arXiv subjects

Ranveer Chandra

Publications and source records attributed to Ranveer Chandra.

At least 19 recordsLinked to original sources

arXiv2026

Personalization as Inverse Planning: Learning Latent Design Intents for Agentic Slide Generation via Structural Denoising

Slide design requires personalizing both deck themes and page layouts. Yet, current AI agent-based methods struggle with fine-grained, page-level design. Solely relying on prespecified templates or user verbose instructions, they fail to capture latent design intents, leaving Page-level Slide Personalization (PSP) unresolved. To close this gap, this work formulates PSP as an inverse planning problem. We propose to learn a design intent without assuming any knowledge of the specific executing tools (e.g., PowerPoint, Beamer) being used. However, relinquishing control over these tools makes the problem intractable to optimize end-to-end. To overcome this, we propose SPIRE, a principled framework to solve PSP approximately. By intentionally corrupting the visual structures of clean slides, SPIRE creates a verifiable task to denoise the corruption, whereby two agents learn to collaboratively refine executable designs via reinforcement learning (RL). We present a proof that structural denoising is a consistent surrogate for PSP, and that the multi-agent formulation strictly reduces policy gradient variance in RL. Extensive experiments demonstrate the superiority of SPIRE.

cs.AI↗

arXiv2026

GRIP: Feedback-Guided Prompt Retrieval for Large Multimodal Models

In-Context Learning (ICL) has become a powerful mechanism for adapting Large Language Models (LLMs) to new tasks without fine-tuning. Extending this concept to Large Multimodal Models (LMMs), Multimodal In-Context Learning (M-ICL) relies on retrieving relevant examples, such as images, captions, or question-answer pairs, to guide predictions across tasks like classification, captioning, and visual question answering (VQA). Most existing approaches select in-context examples based on feature-space similarity, assuming that semantically similar samples provide the most useful context. However, our systematic analysis reveals that this assumption does not always hold: visually similar examples are not necessarily those that most effectively enhance in-context learning performance. To address this, we propose the Guided Retrieval of In-context Prompts (GRIP), a learnable vision-only retrieval framework that leverages feedback from LMMs to identify examples that truly improve model predictions. GRIP learns to distinguish beneficial from detrimental in-context examples through contrastive training, refining retrieval beyond pure similarity. Across three multimodal tasks, namely classification, captioning, and VQA, GRIP improves consistently over similarity-based retrieval on Qwen2.5-VL-7B, with its strongest gains in classification on Idefics2-8B. Moreover, we demonstrate that retrievers trained with feedback from one open LMM can be transferred to other models without retraining, including closed-source GPT-4o and Gemini, enabling scalable and cost-efficient deployment of M-ICL. Code will be published upon acceptance.

cs.CV↗

arXiv2026

Diagnosing Capability Gaps in Fine-Tuning Data

Fine-tuning large language models (LLMs) for domain-specific tasks requires training datasets that comprehensively cover the target capabilities a practitioner needs. Yet identifying which capabilities a dataset fails to support, and doing so before an expensive fine-tuning run, remains a largely unsolved problem. We introduce GoalCover, a framework that helps practitioners systematically detect capability gaps in fine-tuning datasets through interactive goal decomposition and automated coverage assessment. GoalCover guides a practitioner through structured decomposition of a high-level goal into atomic, independently evaluable subgoals; assigns each training sample an LLM-based alignment score against every subgoal; and surfaces missing capabilities through automated analysis of low-scoring sample explanations. We validate the framework along two complementary axes. First, through controlled corruption experiments across three domains (medical QA, legal summarization, code generation), we show that GoalCover reliably distinguishes targeted from non-targeted capability impacts: target subgoals degrade by 25.6% on average versus 2.1% for non-target subgoals (Cohen's d=1.24). Second, we demonstrate downstream utility on a financial-summarization Reinforcement Fine-Tuning (RFT) task with Qwen-3-14B: training on GoalCover-filtered data improves the LLM-judge reward from 3.77 to 4.12 (out of 5) over the unfiltered baseline, and combining filtered data with goal-conditioned synthetic samples yields the strongest result (4.20). The two results together show that GoalCover works as a practical pre-fine-tuning diagnostic: it detects capability gaps and produces concrete signal for closing them.

