Ashis Ghosh

Shilpa Mukhopadhyay, Sourav Ganguly, Santosh Mohan Rajkumar et al.

We study robust peak-cost constrained reinforcement learning (RP-CRL), where the objective is to maximize expected reward while controlling the maximum cost encountered along a trajectory. This setting is motivated by safety-critical applications in which a single large violation can be catastrophic and therefore cannot be adequately captured by the standard CMDP framework based on expected cumulative cost. Existing reachability-constrained RL methods adopt Lagrangian-based approaches, yet the underlying duality properties of peak-cost constrained MDPs remain unclear. We show that, unlike standard CMDPs, peak-cost constrained MDPs may not admit zero duality gap. We further consider a robust formulation to address simulator-to-real-world mismatch in the transition dynamics. To solve this problem, we develop a surrogate optimization framework and a robust value estimation method based on integral probability metrics. We prove that, with appropriate hyperparameter choices, the surrogate solution attains the same robust reward value as the original problem while violating the constraint by at most epsilon. Experiments show that the proposed method effectively enforces safety under dynamics perturbations while retaining strong reward performance.

Anamitra Ghosh, Abhiroop Chatterjee, Susmita Ghosh

This article presents DMFNet, a dual-backbone multiscale feature fusion framework with residual feature propagation and spatial attention for remote sensing scene classification. Existing approaches often face challenges in effectively capturing multiscale feature interactions and learning robust feature representations from complex aerial scenes with high intra-class variability and inter-class similarity. To address these limitations, the proposed framework employs two pretrained backbone networks to extract diverse hierarchical feature representations. A multiscale feature fusion mechanism with residual feature propagation is introduced to enhance feature interaction across multiple resolution levels. In addition, a spatial attention module is introduced to emphasize informative spatial regions in multi-object scenes. Further, a two-stage training strategy consisting of backbone freezing followed by selective fine-tuning is adopted to ensure stable optimization and improved generalization. Experiments conducted on the benchmark AID dataset demonstrate that the DMFNet achieves an average accuracy of 97.46\% $\pm$ 0.14\%. Ablative analysis further show the importance of various components in unison.

Jean-Daniel Boissonnat, Hana Dal Poz Kourimska, Arijit Ghosh et al.

In 1942, Freudenthal showed that a simplex in Euclidean space can be subdivided such that the quality (well-shapedness of the simplex, quantified in terms of e.g. fatness) of the simplices in the subdivision is lower bounded. This answered a question of Brouwer. Recently, Brunck discussed the same problem for simplices in two-dimensional spaces of constant curvature and provided a closely related construction. In this paper we generalize Brunck's result to arbitrary dimensional spaces of constant curvature by combining Freudenthal's construction and radial projection. We contrast this approach with Brunck's construction.

Shayne Longpre, Elaine Zhu, Carson Ezell et al.

Flaw reporting for deployed AI systems is fundamental to identifying system failures and improving AI safety. Yet the AI reporting ecosystem is fragmented: researchers who identify flaws often do not know what or where to report, and groups who receive reports rarely share them with other relevant stakeholders. As a result, good-faith reporters duplicate effort by submitting many different forms, and recipients lack standardized, triage-ready information. We audit 12 reporting systems published by AI developers, cybersecurity groups, and AI flaw aggregators, identifying five recurring design challenges spanning discoverability, scope, information collection, coordination, and guidance for strict-liability cases. Building on this analysis and feedback from 49 experts across 32 organizations representing developers, security researchers, and ecosystem coordinators, we introduce FLARE-AI, an open-source AI flaw reporting system designed for interoperability with existing systems. FLARE-AI streamlines flaw report creation by collecting triage-relevant information through conditional logic and early classification, then enables optional dissemination of standardized, machine-readable reports to multiple developers, coordinators, and incident registries from a single submission. By lowering barriers to reporting AI flaws and improving interoperability across stakeholders, FLARE-AI helps break down silos and accelerate remediation across the AI ecosystem.

Debarshi Chanda, Buddha Dev Das, Arijit Ghosh et al.

