Ayush Agarwal

Ayush Agarwal, Ansh Gandhi, Jeremy A. Collins et al.

The scarcity of large-scale, high-quality demonstration data remains a bottleneck in scaling imitation learning for robotic manipulation. We present COBALT, a teleoperation platform designed to democratize robot learning at scale both in simulation and in the real world. By leveraging vectorized environments, our scalable, load-balanced infrastructure supports concurrent teleoperation by multiple users on a single GPU, yielding a significant reduction in teleoperation cost. Operators can connect from nearly anywhere on Earth using commonly available devices, including single or dual smartphones, VR headsets, 3D mice, and keyboards. An inmemory data cache and efficient video streaming keep control and rendering synchronous, sustaining dozens of concurrent users at 20 Hz with sub-100 ms end-to-end latency for up to 8 concurrent users per GPU. We also demonstrate stable operation supporting 256 simulated clients across 8 GPUs, underscoring the system's ability to scale across hardware and within individual servers. We perform a comprehensive user study showing that phone-based teleoperation performs comparably to or better than specialized hardware, enabling faster, more ergonomic data collection. To ensure data quality, COBALT logs a suite of real-time metrics to automatically filter suboptimal demonstrations. We further demonstrate that a structured user training curriculum significantly improves data collection quality. Guided by insights from our user study, we crowdsource the collection of a large-scale, high-quality pilot dataset with 7500+ demonstrations (50+ hours) collected with smartphones across nine countries over five days. We validate the dataset's quality by training state-of-the-art imitation learning algorithms. Please visit \href{https://cobalt-teleop.github.io/}{cobalt-teleop.github.io} for more details.

Luyi Ma, Wanjia Sherry Zhang, Zezhong Fan et al.

On online advertising platforms, newly introduced promotional ads face the cold-start problem, as they lack sufficient user feedback for model training. In this work, we propose LLM-HYPER, a novel framework that treats large language models (LLMs) as hypernetworks to directly generate the parameters of the click-through rate (CTR) estimator in a training-free manner. LLM-HYPER uses few-shot Chain-of-Thought prompting over multimodal ad content (text and images) to infer feature-wise model weights for a linear CTR predictor. By retrieving semantically similar past campaigns via CLIP embeddings and formatting them into prompt-based demonstrations, the LLM learns to reason about customer intent, feature influence, and content relevance. To ensure numerical stability and serviceability, we introduce normalization and calibration techniques that align the generated weights with production-ready CTR distributions. Extensive offline experiments show that LLM-HYPER significantly outperforms cold-start baselines in NDCG$@10$ by 55.9\%. Our real-world online A/B test on one of the top e-commerce platforms in the U.S. demonstrates the strong performance of LLM-HYPER, which drastically reduces the cold-start period and achieves competitive performance. LLM-HYPER has been successfully deployed in production.

Manasi Patwardhan, Ayush Agarwal, Shabbirhussain Bhaisaheb et al.

The performance of Large Language Models (LLMs) for translating Natural Language (NL) queries into SQL varies significantly across databases (DBs). NL queries are often expressed using a domain specific vocabulary, and mapping these to the correct SQL requires an understanding of the embedded domain expressions, their relationship to the DB schema structure. Existing benchmarks rely on unrealistic, ad-hoc query specific textual hints for expressing domain knowledge. In this paper, we propose a systematic framework for associating structured domain statements at the database level. We present retrieval of relevant structured domain statements given a user query using sub-string level match. We evaluate on eleven realistic DB schemas covering diverse domains across five open-source and proprietary LLMs and demonstrate that (1) DB level structured domain statements are more practical and accurate than existing ad-hoc query specific textual domain statements, and (2) Our sub-string match based retrieval of relevant domain statements provides significantly higher accuracy than other retrieval approaches.

Mansi Goel, Ayush Agarwal, Shubham Agrawal et al.