cs.LG↗

arXiv2026

SibylSense: Adaptive Rubric Learning via Memory Tuning and Adversarial Probing

Designing aligned and robust rewards for open-ended generation remains a key barrier to RL post-training. Rubrics provide structured, interpretable supervision, but scaling rubric construction is difficult: expert rubrics are costly, prompted rubrics are often superficial or inconsistent, and fixed-pool discriminative rubrics can saturate and drift, enabling reward hacking. We present SibylSense, an inference-time learning approach that adapts a frozen rubric generator through a tunable memory bank of validated rubric items. Memory is updated via verifier-based item rewards measured by reference-candidate answer discriminative gaps from a handful of examples. SibylSense alternates memory tuning with a rubric-adversarial policy update that produces rubric-satisfying candidate answers, shrinking discriminative gaps and driving the rubric generator to capture new quality dimensions. Experiments on two open-ended tasks show that SibylSense yields more discriminative rubrics and improves downstream RL performance over static and non-adaptive baselines.

cs.CL↗

arXiv2025

Enterprise AI Must Enforce Participant-Aware Access Control

Large language models (LLMs) are increasingly deployed in enterprise settings where they interact with multiple users and are trained or fine-tuned on sensitive internal data. While fine-tuning enhances performance by internalizing domain knowledge, it also introduces a critical security risk: leakage of confidential training data to unauthorized users. These risks are exacerbated when LLMs are combined with Retrieval-Augmented Generation (RAG) pipelines that dynamically fetch contextual documents at inference time. We demonstrate data exfiltration attacks on AI assistants where adversaries can exploit current fine-tuning and RAG architectures to leak sensitive information by leveraging the lack of access control enforcement. We show that existing defenses, including prompt sanitization, output filtering, system isolation, and training-level privacy mechanisms, are fundamentally probabilistic and fail to offer robust protection against such attacks. We take the position that only a deterministic and rigorous enforcement of fine-grained access control during both fine-tuning and RAG-based inference can reliably prevent the leakage of sensitive data to unauthorized recipients. We introduce a framework centered on the principle that any content used in training, retrieval, or generation by an LLM is explicitly authorized for \emph{all users involved in the interaction}. Our approach offers a simple yet powerful paradigm shift for building secure multi-user LLM systems that are grounded in classical access control but adapted to the unique challenges of modern AI workflows. Our solution has been deployed in Microsoft Copilot Tuning, a product offering that enables organizations to fine-tune models using their own enterprise-specific data.

cs.CR↗

arXiv2025

RLTHF: Targeted Human Feedback for LLM Alignment

Fine-tuning large language models (LLMs) to align with user preferences is challenging due to the high cost of quality human annotations in Reinforcement Learning from Human Feedback (RLHF) and the generalizability limitations of AI Feedback. To address these challenges, we propose RLTHF, a human-AI hybrid framework that combines LLM-based initial alignment with selective human annotations to achieve full-human annotation alignment with minimal effort. RLTHF identifies hard-to-annotate samples mislabeled by LLMs using a reward model's reward distribution and iteratively enhances alignment by integrating strategic human corrections while leveraging LLM's correctly labeled samples. Evaluations on HH-RLHF and TL;DR datasets show that RLTHF reaches full-human annotation-level alignment with only 6-7% of the human annotation effort. Furthermore, models trained on RLTHF's curated datasets for downstream tasks outperform those trained on fully human-annotated datasets, underscoring the effectiveness of RLTHF.

cs.CL↗

arXiv2025

Orchestration for Domain-specific Edge-Cloud Language Models

The remarkable performance of Large Language Models (LLMs) has inspired many applications, which often necessitate edge-cloud collaboration due to connectivity, privacy, and cost considerations. Traditional methods primarily focus on selecting the best LLM model for optimizing performance, while neglecting the critical interplay between the components of the LLM serving pipeline (context retrieval, query preprocessing, etc.) or the changing latency and cost constraints. We introduce ECO-LLM (Edge-Cloud Orchestrator for LLMs), a novel system that reframes this problem as a joint optimization challenge and solves it by systematically exploring component configurations and dynamically selecting optimal strategies at the query level. ECO-LLM consists of two components: (1) the ECO-LLM Emulator, which efficiently explores the vast configuration space utilizing query clustering and pareto-optimal path selection, gathering domain-specific performance metrics without exhaustive evaluation; and (2) the ECO-LLM Runtime, which leverages these metrics to dynamically select optimal resolution strategies for user queries while meeting user-defined Service Level Objectives (SLOs). We evaluate ECO-LLM on a smart home and a smart car assistant scenarios. With an exhaustive exploration of all possible configurations for seen queries, ECO-LLM outperforms cloud-based models like GPT-4o in terms of accuracy (90% vs. 74% on average) while reducing costs by 90% and latency by 55%, demonstrating the value of its joint optimization at the query level. In practical deployment for previously unseen queries, ECO-LLM selects configurations that reduce costs by 62% or improve response times by 62% on average compared to state-of-the-art model routing approaches, while maintaining higher accuracy and consistently adhering to specified latency and cost constraints.