We study the problem of estimating the number of edges in an undirected, unweighted graph using sublinear query access. We consider a query model that preserves the structure of Independent Set (IS) queries, but augments their output with a random certificate: given a vertex subset, the oracle returns a uniformly random edge from the induced subgraph if one exists, and returns null otherwise. Using this access, we give a randomized algorithm that outputs a $(1 \pm \varepsilon)$-approximation to the number of edges with constant success probability using $\widetilde{O}(\log^{2} m)$ queries. This implies an exponential separation from both standard IS queries and global random edge-sampling models: estimating the number of edges using standard IS queries require $\widetildeΘ\!\left(\min\left\{\sqrt{m},\, \frac{n}{\sqrt{m}}\right\}\right)$ queries, while direct random edge-sample access requires $\widetildeΘ(\sqrt{m})$ samples. Beyond separation in query complexity, our algorithm is output-sensitive: its query complexity is polylogarithmic in the number of edges in the graph. This aligns with the classical objective in group testing, where one seeks algorithms that are both worst-case optimal and instance-adaptive. Conceptually, our model connects group testing, the decision-versus-counting dichotomy, graph property testing, and the "power of a random certificate", and can be viewed as a structured form of conditional sampling of edges in graphs.

Victor-Alexandru Pădurean, Kaitlin Riegel, Alkis Gotovos et al.

As Generative AI (GenAI) becomes increasingly central to software development, CS education is integrating prompt-centered workflows where students describe intended program behavior in natural language to elicit code. However, professional practice requires careful review and verification of GenAI-generated code that may appear correct while containing subtle faults. This creates a challenge for CS1-level activities, where current models often solve tasks correctly and reduce students' incentive to closely inspect generated outputs. We investigate how prompt-centered programming activities can be adapted to better foster these practices. Specifically, we explore an approach where realistic, runnable bugs are injected into otherwise correct solutions, thus requiring students to read and repair generated outputs. We analyzed 2,636 sessions from 917 students, and examined behavior across instances of naturally occurring prompt-related failures and deliberately injected bugs within each session. Our findings show that students responded differently across bug sources. Deliberately injected bugs more often led to direct code edits and higher next-attempt success, suggesting localized repair of near-miss solutions. Prompt-related failures instead more often led students to refine prompts by clarifying constraints, updating function signatures, adding edge cases, or reframing the task. Student reflections reinforce the emphasis on review and repair, describing useful practice in code understanding, code review, and debugging, as well as a more careful verification mindset and greater awareness of GenAI limitations. Ultimately, prompt-related failures and injected bugs together support a pedagogically useful GenAI workflow, where students practice both specification refinement through prompts and debugging through code editing.

Shen Zheng, Anurag Ghosh, Gaurav Parmar et al.

Image-to-image (I2I) translation has achieved strong results in tasks like human relighting and driving scene translation using latent diffusion models (LDMs). However, compact LDMs often struggle to preserve fine-grained structures because the encoder compresses high-resolution inputs into a spatially downsampled latent space. To address this issue, we propose a simple saliency-guided warp-unwarp framework that reallocates spatial representation toward salient regions before encoding, enabling better preservation of structural details without increasing latent resolution. The warped image is processed by the original diffusion model and then mapped back via an inverse warp. In addition, we propose a simple and efficient outpainting-based synthetic data generation pipeline to produce high-quality paired data for image relighting. Our method is model-agnostic, requires no architectural modification, and introduces negligible computational overhead. Experiments on human relighting, driving scene relighting, and translation demonstrate improved structural preservation, lighting faithfulness, and image quality, with our framework extending naturally to video via frame-by-frame application with good temporal stability. Project Webpage: https://shenzheng2000.github.io/WarpI2I.github.io

Matteo Farina, Vishaal Udandarao, Thao Nguyen et al.

Building performant Vision-Language Models (VLMs) requires carefully curating large-scale training datasets, yet the community lacks systematic benchmarks for evaluating such curation strategies. We introduce DataComp for VLMs (DCVLM), a benchmark for controlled data-centric experiments to improve VLM training. As part of DCVLM, we collect 160 datasets spanning four data types -- image-caption pairs, multimodal interleaved documents, text-only, and instruction-tuning data -- into a corpus of 6T multimodal tokens. DCVLM allows participants to test curation strategies (filtering, mixing, formatting, sampling) across 1B-8B models and 6.25B-200B token budgets. Models are then evaluated on a carefully selected suite of up to 52 downstream benchmarks across 9 domains. We conduct extensive experiments on DCVLM and find that data mixing, not filtering, is key to a high-quality training dataset: instruction-heavy mixtures scale better than caption-heavy ones, with gains widening at larger scales. The resulting dataset, DCVLM-Baseline, enables training an 8B VLM to 63.6% accuracy on our 33-task core suite with 200B training tokens. Compared to FineVision, the state-of-the-art open VLM training dataset, this represents an improvement of +5.4pp. DCVLM and all accompanying artifacts will be made publicly available at https://www.datacomp.ai/dcvlm/.

