Arvind Srinivasan

Arvind Srinivasan, Niklas Elmqvist

Where people look during shared activity carries coordination cues that speech and gesture cannot replace, but these patterns remain invisible to participants. XR headsets make gaze available as real-time input, yet few systems feed it back visually. We frame our work using the Attention-Aware Pipeline (Capture, Record, Revisualize), whose feedback loop means the systems visual response alters what users attend to next, triggering further responses. This generates design tensions whose form depends on each stages configuration. We trace the pipeline through three systems casting attention as a mirror (reflecting gaze history), a medium (sharing it across collaborators), and a mediator (intervening through diminished reality). Each encountered a tension the loop predicted, motivating the next. A formative eye-tracking study of four musicians surfaced attentional tunneling and near-total disconnection, confirming the need for intervention. We present these tensions and a next step: testing whether subtractive intervention reduces tunneling for a single sight-reader.

Arvind Srinivasan, Tobias Rau, Michael Sedlmair

Visual attention is central to ensemble coordination, yet how musicians allocate gaze during naturalistic rehearsal remains poorly understood. We present a pilot study using mobile eye tracking to examine gaze behaviour in a four-member band across three songs, each practiced twice. Musicians wore Pupil Labs Neon eye trackers, and YOLOv8-assisted scene annotations mapped fixations to ensemble members and objects in view. Analyzing fixation matrices, transition matrices, temporal scarf plots, and dwell-transition correlations, we uncover a hub-and-spoke attention topology: the session leader was the dominant gaze target for all members, while the learning guitarist concentrated up to 97% of interpersonal dwell on this single reference. Between attempts, gaze transitions decreased by up to 65% on average for unfamiliar material (up to 82% for individual participants) as scanning stabilized. Scarf plots reveal how teaching breakdowns fragment attention and uninterrupted runs consolidate it. Post-session participant reflections align with the quantitative patterns, and we discuss implications for gaze-aware tools in ensemble pedagogy.

Arvind Srinivasan, Niklas Elmqvist

Group awareness--the ability to perceive the activities of collaborators in a shared space--is a vital mechanism to support effective coordination and joint data analysis in collaborative visualization. We introduce collaborative attention-aware visualizations (CAAVs) that track, record, and revisualize the collective attention of multiple users over time. We implement this concept in HeedVision, a standards-compliant WebXR system built with React Three Fiber that runs on modern AR/VR headsets, and complement it with proof-of-concept implementations covering the remaining three quadrants of our design space--varying presentation (embedded vs. separated) and situatedness (world space vs. camera space). Through a mixed-methods exploratory study where pairs of co-located analysts performed visual search tasks in a shared immersive AR environment, we investigate how attention revisualization affects collaborative coordination in immersive analytics. Our results show that CAAVs improve spatial coordination, search efficiency, and task load distribution among collaborators, though benefits vary by context, favoring abstract environments lacking natural landmarks. This work extends attention awareness to multi-user settings and provides empirical evidence for its context-dependent benefits in collaborative immersive analytics environments.

Mononito Goswami, Vedant Sanil, Arjun Choudhry et al. · cmu

Machine learning (ML) models are only as good as the data they are trained on. But recent studies have found datasets widely used to train and evaluate ML models, e.g. ImageNet, to have pervasive labeling errors. Erroneous labels on the train set hurt ML models' ability to generalize, and they impact evaluation and model selection using the test set. Consequently, learning in the presence of labeling errors is an active area of research, yet this field lacks a comprehensive benchmark to evaluate these methods. Most of these methods are evaluated on a few computer vision datasets with significant variance in the experimental protocols. With such a large pool of methods and inconsistent evaluation, it is also unclear how ML practitioners can choose the right models to assess label quality in their data. To this end, we propose a benchmarking environment AQuA to rigorously evaluate methods that enable machine learning in the presence of label noise. We also introduce a design space to delineate concrete design choices of label error detection models. We hope that our proposed design space and benchmark enable practitioners to choose the right tools to improve their label quality and that our benchmark enables objective and rigorous evaluation of machine learning tools facing mislabeled data.