Food touches our lives through various endeavors, including flavor, nourishment, health, and sustainability. Recipes are cultural capsules transmitted across generations via unstructured text. Automated protocols for recognizing named entities, the building blocks of recipe text, are of immense value for various applications ranging from information extraction to novel recipe generation. Named entity recognition is a technique for extracting information from unstructured or semi-structured data with known labels. Starting with manually-annotated data of 6,611 ingredient phrases, we created an augmented dataset of 26,445 phrases cumulatively. Simultaneously, we systematically cleaned and analyzed ingredient phrases from RecipeDB, the gold-standard recipe data repository, and annotated them using the Stanford NER. Based on the analysis, we sampled a subset of 88,526 phrases using a clustering-based approach while preserving the diversity to create the machine-annotated dataset. A thorough investigation of NER approaches on these three datasets involving statistical, fine-tuning of deep learning-based language models and few-shot prompting on large language models (LLMs) provides deep insights. We conclude that few-shot prompting on LLMs has abysmal performance, whereas the fine-tuned spaCy-transformer emerges as the best model with macro-F1 scores of 95.9%, 96.04%, and 95.71% for the manually-annotated, augmented, and machine-annotated datasets, respectively.

Ayush Agarwal, Raghavendra Ramachandra, Sushma Venkatesh et al.

In the domain of Extended Reality (XR), particularly Virtual Reality (VR), extensive research has been devoted to harnessing this transformative technology in various real-world applications. However, a critical challenge that must be addressed before unleashing the full potential of XR in practical scenarios is to ensure robust security and safeguard user privacy. This paper presents a systematic survey of the utility of biometric characteristics applied in the XR environment. To this end, we present a comprehensive overview of the different types of biometric modalities used for authentication and representation of users in a virtual environment. We discuss different biometric vulnerability gateways in general XR systems for the first time in the literature along with taxonomy. A comprehensive discussion on generating and authenticating biometric-based photorealistic avatars in XR environments is presented with a stringent taxonomy. We also discuss the availability of different datasets that are widely employed in evaluating biometric authentication in XR environments together with performance evaluation metrics. Finally, we discuss the open challenges and potential future work that need to be addressed in the field of biometrics in XR.

Yunjiang Jiang, Ayush Agarwal, Yang Liu et al.

Scaling large recommendation systems requires advancing three major frontiers: processing longer user histories, expanding candidate sets, and increasing model capacity. While promising, transformers' computational cost scales quadratically with the user sequence length and linearly with the number of candidates. This trade-off makes it prohibitively expensive to expand candidate sets or increase sequence length at inference, despite the significant performance improvements. We introduce \textbf{LIME}, a novel architecture that resolves this trade-off. Through two key innovations, LIME fundamentally reduces computational complexity. First, low-rank ``link embeddings" enable pre-computation of attention weights by decoupling user and candidate interactions, making the inference cost nearly independent of candidate set size. Second, a linear attention mechanism, \textbf{LIME-XOR}, reduces the complexity with respect to user sequence length from quadratic ($O(N^2)$) to linear ($O(N)$). Experiments on public and industrial datasets show LIME achieves near-parity with state-of-the-art transformers but with a 10$\times$ inference speedup on large candidate sets or long sequence lengths. When tested on a major recommendation platform, LIME improved user engagement while maintaining minimal inference costs with respect to candidate set size and user history length, establishing a new paradigm for efficient and expressive recommendation systems.

Ashish Ranjan, Ayush Agarwal, Shalin Barot et al.

Temporal set prediction involves forecasting the elements that will appear in the next set, given a sequence of prior sets, each containing a variable number of elements. Existing methods often rely on intricate architectures with substantial computational overhead, which hampers their scalability. In this work, we introduce a novel and scalable framework that leverages permutation-equivariant and permutation-invariant transformations to efficiently model set dynamics. Our approach significantly reduces both training and inference time while maintaining competitive performance. Extensive experiments on multiple public benchmarks show that our method achieves results on par with or superior to state-of-the-art models across several evaluation metrics. These results underscore the effectiveness of our model in enabling efficient and scalable temporal set prediction.

Anatoly Shusterman, Ayush Agarwal, Sioli O'Connell et al.

The "eternal war in cache" has reached browsers, with multiple cache-based side-channel attacks and countermeasures being suggested. A common approach for countermeasures is to disable or restrict JavaScript features deemed essential for carrying out attacks. To assess the effectiveness of this approach, in this work we seek to identify those JavaScript features which are essential for carrying out a cache-based attack. We develop a sequence of attacks with progressively decreasing dependency on JavaScript features, culminating in the first browser-based side-channel attack which is constructed entirely from Cascading Style Sheets (CSS) and HTML, and works even when script execution is completely blocked. We then show that avoiding JavaScript features makes our techniques architecturally agnostic, resulting in microarchitectural website fingerprinting attacks that work across hardware platforms including Intel Core, AMD Ryzen, Samsung Exynos, and Apple M1 architectures. As a final contribution, we evaluate our techniques in hardened browser environments including the Tor browser, Deter-Fox (Cao el al., CCS 2017), and Chrome Zero (Schwartz et al., NDSS 2018). We confirm that none of these approaches completely defend against our attacks. We further argue that the protections of Chrome Zero need to be more comprehensively applied, and that the performance and user experience of Chrome Zero will be severely degraded if this approach is taken.