Ananya Kunisetty, Avishek Ghosh

We study a multi-agent multi-armed bandit problem in the competitive setup with two-sided matching markets under a human centric decision making model. To capture human preferences, we use cumulative prospect theory (CPT) that weighs the actions of the agent in a nonlinear fashion using a ($α$-Hölder continuous) weight function. CPT has been widely used in behavioral economics and risk sensitive machine learning to emulate human preferences. We analyze the state-of-the-art learning algorithm with CPT weight distorted rewards and obtain a player optimal regret of $\mathcal{O}(K\log T \left(\frac{1}Δ\right)^{2/α})$, where $K$ denotes the number of arms, $T$ is the learning horizon, and $Δ$ represents (suitably defined) players' minimum preference gap. Noticing the dependence on $Δ$ to be sub-optimal, we further improve this regret by judiciously selecting the active set of arms during exploration, which removes the dependence on $K$ in the dominant term and achieves an improved (optimal) regret guarantees in the setting where the number of arms $K$ is significantly larger than the number of players $N$. In addition, we consider adversarial markets where the observed rewards of the agents may be corrupted. We propose and analyze algorithms for robust markets with CPT as risk sensitive measure in both settings where the total corruption budget is known and where it is unknown, and establish logarithmic player-optimal regret guarantees in both cases.

Wojciech Czerwiński, Daniel Dadush, Ekin Ergen et al.

In online vector balancing, vectors $t_1,\dots,t_n$ arrive one by one from a given set $T$ and the goal is to assign signs $s_1,\dots,s_n\in\{\pm1\}$ in an online manner so as to minimize the largest norm of any signed prefix sum $\sum_{i=1}^ks_i t_i$, $k \in [n]$. In this paper, we analyze the natural Euclidean greedy vector balancing algorithm for this problem: at each step $k$, the sign $s_k\in\{\pm1\}$ is chosen so that $s_k t_k$ has non-positive inner product with $\sum_{i=1}^{k-1} s_i\cdot t_i$. Our main result is the first finite bound, independent of the sequence length $n$, on the performance of greedy whenever $T$ is finite. When $T \subset \mathbb{R}^d$ consists of unit vectors, we prove that the signed sums produced by greedy have Euclidean norm at most $(2/δ_T)^{d-1}$, where $δ_T$ is the minimum non-zero distance between vectors in $T$ and subspaces spanned by vectors in $T$. The same upper bound holds when the sequences are composed of scaled down vectors in $T$. We also provide a simple set $T$ for which $Ω(\sqrt{d}/δ_T)$ is a lower bound. We analyze the greedy algorithm by proving the existence of a bounded convex $K_T$ that is $T$-absorbing: $\forall x\in K_T$ and $t \in\pm T$, $\langle x,t\rangle\leq0\Rightarrow x+t\in K_T$. We give an explicit construction of a set $K_T$ contained in a ball of radius $(2/δ_T)^{d-1}$, based on chains of subspaces spanned by vectors in $T$, which may be of independent interest. We generalize our greedy vector balancing bound to online vector partitioning, where the sequence $t_1,\dots,t_n$ must be partitioned in an online manner into $p$ subsequences. As an application, we prove a special case of a conjecture of Bosman et al. (arxiv:2402.19259), showing that a lexicographic version of total completion time scheduling under scenarios is polynomial time solvable when the number of scenarios is fixed.

Jan Batzner, Sree Harsha Nelaturu, Anastassia Kornilova et al.