Lingxiao Zhao, Saurabh Sawlani, Arvind Srinivasan et al.

Graph-based anomaly detection finds numerous applications in the real-world. Thus, there exists extensive literature on the topic that has recently shifted toward deep detection models due to advances in deep learning and graph neural networks (GNNs). A vast majority of prior work focuses on detecting node/edge/subgraph anomalies within a single graph, with much less work on graph-level anomaly detection in a graph database. This work aims to fill two gaps in the literature: We (1) design GLAM, an end-to-end graph-level anomaly detection model based on GNNs, and (2) focus on unsupervised model selection, which is notoriously hard due to lack of any labels, yet especially critical for deep NN based models with a long list of hyper-parameters. Further, we propose a new pooling strategy for graph-level embedding, called MMD-pooling, that is geared toward detecting distribution anomalies which has not been considered before. Through extensive experiments on 15 real-world datasets, we show that (i) GLAM outperforms node-level and two-stage (i.e. not end-to-end) baselines, and (ii) model selection picks a significantly more effective model than expectation (i.e. average) -- without using any labels -- among candidates with otherwise large variation in performance.

Zijian Ding, Arvind Srinivasan, Stephen MacNeil et al.

Cross-domain analogical reasoning is a core creative ability that can be challenging for humans. Recent work has shown some proofs-of concept of Large language Models' (LLMs) ability to generate cross-domain analogies. However, the reliability and potential usefulness of this capacity for augmenting human creative work has received little systematic exploration. In this paper, we systematically explore LLMs capacity to augment cross-domain analogical reasoning. Across three studies, we found: 1) LLM-generated cross-domain analogies were frequently judged as helpful in the context of a problem reformulation task (median 4 out of 5 helpfulness rating), and frequently (~80% of cases) led to observable changes in problem formulations, and 2) there was an upper bound of 25% of outputs bring rated as potentially harmful, with a majority due to potentially upsetting content, rather than biased or toxic content. These results demonstrate the potential utility -- and risks -- of LLMs for augmenting cross-domain analogical creativity.

Junze Liu, Kun Qian, Florian Dubost et al.

Vision-language models (VLMs) exhibit a striking paradox: they can generate executable code that reconstructs a 3D scene from geometric primitives with correct object counts, classes, and approximate positions, yet the same models fail at simpler spatial questions on the same image. We show that 3D geometric primitives (cubes, spheres, cylinders, expressed in executable code) serve as a powerful intermediate representation for spatial understanding, and exploit this through three contributions. First, we introduce \textbf{\textsc{SpatialBabel}}, a benchmark evaluating fourteen VLMs on primitive-based 3D scene reconstruction across six \emph{scene-code languages} (programming languages and declarative formats for 3D primitive scenes), revealing that a single model's object-detection F1 can vary by up to $5.7\times$ across languages. Second, we propose \textbf{Code-CoT} (Code Chain-of-Thought), a training-free inference strategy that routes spatial reasoning through primitive-based code generation. Code-CoT lifts the SpatialBabel-QA-Score by up to $+6.4$\% on primitive scenes and real-photo CV-Bench-3D accuracy by $+5.0$\% for VLMs with strong coding capabilities. Third, we propose \textbf{S$^{3}$-FT} (Self-Supervised Spatial Fine-Tuning), which self-supervisedly distills primitive spatial knowledge into general visual reasoning by parsing the model's own Three.js primitive-reconstructions into structured annotations and fine-tuning on the result, with \emph{no human labels and no teacher model}. Training on primitive images alone, S$^3$-FT improves Qwen3-VL-8B by $+4.6$ to $+8.6$\% on SpatialBabel-Primitive-QA, $+9.7$\% on CV-Bench-2D, and $+17$\% on HallusionBench; the recipe transfers across model families. These results establish geometric primitives in code as both a diagnostic and a transferable spatial vocabulary for VLMs. We will release all artifacts upon publication.

