Search

Despite being an integral tool for finding health-related information online, YouTube has faced criticism for disseminating COVID-19 misinformation globally to its users. Yet, prior audit studies have predominantly investigated YouTube within the Global North contexts, often overlooking the Global South. To address this gap, we conducted a comprehensive 10-day geolocation-based audit on YouTube to compare the prevalence of COVID-19 misinformation in search results between the United States (US) and South Africa (SA), the countries heavily affected by the pandemic in the Global North and the Global South, respectively. For each country, we selected 3 geolocations and placed sock-puppets, or bots emulating "real" users, that collected search results for 48 search queries sorted by 4 search filters for 10 days, yielding a dataset of 915K results. We found that 31.55% of the top-10 search results contained COVID-19 misinformation. Among the top-10 search results, bots in SA faced significantly more misinformative search results than their US counterparts. Overall, our study highlights the contrasting algorithmic behaviors of YouTube search between two countries, underscoring the need for the platform to regulate algorithmic behavior consistently across different regions of the Globe., Comment: 28 pages. Under submission

In this article, we focus on the pre-training of visual autonomous driving agents in the context of imitation learning. Current methods often rely on a classification-based pre-training, which we hypothesise to be holding back from extending capabilities of implicit image understanding. We propose pre-training the visual encoder of a driving agent using the self-distillation with no labels (DINO) method, which relies on a self-supervised learning paradigm.% and is trained on an unrelated task. Our experiments in CARLA environment in accordance with the Leaderboard benchmark reveal that the proposed pre-training is more efficient than classification-based pre-training, and is on par with the recently proposed pre-training based on visual place recognition (VPRPre).

On the whole, the U.S. Algorithmic Accountability Act of 2022 (US AAA) is a pragmatic approach to balancing the benefits and risks of automated decision systems. Yet there is still room for improvement. This commentary highlights how the US AAA can both inform and learn from the European Artificial Intelligence Act (EU AIA)., Comment: Minds & Machines (2022)

Given a task in the form of a basic description and its training examples, prompt optimization is the problem of synthesizing the given information into a text prompt for a large language model (LLM). Humans solve this problem by also considering the different facets that define a task (e.g., counter-examples, explanations, analogies) and including them in the prompt. However, it is unclear whether existing algorithmic approaches, based on iteratively editing a given prompt or automatically selecting a few in-context examples, can cover the multiple facets required to solve a complex task. In this work, we view prompt optimization as that of learning multiple facets of a task from a set of training examples. We identify and exploit structure in the prompt optimization problem -- first, we find that prompts can be broken down into loosely coupled semantic sections that have a relatively independent effect on the prompt's performance; second, we cluster the input space and use clustered batches so that the optimization procedure can learn the different facets of a task across batches. The resulting algorithm, UniPrompt, consists of a generative model to generate initial candidates for each prompt section; and a feedback mechanism that aggregates suggested edits from multiple mini-batches into a conceptual description for the section. Empirical evaluation on multiple datasets and a real-world task shows that prompts generated using UniPrompt obtain higher accuracy than human-tuned prompts and those from state-of-the-art methods. In particular, our algorithm can generate long, complex prompts that existing methods are unable to generate. Code for UniPrompt will be available at \url{https://aka.ms/uniprompt}.

Attention mechanisms are becoming increasingly popular, being used in neural network models in multiple domains such as natural language processing (NLP) and vision applications, especially at the edge. However, attention layers are difficult to map onto existing neuro accelerators since they have a much higher density of non-linear operations, which lead to inefficient utilization of today's vector units. This work introduces NOVA, a NoC-based Vector Unit that can perform non-linear operations within the NoC of the accelerators, and can be overlaid onto existing neuro accelerators to map attention layers at the edge. Our results show that the NOVA architecture is up to 37.8x more power-efficient than state-of-the-art hardware approximators when running existing attention-based neural networks., Comment: 6 pages, 8 figures