AI evaluations are widely used for testing and understanding progress. However, the diverse evaluators bring with them inconsistencies that challenge analysis and comparison. First, results are saved in incompatible formats, scattered across leaderboards, papers, blog posts, evaluation harness logs, and custom repositories. Second, results are created by different evaluation frameworks, which produce divergent scores for nominally identical evaluations and record metadata inconsistently, hindering comparison, cross-community evaluation science, cost reduction, and reuse. We introduce Every Eval Ever, the first shared schema and community-crowdsourced repository for AI evaluation results. The schema standardizes how evaluations are represented in a unified, single JSON document. It is source-agnostic by design, ingesting results from evaluation harnesses and papers alike, and optionally stores per-instance outputs for fine-grained analysis. We contribute: (i) a community-governed metadata schema with a companion instance-level schema, the first standardization effort of its kind; (ii) automatic converters from popular formats, evaluation harnesses, and leaderboards to the unified schema; and (iii) a crowdsourced community database hosted on Hugging Face, currently spanning to date 22,235 models, 2,273 unique benchmarks, and 31 evaluation formats.

Tanmoy Kanti Halder, Akash Ghosh, Subhadip Baidya et al.

Multimodal Large Language Models (MLLMs) have shown promising reasoning capabilities in general domains, yet their performance remains limited in specialized settings such as healthcare, especially in multilingual and low-resource scenarios. This gap is critical in regions like rural India, where patients often express complex medical queries in native Indic languages and rely on multimodal inputs such as medical images. Existing English-centric MLLMs struggle to support such use cases, limiting equitable access to AI-driven healthcare assistance. To address this challenge, we introduce ArogyaBodha, a large-scale multilingual multimodal medical question-answer dataset constructed from eight heterogeneous sources, covering 31 body systems, six imaging modalities, and 21 clinical domains across English and seven major Indian languages. We further propose ArogyaSutra, an actor-critic-based multi-agent framework that integrates tool grounding with dual-memory mechanisms for step-wise, reasoning-aware decision making, and uses stored actor-critic simulation trajectories for distillation. Experiments show that our dataset and framework improve multilingual medical reasoning accuracy across all Indic languages, with ablations validating the contribution of each component. The source code and dataset are available at: https://iitp-cse.github.io/ ArogyaSutra/

Addis Fuhr, Zachary R. Fox, David Parker et al.

Two-dimensional magnets offer compelling platforms for spintronics and quantum technologies, yet predicting their magnetic ground states, moments, and anisotropy remains challenging. This limitation primarily arises because existing machine-learning representations encode chemical environments without capturing the symmetry or exchange physics that govern magnetism. In this work, we introduce the symmetry-electronic fingerprint (SEF), a physically interpretable representation that encodes crystallographic symmetry operations, Wyckoff-site geometry, together with site-resolved electronic structure. Combined with ensemble learning with random forests, the SEF accurately classifies magnetic ordering while regressing moments alongside anisotropy energies while simultaneously resolving the distinct regimes of itinerant Stoner ferromagnetism from localized superexchange. What sets the SEF-trained models apart is that regions of elevated model uncertainty are not a failure but a diagnostic, identifying materials where these mechanisms compete. First-principles calculations on Co- and Ni-based halides and oxides confirm that these regions correspond to genuine near-degenerate FM and AFM phases with magnetic frustration, suppressed anisotropy, and emergent non-collinear ordering. By encoding symmetry together with exchange physics directly into the representation unlike conventional descriptors, the SEF transforms model uncertainty into a compass pointing toward two-dimensional materials where small perturbations drive transitions between collinear, frustrated, or non-collinear magnetic phases.

Anisha Saha, Vaibhav Rathore, Abhisek Tiwari et al.

The growing adoption of AI in healthcare, particularly in preventive care, highlights the critical need for accessibility and precision in Medical Question Answering (MedQA). In recent years, significant efforts have been made to develop multi-span medical question-answering systems, where the answer to a query may span multiple sections or paragraphs of a source document. However, existing systems fall short of aligning with real-world scenarios, where source documents often include both textual and visual content, requiring answers to incorporate images for better comprehension. To address this gap, we propose $M^3QAFrame$, a multi-modal, multi-span medical question-answering framework that leverages visual cues to enhance the generation of comprehensive answers drawn from diverse textual and visual spans. The model takes the context, query, and images as input and outputs an answer containing both textual answers and relevant images. The text and image embeddings are processed using a transformer-based architecture to determine the sentence and image relevance. We curate a multi-modal, multi-span medical question-answering ($M^3 QuestionIng$) dataset containing queries, medical contexts, associated medical images, and extractive answers. Additionally, each query-answer pair is labeled with user intent and query type to enhance query and context comprehension. Extensive experiments show that our approach consistently outperforms existing methods across various evaluation metrics.