Navita Goyal, Roodram Paneri, Ayush Agarwal et al.

Content is created for a well-defined purpose, often described by a metric or signal represented in the form of structured information. The relationship between the goal (metrics) of target content and the content itself is non-trivial. While large-scale language models show promising text generation capabilities, guiding the generated text with external metrics is challenging. These metrics and content tend to have inherent relationships and not all of them may be of consequence. We introduce CaM-Gen: Causally aware Generative Networks guided by user-defined target metrics incorporating the causal relationships between the metric and content features. We leverage causal inference techniques to identify causally significant aspects of a text that lead to the target metric and then explicitly guide generative models towards these by a feedback mechanism. We propose this mechanism for variational autoencoder and Transformer-based generative models. The proposed models beat baselines in terms of the target metric control while maintaining fluency and language quality of the generated text. To the best of our knowledge, this is one of the early attempts at controlled generation incorporating a metric guide using causal inference.

Dhritimaan Das, Ayush Agarwal, Pratik Chattopadhyay et al.

Gait of a person refers to his/her walking pattern, and according to medical studies gait of every individual is unique. Over the past decade, several computer vision-based gait recognition approaches have been proposed in which walking information corresponding to a complete gait cycle has been used to construct gait features for person identification. These methods compute gait features with the inherent assumption that a complete gait cycle is always available. However, in most public places occlusion is an inevitable occurrence, and due to this, only a fraction of a gait cycle gets captured by the monitoring camera. Unavailability of complete gait cycle information drastically affects the accuracy of the extracted features, and till date, only a few occlusion handling strategies to gait recognition have been proposed. But none of these performs reliably and robustly in the presence of a single cycle with incomplete information, and because of this practical application of gait recognition is quite limited. In this work, we develop deep learning-based algorithm to accurately identify the affected frames as well as predict the missing frames to reconstruct a complete gait cycle. While occlusion detection has been carried out by employing a VGG-16 model, the model for frame reconstruction is based on Long-Short Term Memory network that has been trained to optimize a multi-objective function based on dice coefficient and cross-entropy loss. The effectiveness of the proposed occlusion reconstruction algorithm is evaluated by computing the accuracy of the popular Gait Energy Feature on the reconstructed sequence. Experimental evaluation on public data sets and comparative analysis with other occlusion handling methods verify the effectiveness of our approach.

Mansi Goel, Ayush Agarwal, Deepak Thukral et al.

This paper presents Fiducia, a food review system involving a pipeline which processes restaurant-related reviews obtained from Zomato (India's largest restaurant search and discovery service). Fiducia is specific to popular cafe food items and manages to identify relevant information pertaining to each item separately in the reviews. It uses a sentiment check on these pieces of text and accordingly suggests an appropriate restaurant for the particular item depending on user-item and item-item similarity. Experimental results show that the sentiment analyzer module of Fiducia achieves an accuracy of over 85% and our final recommender system achieves an RMSE of about 1.01 beating other baselines.

Le Hou, Ayush Agarwal, Dimitris Samaras et al.

Hematoxylin and Eosin stained histopathology image analysis is essential for the diagnosis and study of complicated diseases such as cancer. Existing state-of-the-art approaches demand extensive amount of supervised training data from trained pathologists. In this work we synthesize in an unsupervised manner, large histopathology image datasets, suitable for supervised training tasks. We propose a unified pipeline that: a) generates a set of initial synthetic histopathology images with paired information about the nuclei such as segmentation masks; b) refines the initial synthetic images through a Generative Adversarial Network (GAN) to reference styles; c) trains a task-specific CNN and boosts the performance of the task-specific CNN with on-the-fly generated adversarial examples. Our main contribution is that the synthetic images are not only realistic, but also representative (in reference styles) and relatively challenging for training task-specific CNNs. We test our method for nucleus segmentation using images from four cancer types. When no supervised data exists for a cancer type, our method without supervision cost significantly outperforms supervised methods which perform across-cancer generalization. Even when supervised data exists for all cancer types, our approach without supervision cost performs better than supervised methods.