Recent studies have shown direct ion heating (vashistha2020new,Juneja_2023) by lasers EM (Electromagnetic) wave interacting with a plasma threaded by an external uniform magnetic field. The EM wave frequency was near the lower hybrid (LH) resonance frequency. The LH resonance occurs at the edge of the pass band of the magnetized dispersion relation. The group speed of the wave is negligible at resonance. In these studies, the energy absorption remains essentially confined at the plasma surface. However, to heat the ions in the bulk plasma and at a desired location, a tailored inhomogeneous external magnetic field profile has been chosen here. The strength of the magnetic field at the plasma edge is such that the EM wave frequency lies inside the pass band, where the group velocity has a significant value. It enables the wave to enter the bulk plasma. The external magnetic field is then spatially tailored appropriately to have the LH resonance at a desired spatial location inside the plasma. The Particle-In-Cell (PIC) simulations using the OSIRIS4.0 platform have been carried out, which demonstrates that the EM wave pulse comes to a standstill at the location of the resonance. The wave pulse is observed to break down subsequently, and the energy consequently goes dominantly to the local plasma ions. The absorption is significantly enhanced compared to the case in which the magnetic field profile was homogeneous. The dependence of absorption on the choice of magnetic field profile, the laser intensity, etc., has also been carried out., Comment: 10 pages, 17 figures

The article argues that AI can enhance the measurement and implementation of democratic processes within political parties, known as Intra-Party Democracy (IPD). It identifies the limitations of traditional methods for measuring IPD, which often rely on formal parameters, self-reported data, and tools like surveys. Such limitations lead to the collection of partial data, rare updates, and significant demands on resources. To address these issues, the article suggests that specific data management and Machine Learning (ML) techniques, such as natural language processing and sentiment analysis, can improve the measurement (ML about) and practice (ML for) of IPD. The article concludes by considering some of the principal risks of ML for IPD, including concerns over data privacy, the potential for manipulation, and the dangers of overreliance on technology.

Despite demonstrating emergent reasoning abilities, Large Language Models (LLMS) often lose track of complex, multi-step reasoning. Existing studies show that providing guidance via decomposing the original question into multiple subproblems elicits more robustness in LLM reasoning -- a decomposer generates the subproblems, and a solver solves each of these subproblems. However, these techniques fail to accommodate coordination between the decomposer and the solver modules (either in a single model or different specialized ones) -- the decomposer does not keep track of the ability of the solver to follow the decomposed reasoning. In this paper, we propose LM2 to address these challenges. LM2 modularizes the decomposition, solution, and verification into three different language models. The decomposer module identifies the key concepts necessary to solve the problem and generates step-by-step subquestions according to the reasoning requirement. The solver model generates the solution to the subproblems that are then checked by the verifier module; depending upon the feedback from the verifier, the reasoning context is constructed using the subproblems and the solutions. These models are trained to coordinate using policy learning. Exhaustive experimentation suggests the superiority of LM2 over existing methods on in- and out-domain reasoning problems, outperforming the best baselines by $8.1\%$ on MATH, $7.71\%$ on JEEBench, and $9.7\%$ on MedQA problems (code available at https://github.com/LCS2-IIITD/Language_Model_Multiplex).

Top-$2$ methods have become popular in solving the best arm identification (BAI) problem. The best arm, or the arm with the largest mean amongst finitely many, is identified through an algorithm that at any sequential step independently pulls the empirical best arm, with a fixed probability $\beta$, and pulls the best challenger arm otherwise. The probability of incorrect selection is guaranteed to lie below a specified $\delta >0$. Information theoretic lower bounds on sample complexity are well known for BAI problem and are matched asymptotically as $\delta \rightarrow 0$ by computationally demanding plug-in methods. The above top 2 algorithm for any $\beta \in (0,1)$ has sample complexity within a constant of the lower bound. However, determining the optimal $\beta$ that matches the lower bound has proven difficult. In this paper, we address this and propose an optimal top-2 type algorithm. We consider a function of allocations anchored at a threshold. If it exceeds the threshold then the algorithm samples the empirical best arm. Otherwise, it samples the challenger arm. We show that the proposed algorithm is optimal as $\delta \rightarrow 0$. Our analysis relies on identifying a limiting fluid dynamics of allocations that satisfy a series of ordinary differential equations pasted together and that describe the asymptotic path followed by our algorithm. We rely on the implicit function theorem to show existence and uniqueness of these fluid ode's and to show that the proposed algorithm remains close to the ode solution.