Jiacheng Lin, Kun Qian, Arvind Srinivasan et al.

Personalization today is fundamentally platform-centric: services build user representations from the behavioral fragments they observe. Yet no platform can construct a complete picture of the user, as competitive incentives, legal constraints, user privacy concerns, and epistemic limits create persistent data barriers. This paper argues for a shift from platform-centric personalization to user-governed personalization, where only the user can integrate fragmented contexts across platforms and the offline world. The key asymmetry lies in data access: only users can aggregate their own cross-platform and offline information. Large language model (LLM) agents make such integration practically feasible for the first time by enabling reasoning over heterogeneous personal data and transforming users' cross-context information into actionable personalization capabilities. We provide proof-of-concept evidence that users equipped with cross-platform data exports and an off-the-shelf LLM agent can outperform single-platform personalization baselines. We conclude by outlining a research agenda for building scalable user-governed personalization systems.

Jiacheng Lin, Zhongruo Wang, Kun Qian et al.

Supervised Fine-Tuning (SFT) on domain-specific datasets is a common approach to adapt Large Language Models (LLMs) to specialized tasks but is often believed to degrade their general capabilities. In this work, we revisit this trade-off and present both empirical and theoretical insights. First, we show that SFT does not always hurt: using a smaller learning rate can substantially mitigate general performance degradation while preserving comparable target-domain performance. We then provide a theoretical analysis that explains these phenomena and further motivates a new method, Token-Adaptive Loss Reweighting (TALR). Building on this, and recognizing that smaller learning rates alone do not fully eliminate general-performance degradation in all cases, we evaluate a range of strategies for reducing general capability loss, including L2 regularization, LoRA, model averaging, FLOW, and our proposed TALR. Experimental results demonstrate that while no method completely eliminates the trade-off, TALR consistently outperforms these baselines in balancing domain-specific gains and general capabilities. Finally, we distill our findings into practical guidelines for adapting LLMs to new domains: (i) using a small learning rate to achieve a favorable trade-off, and (ii) when a stronger balance is further desired, adopt TALR as an effective strategy.

Arvind Srinivasan, Chien-Chung Chan

We present a new secret sharing algorithm that provides the storage efficiency of an Information Dispersal Algorithm (IDA) while providing perfect secret sharing. We achieve this by mixing the input message with random bytes generated using Repeatable Random Sequence Generator (RRSG). We also use the data from the RRSG to provide random polynomial evaluation points and optionally compute the polynomials on random isomorphic fields rather than a single fixed field.

Arvind Srinivasan, Aprameya Bharadwaj, Aveek Saha et al.

Large scale video retrieval is a field of study with a lot of ongoing research. Most of the work in the field is on video retrieval through text queries using techniques such as VSE++. However, there is little research done on video retrieval through image queries, and the work that has been done in this field either uses image queries from within the video dataset or iterates through videos frame by frame. These approaches are not generalized for queries from outside the dataset and do not scale well for large video datasets. To overcome these issues, we propose a new approach for video retrieval through image queries where an undirected graph is constructed from the combined set of frames from all videos to be searched. The node features of this graph are used in the task of video retrieval. Experimentation is done on the MSR-VTT dataset by using query images from outside the dataset. To evaluate this novel approach P@5, P@10 and P@20 metrics are calculated. Two different ResNet models namely, ResNet-152 and ResNet-50 are used in this study.

Arvind Srinivasan, Aprameya Bharadwaj, Manasa Sathyan et al.

In this paper we improve the image embeddings generated in the graph neural network solution for few shot learning. We propose alternate architectures for existing networks such as Inception-Net, U-Net, Attention U-Net, and Squeeze-Net to generate embeddings and increase the accuracy of the models. We improve the quality of embeddings created at the cost of the time taken to generate them. The proposed implementations outperform the existing state of the art methods for 1-shot and 5-shot learning on the Omniglot dataset. The experiments involved a testing set and training set which had no common classes between them. The results for 5-way and 10-way/20-way tests have been tabulated.