cs.DB↗

arXiv2025

Direct Reasoning Optimization: Token-Level Reasoning Reflectivity Meets Rubric Gates for Unverifiable Tasks

Reinforcement learning (RL) training of large language models (LLMs) on unverifiable tasks is challenging even when a reasonable-quality reference answer is available. We propose a constrained RL training framework that (i) optimizes a token-level dense Reasoning Reflection Reward (R3) aligned with reasoning quality, and (ii) enforces rubric-gating as feasibility constraints at the rollout group level. R3 measures the model's token-level certainty of a reference answer under its chain-of-thought (CoT) prefix, and selectively emphasizes tokens with high cross-rollout variance, which we call reasoning-reflective tokens, that would otherwise be diluted by the bulk of low-variance tokens. The same variance signal also drives a filter that discards queries with insufficient signal for comparative learning. Rubric-gating complements R3 by operationalizing principled task criteria as hard accept/reject checks on final answers. Empirically, across four datasets spanning scientific writing, medicine, legal contracts, and finance, our framework outperforms strong baselines, achieves faster, more sample-efficient learning, and respects feasibility constraints.

cs.CL↗

arXiv2025

Design and Implementation of ARA Wireless Living Lab for Rural Broadband and Applications

Addressing the broadband gap between rural and urban regions requires rural-focused wireless research and innovation. In the meantime, rural regions provide rich, diverse use cases of advanced wireless, and they offer unique real-world settings for piloting applications that advance the frontiers of wireless systems (e.g., teleoperation of ground and aerial vehicles). To fill the broadband gap and to leverage the unique opportunities that rural regions provide for piloting advanced wireless applications, we design and implement the ARA wireless living lab for research and innovation in rural wireless systems and their applications in precision agriculture, community services, and so on. ARA focuses on the unique community, application, and economic context of rural regions, and it features the first-of-its-kind, real-world deployment of long-distance, high-capacity terrestrial wireless x-haul and access platforms as well as low-earth-orbit (LEO) satellite communications platforms across a rural area of diameter over 30 km. With both software-defined radios and programmable COTS systems, and through effective orchestration of these wireless resources with fiber as well as compute resources embedded end-to-end across user equipment (UE), base stations (BS), edge, and cloud, including support for Bring Your Own Device (BYOD), ARA offers programmability, performance, robustness, and heterogeneity at the same time, thus enabling rural-focused co-evolution of wireless and applications while helping advance the frontiers of wireless systems in domains such as Open RAN, NextG, and agriculture applications.

cs.NI↗

arXiv2025

SARLink: Satellite Backscatter Connectivity using Synthetic Aperture Radar

SARLink is a passive satellite backscatter communication system that uses existing spaceborne synthetic aperture radar (SAR) imaging satellites to provide connectivity in remote regions. It achieves orders of magnitude more range than traditional backscatter systems, enabling communication between a passive ground node and a satellite in low earth orbit. The system is composed of a cooperative ground target, a SAR satellite, and a data processing algorithm. A mechanically modulating reflector was designed to apply amplitude modulation to ambient SAR backscatter signals by changing its radar cross section. These communication bits are extracted from the raw SAR data using an algorithm that leverages subaperture processing to detect multiple bits from a target in a single image dataset. A theoretical analysis of this communication system using on-off keying is presented, including the expected signal model, throughput, and bit error rate. The results suggest a 5.5 ft by 5.5 ft modulating corner reflector could send 60 bits every satellite pass, enough to support low bandwidth sensor data and messages. Using Sentinel-1A, a SAR satellite at an altitude of 693~km, we deployed static and modulating reflectors to evaluate the system. The results, successfully detecting the changing state of a modulating ground target, demonstrate our algorithm's effectiveness for extracting bits, paving the way for ultra-long-range, low-power satellite backscatter communication.

eess.SP↗

arXiv2025

TerraTrace: Temporal Signature Land Use Mapping System

Understanding land use over time is critical to tracking events related to climate change, like deforestation. However, satellite-based remote sensing tools which are used for monitoring struggle to differentiate vegetation types in farms and orchards from forests. We observe that metrics such as the Normalized Difference Vegetation Index (NDVI), based on plant photosynthesis, have unique temporal signatures that reflect agricultural practices and seasonal cycles. We analyze yearly NDVI changes on 20 farms for 10 unique crops. Initial results show that NDVI curves are coherent with agricultural practices, are unique to each crop, consistent globally, and can differentiate farms from forests. We develop a novel longitudinal NDVI dataset for the state of California from 2020-2023 with 500~m resolution and over 70 million points. We use this to develop the TerraTrace platform, an end-to-end analytic tool that classifies land use using NDVI signatures and allows users to query the system through an LLM chatbot and graphical interface.

