidelity, precision, and speed. The dataset along with the code for ASAP and baselines can be accessed here: https://asap-benchmark.github.io/.

The use of appearance codes in recent work on generative modeling has enabled novel view renders with variable appearance and illumination, such as day-time and night-time renders of a scene. A major limitation of this technique is the need to re-train new appearance codes for every scene on inference, so in this work we address this problem proposing a framework that learns a joint embedding space for the appearance and structure of the scene by enforcing a contrastive loss constraint between different modalities. We apply our framework to a simple Variational Auto-Encoder model on the RADIATE dataset \cite{sheeny2021radiate} and qualitatively demonstrate that we can generate new renders of night-time photos using day-time appearance codes without additional optimization iterations. Additionally, we compare our model to a baseline VAE that uses the standard per-image appearance code technique and show that our approach achieves generations of similar quality without learning appearance codes for any unseen images on inference.

The human recognition system has presented the remarkable ability to effortlessly learn novel knowledge from only a few trigger events based on prior knowledge, which is called insight learning. Mimicking such behavior on Knowledge Graph Reasoning (KGR) is an interesting and challenging research problem with many practical applications. Simultaneously, existing works, such as knowledge embedding and few-shot learning models, have been limited to conducting KGR in either “seen-to-seen” or “unseen-to-unseen” scenarios. To this end, we propose a neural insight learning framework named Eureka to bridge the “seen” to “unseen” gap. Eureka is empowered to learn the seen relations with sufficient training triples while providing the flexibility of learning unseen relations given only one trigger without sacrificing its performance on seen relations. Eureka meets our expectation of the model to acquire seen and unseen relations at no extra cost, and eliminate the need to retrain when encountering emerging unseen relations. Experimental results on two real-world datasets demonstrate that the proposed framework also outperforms various state-of-the-art baselines on datasets of both seen and unseen relations.

Click-through rate (CTR) prediction plays a critical role in recommender systems and online advertising. The data used in these applications are multi-field categorical data, where each feature belongs to one field. Field information is proved to be important and there are several works considering fields in their models. In this paper, we proposed a novel approach to model the field information effectively and efficiently. The proposed approach is a direct improvement of FwFM, and is named as Field-matrixed Factorization Machines (FmFM, or $FM^2$). We also proposed a new explanation of FM and FwFM within the FmFM framework, and compared it with the FFM. Besides pruning the cross terms, our model supports field-specific variable dimensions of embedding vectors, which acts as soft pruning. We also proposed an efficient way to minimize the dimension while keeping the model performance. The FmFM model can also be optimized further by caching the intermediate vectors, and it only takes thousands of floating-point operations (FLOPs) to make a prediction. Our experiment results show that it can out-perform the FFM, which is more complex. The FmFM model's performance is also comparable to DNN models which require much more FLOPs in runtime., Comment: In Proceedings of the Web Conference 2021 (WWW 2021), April 19-23, 2021, Ljubljana, Slovenia. 10 pages

Although a flow cytometer, being one of the most popular research and clinical tools for biomedicine, can analyze cells based on the cell size, internal structures such as granularity, and molecular markers, it provides little information about the physical properties of cells such as cell stiffness and physical interactions between the cell membrane and fluid. In this paper, we propose a computational cell analysis technique using cells' different equilibrium positions in a laminar flow. This method utilizes a spatial coding technique to acquire the spatial position of the cell in a microfluidic channel and then uses mathematical algorithms to calculate the ratio of cell mixtures. Most uniquely, the invented computational cell analysis technique can unequivocally detect the subpopulation of each cell type without labeling even when the cell type shows a substantial overlap in the distribution plot with other cell types, a scenario limiting the use of conventional flow cytometers and machine learning techniques. To prove this concept, we have applied the computation method to distinguish live and fixed cancer cells without labeling, count neutrophils from human blood, and distinguish drug treated cells from untreated cells. Our work paves the way for using computation algorithms and fluidic dynamic properties for cell classification, a label-free method that can potentially classify over 200 types of human cells. Being a highly cost-effective cell analysis method complementary to flow cytometers, our method can offer orthogonal tests in companion with flow cytometers to provide crucial information for biomedical samples.

High-throughput single cell imaging is a critical enabling and driving technology in molecular and cellular biology, biotechnology, medicine and related areas. Imaging flow cytometry combines the single-cell imaging capabilities of microscopy with the high-throughput capabilities of conventional flow cytometry. Recent advances in imaging flow cytometry are remarkably revolutionizing single-cell analysis. This article describes recent imaging flow cytometry technologies and their challenges.

1

1