There has been significant interest lately in the study of Electromagnetic (EM) waves interacting with magnetized plasmas. The variety of resonances and the existence of several pass and stop bands in the dispersion curve for different orientations of the magnetic field offer new mechanisms of EM wave energy absorption (PhysRevE.105.055209, Juneja_2023,vashistha2022localized,vashistha2020new). By an appropriate choice of inhomogeneous magnetic field, one can construct a configuration wherein the same EM wave pulse encounters more than one resonance in the plasma. A 2-D Particle - In - Cell (PIC) simulation using the OSIRIS4.0 platform has been carried out for the case of dual resonance. It is observed that in the presence of dual resonance, there is a significant enhancement in leading to almost complete absorption of laser energy by the plasma in certain cases. A detailed study of the influence of the relative location of the resonances, the effect of high input EM wave intensity, etc., has also been carried out., Comment: 11 pages, 10 figures

As more users turn to video-sharing platforms like YouTube as an information source, they may consume misinformation despite their best efforts. In this work, we investigate ways that users can better assess the credibility of videos by first exploring how users currently determine credibility using existing signals on platforms and then by introducing and evaluating new credibility-based signals. We conducted 12 contextual inquiry interviews with YouTube users, determining that participants used a combination of existing signals, such as the channel name, the production quality, and prior knowledge, to evaluate credibility, yet sometimes stumbled in their efforts to do so. We then developed Viblio, a prototype system that enables YouTube users to view and add citations and related information while watching a video based on our participants' needs. From an evaluation with 12 people, all participants found Viblio to be intuitive and useful in the process of evaluating a video's credibility and could see themselves using Viblio in the future.

In this draft we consider the problem of forecasting rainfall across India during the four monsoon months, one day as well as three days in advance. We train neural networks using historical daily gridded precipitation data for India obtained from IMD for the time period $1901- 2022$, at a spatial resolution of $1^{\circ} \times 1^{\circ}$. This is compared with the numerical weather prediction (NWP) forecasts obtained from NCEP (National Centre for Environmental Prediction) available for the period 2011-2022. We conduct a detailed country wide analysis and separately analyze some of the most populated cities in India. Our conclusion is that forecasts obtained by applying deep learning to historical rainfall data are more accurate compared to NWP forecasts as well as predictions based on persistence. On average, compared to our predictions, forecasts from NCEP-NWP model have about 34% higher error for a single day prediction, and over 68% higher error for a three day prediction. Similarly, persistence estimates report a 29% higher error in a single day forecast, and over 54% error in a three day forecast. We further observe that data up to 20 days in the past is useful in reducing errors of one and three day forecasts, when a transformer based learning architecture, and to a lesser extent when an LSTM is used. A key conclusion suggested by our preliminary analysis is that NWP forecasts can be substantially improved upon through more and diverse data relevant to monsoon prediction combined with carefully selected neural network architecture.

There is a growing demand for transparency in search engines to understand how search results are curated and to enhance users' trust. Prior research has introduced search result explanations with a focus on how to explain, assuming explanations are beneficial. Our study takes a step back to examine if search explanations are needed and when they are likely to provide benefits. Additionally, we summarize key characteristics of helpful explanations and share users' perspectives on explanation features provided by Google and Bing. Interviews with non-technical individuals reveal that users do not always seek or understand search explanations and mostly desire them for complex and critical tasks. They find Google's search explanations too obvious but appreciate the ability to contest search results. Based on our findings, we offer design recommendations for search engines and explanations to help users better evaluate search results and enhance their search experience.