eess.IV↗

arXiv2024

Enabling Adoption of Regenerative Agriculture through Soil Carbon Copilots

Mitigating climate change requires transforming agriculture to minimize environ mental impact and build climate resilience. Regenerative agricultural practices enhance soil organic carbon (SOC) levels, thus improving soil health and sequestering carbon. A challenge to increasing regenerative agriculture practices is cheaply measuring SOC over time and understanding how SOC is affected by regenerative agricultural practices and other environmental factors and farm management practices. To address this challenge, we introduce an AI-driven Soil Organic Carbon Copilot that automates the ingestion of complex multi-resolution, multi-modal data to provide large-scale insights into soil health and regenerative practices. Our data includes extreme weather event data (e.g., drought and wildfire incidents), farm management data (e.g., cropland information and tillage predictions), and SOC predictions. We find that integrating public data and specialized models enables large-scale, localized analysis for sustainable agriculture. In comparisons of agricultural practices across California counties, we find evidence that diverse agricultural activity may mitigate the negative effects of tillage; and that while extreme weather conditions heavily affect SOC, composting may mitigate SOC loss. Finally, implementing role-specific personas empowers agronomists, farm consultants, policymakers, and other stakeholders to implement evidence-based strategies that promote sustainable agriculture and build climate resilience.

cs.IR↗

arXiv2024

Wireless Spectrum in Rural Farmlands: Status, Challenges and Opportunities

Due to factors such as low population density and expansive geographical distances, network deployment falls behind in rural regions, leading to a broadband divide. Wireless spectrum serves as the blood and flesh of wireless communications. Shared white spaces such as those in the TVWS and CBRS spectrum bands offer opportunities to expand connectivity, innovate, and provide affordable access to high-speed Internet in under-served areas without additional cost to expensive licensed spectrum. However, the current methods to utilize these white spaces are inefficient due to very conservative models and spectrum policies, causing under-utilization of valuable spectrum resources. This hampers the full potential of innovative wireless technologies that could benefit farmers, small Internet Service Providers (ISPs) or Mobile Network Operators (MNOs) operating in rural regions. This study explores the challenges faced by farmers and service providers when using shared spectrum bands to deploy their networks while ensuring maximum system performance and minimizing interference with other users. Additionally, we discuss how spatiotemporal spectrum models, in conjunction with database-driven spectrum-sharing solutions, can enhance the allocation and management of spectrum resources, ultimately improving the efficiency and reliability of wireless networks operating in shared spectrum bands.

cs.NI↗

arXiv2024

Exploring the Efficiency of Renewable Energy-based Modular Data Centers at Scale

Modular data centers (MDCs) that can be placed right at the energy farms and powered mostly by renewable energy, are proven to be a flexible and effective approach to lowering the carbon footprint of data centers. However, the main challenge of using renewable energy is the high variability of power produced, which implies large volatility in powering computing resources at MDCs, and degraded application performance due to the task evictions and migrations. This causes challenges for platform operators to decide the MDC deployment. To this end, we present SkyBox, a framework that employs a holistic and learning-based approach for platform operators to explore the efficient use of renewable energy with MDC deployment across geographical regions. SkyBox is driven by the insights based on our study of real-world power traces from a variety of renewable energy farms -- the predictable production of renewable energy and the complementary nature of energy production patterns across different renewable energy sources and locations. With these insights, SkyBox first uses the coefficient of variation metric to select the qualified renewable farms, and proposes a subgraph identification algorithm to identify a set of farms with complementary energy production patterns. After that, SkyBox enables smart workload placement and migrations to further tolerate the power variability. Our experiments with real power traces and datacenter workloads show that SkyBox has the lowest carbon emissions in comparison with current MDC deployment approaches. SkyBox also minimizes the impact of the power variability on cloud virtual machines, enabling rMDCs a practical solution of efficiently using renewable energy.