Diffusion models when conditioned on text prompts, generate realistic-looking images with intricate details. But most of these pre-trained models fail to generate accurate images when it comes to human features like hands, teeth, etc. We hypothesize that this inability of diffusion models can be overcome through well-annotated good-quality data. In this paper, we look specifically into improving the hand-object-interaction image generation using diffusion models. We collect a well annotated hand-object interaction synthetic dataset curated using Prompt-Propose-Verify framework and finetune a stable diffusion model on it. We evaluate the image-text dataset on qualitative and quantitative metrics like CLIPScore, ImageReward, Fedility, and alignment and show considerably better performance over the current state-of-the-art benchmarks., Comment: Accepted at International Workshop on AI for Digital Human in AAAI Conference on Articial Intelligence (AAAI, 2024)

Privacy labels -- standardized, compact representations of data collection and data use practices -- are often presented as a solution to the shortcomings of privacy policies. Apple introduced mandatory privacy labels for apps in its App Store in December 2020; Google introduced mandatory labels for Android apps in July 2022. iOS app privacy labels have been evaluated and critiqued in prior work. In this work, we evaluated Android Data Safety Labels and explored how differences between the two label designs impact user comprehension and label utility. We conducted a between-subjects, semi-structured interview study with 12 Android users and 12 iOS users. While some users found Android Data Safety Labels informative and helpful, other users found them too vague. Compared to iOS App Privacy Labels, Android users found the distinction between data collection groups more intuitive and found explicit inclusion of omitted data collection groups more salient. However, some users expressed skepticism regarding elided information about collected data type categories. Most users missed critical information due to not expanding the accordion interface, and they were surprised by collection practices excluded from Android's definitions. Our findings also revealed that Android users generally appreciated information about security practices included in the labels, and iOS users wanted that information added., Comment: This paper has been accepted to Privacy Enhancing Technologies Symposium (PETS) 2024

The guiding and control of ultrahigh flux, femtosecond relativistic electron pulses through solid density matter is of great importance for many areas of high energy density science. Efforts so far include the use of magnetic fields generated by the propagation of the electron pulse itself or the application of hundreds of Tesla magnitudes, pulsed external magnetic fields driven by either short pulse lasers or electrical pulses. Here we experimentally demonstrate the guiding of hundreds of keV mega-ampere electron pulses in a magnetized neodymium solid that has a very modest, easily available static field of 0.1 tesla. The electron pulses driven by an ultrahigh intensity, 30 femtosecond laser are shown to propagate beam-like, a distance as large as 5 mm in a high Z target (neodymium), their collimation improved and flux density enhanced nearly by a factor of 3. Particle-in-cell simulations in the appropriate parameter regime match the experimental observations. In addition, the simulations predict the occurrence of a novel, near-monochromatic feature towards the high energy end of the electron energy spectrum, which is tunable by the applied magnetic field strength. These results may prove valuable for fast electron beam-driven radiation sources, fast ignition of laser fusion, and laboratory astrophysics., Comment: 10 pages, 5 figures

We consider a queuing network that opens at a specified time, where customers are non-atomic and belong to different classes. Each class has its own route, and as is typical in the literature, the costs are a linear function of waiting and service completion time. We restrict ourselves to a two class, two queue network: this simplification is well motivated as the diversity in solution structure as a function of problem parameters is substantial even in this simple setting (e.g., a specific routing structure involves eight different regimes), suggesting a combinatorial blow up as the number of queues, routes and customer classes increase. We identify the unique Nash equilibrium customer arrival profile when the customer linear cost preferences are different. This profile is a function of problem parameters including the size of each class, service rates at each queue, and customer cost preferences. When customer cost preferences match, under certain parametric settings, the equilibrium arrival profiles may not be unique and may lie in a convex set. We further make a surprising observation that in some parametric settings, customers in one class may arrive in disjoint intervals. Further, the two classes may arrive in contiguous intervals or in overlapping intervals, and at varying rates within an interval, depending upon the problem parameters.