cs.DC↗

arXiv2024

Injecting New Knowledge into Large Language Models via Supervised Fine-Tuning

In recent years, Large Language Models (LLMs) have shown remarkable performance in generating human-like text, proving to be a valuable asset across various applications. However, adapting these models to incorporate new, out-of-domain knowledge remains a challenge, particularly for facts and events that occur after the model's knowledge cutoff date. This paper investigates the effectiveness of Supervised Fine-Tuning (SFT) as a method for knowledge injection in LLMs, specifically focusing on the domain of recent sporting events. We compare different dataset generation strategies -- token-based and fact-based scaling -- to create training data that helps the model learn new information. Our experiments on GPT-4 demonstrate that while token-based scaling can lead to improvements in Q&A accuracy, it may not provide uniform coverage of new knowledge. Fact-based scaling, on the other hand, offers a more systematic approach to ensure even coverage across all facts. We present a novel dataset generation process that leads to more effective knowledge ingestion through SFT, and our results show considerable performance improvements in Q&A tasks related to out-of-domain knowledge. This study contributes to the understanding of domain adaptation for LLMs and highlights the potential of SFT in enhancing the factuality of LLM responses in specific knowledge domains.

cs.CL↗

arXiv2024

Earth+: on-board satellite imagery compression leveraging historical earth observations

With the increasing deployment of earth observation satellite constellations, the downlink (satellite-to-ground) capacity often limits the freshness, quality, and coverage of the imagery data available to applications on the ground. To overcome the downlink limitation, we present Earth+, a new satellite imagery compression system that, instead of compressing each image individually, pinpoints and downloads only recent imagery changes with respect to the history reference images. To minimize the amount of changes, it is critical to make reference images as fresh as possible. Earth+ enables each satellite to choose fresh reference images from not only its own history images but also past images of other satellites from an entire satellite constellation. To share reference images across satellites, Earth+ utilizes the limited capacity of the existing uplink (ground-to-satellite) by judiciously selecting and compressing reference images while still allowing accurate change detection. In short, Earth+ is the first to make reference-based compression efficient, by enabling constellation-wide sharing of fresh reference images across satellites. Our evaluation shows that Earth+ can reduce the downlink usage by a factor of 3.3 compared to state-of-the-art on-board image compression techniques while not sacrificing image quality, or using more on-board computing or storage resources, or more uplink bandwidth than currently available.

cs.NI↗

arXiv2024

RAG vs Fine-tuning: Pipelines, Tradeoffs, and a Case Study on Agriculture

There are two common ways in which developers are incorporating proprietary and domain-specific data when building applications of Large Language Models (LLMs): Retrieval-Augmented Generation (RAG) and Fine-Tuning. RAG augments the prompt with the external data, while fine-Tuning incorporates the additional knowledge into the model itself. However, the pros and cons of both approaches are not well understood. In this paper, we propose a pipeline for fine-tuning and RAG, and present the tradeoffs of both for multiple popular LLMs, including Llama2-13B, GPT-3.5, and GPT-4. Our pipeline consists of multiple stages, including extracting information from PDFs, generating questions and answers, using them for fine-tuning, and leveraging GPT-4 for evaluating the results. We propose metrics to assess the performance of different stages of the RAG and fine-Tuning pipeline. We conduct an in-depth study on an agricultural dataset. Agriculture as an industry has not seen much penetration of AI, and we study a potentially disruptive application - what if we could provide location-specific insights to a farmer? Our results show the effectiveness of our dataset generation pipeline in capturing geographic-specific knowledge, and the quantitative and qualitative benefits of RAG and fine-tuning. We see an accuracy increase of over 6 p.p. when fine-tuning the model and this is cumulative with RAG, which increases accuracy by 5 p.p. further. In one particular experiment, we also demonstrate that the fine-tuned model leverages information from across geographies to answer specific questions, increasing answer similarity from 47% to 72%. Overall, the results point to how systems built using LLMs can be adapted to respond and incorporate knowledge across a dimension that is critical for a specific industry, paving the way for further applications of LLMs in other industrial domains.

cs.CL↗

arXiv2024

Domain Adaptation for Sustainable Soil Management using Causal and Contrastive Constraint Minimization

Monitoring organic matter is pivotal for maintaining soil health and can help inform sustainable soil management practices. While sensor-based soil information offers higher-fidelity and reliable insights into organic matter changes, sampling and measuring sensor data is cost-prohibitive. We propose a multi-modal, scalable framework that can estimate organic matter from remote sensing data, a more readily available data source while leveraging sparse soil information for improving generalization. Using the sensor data, we preserve underlying causal relations among sensor attributes and organic matter. Simultaneously we leverage inherent structure in the data and train the model to discriminate among domains using contrastive learning. This causal and contrastive constraint minimization ensures improved generalization and adaptation to other domains. We also shed light on the interpretability of the framework by identifying attributes that are important for improving generalization. Identifying these key soil attributes that affect organic matter will aid in efforts to standardize data collection efforts.

cs.LG↗