Large Language Models (LLMs) prompted to generate chain-of-thought (CoT) exhibit impressive reasoning capabilities. Recent attempts at prompt decomposition toward solving complex, multi-step reasoning problems depend on the ability of the LLM to simultaneously decompose and solve the problem. A significant disadvantage is that foundational LLMs are typically not available for fine-tuning, making adaptation computationally prohibitive. We believe (and demonstrate) that problem decomposition and solution generation are distinct capabilites, better addressed in separate modules, than by one monolithic LLM. We introduce DaSLaM, which uses a decomposition generator to decompose complex problems into subproblems that require fewer reasoning steps. These subproblems are answered by a solver. We use a relatively small (13B parameters) LM as the decomposition generator, which we train using policy gradient optimization to interact with a solver LM (regarded as black-box) and guide it through subproblems, thereby rendering our method solver-agnostic. Evaluation on multiple different reasoning datasets reveal that with our method, a 175 billion parameter LM (text-davinci-003) can produce competitive or even better performance, compared to its orders-of-magnitude larger successor, GPT-4. Additionally, we show that DaSLaM is not limited by the solver's capabilities as a function of scale; e.g., solver LMs with diverse sizes give significant performance improvement with our solver-agnostic decomposition technique. Exhaustive ablation studies evince the superiority of our modular finetuning technique over exorbitantly large decomposer LLMs, based on prompting alone., Comment: EMNLP 2023 (Typos corrected)

Learning paradigms based purely on offline data as well as those based solely on sequential online learning have been well-studied in the literature. In this paper, we consider combining offline data with online learning, an area less studied but of obvious practical importance. We consider the stochastic $K$-armed bandit problem, where our goal is to identify the arm with the highest mean in the presence of relevant offline data, with confidence $1-\delta$. We conduct a lower bound analysis on policies that provide such $1-\delta$ probabilistic correctness guarantees. We develop algorithms that match the lower bound on sample complexity when $\delta$ is small. Our algorithms are computationally efficient with an average per-sample acquisition cost of $\tilde{O}(K)$, and rely on a careful characterization of the optimality conditions of the lower bound problem., Comment: 45 pages, 5 figures

As the use of online platforms continues to grow across all demographics, users often express a desire to feel represented in the content. To improve representation in search results and recommendations, we introduce end-to-end diversification, ensuring that diverse content flows throughout the various stages of these systems, from retrieval to ranking. We develop, experiment, and deploy scalable diversification mechanisms in multiple production surfaces on the Pinterest platform, including Search, Related Products, and New User Homefeed, to improve the representation of different skin tones in beauty and fashion content. Diversification in production systems includes three components: identifying requests that will trigger diversification, ensuring diverse content is retrieved from the large content corpus during the retrieval stage, and finally, balancing the diversity-utility trade-off in a self-adjusting manner in the ranking stage. Our approaches, which evolved from using Strong-OR logical operator to bucketized retrieval at the retrieval stage and from greedy re-rankers to multi-objective optimization using determinantal point processes for the ranking stage, balances diversity and utility while enabling fast iterations and scalable expansion to diversification over multiple dimensions. Our experiments indicate that these approaches significantly improve diversity metrics, with a neutral to a positive impact on utility metrics and improved user satisfaction, both qualitatively and quantitatively, in production. An accessible PDF of this article is available at https://drive.google.com/file/d/1p5PkqC-sdtX19Y_IAjZCtiSxSEX1IP3q/view, Comment: In Proceedings of the 2023 ACM Conference on Fairness, Accountability, and Transparency (FAccT '23), June 12--15, 2023, Chicago, IL, USA

Knowledge graphs are an increasingly common data structure for representing biomedical information. These knowledge graphs can easily represent heterogeneous types of information, and many algorithms and tools exist for querying and analyzing graphs. Biomedical knowledge graphs have been used in a variety of applications, including drug repurposing, identification of drug targets, prediction of drug side effects, and clinical decision support. Typically, knowledge graphs are constructed by centralization and integration of data from multiple disparate sources. Here, we describe BioThings Explorer, an application that can query a virtual, federated knowledge graph derived from the aggregated information in a network of biomedical web services. BioThings Explorer leverages semantically precise annotations of the inputs and outputs for each resource, and automates the chaining of web service calls to execute multi-step graph queries. Because there is no large, centralized knowledge graph to maintain, BioThing Explorer is distributed as a lightweight application that dynamically retrieves information at query time. More information can be found at https://explorer.biothings.io, and code is available at https://github.com/biothings/biothings_explorer.