Searcharxiv⌕ Search

arXiv subjects

Koyel Mukherjee

Publications and source records attributed to Koyel Mukherjee.

At least 19 recordsLinked to original sources
arXiv2026

Online Learning with LLM Experts from Limited Feedback

We study adaptive routing of prompts to large language model (LLM) experts to maximize response quality in an online setting with limited feedback. We formulate it as a bandit problem with $K$ actions that represent experts and $d$ features that encode prompts, over a horizon of $T$ rounds. We propose algorithms that strategically select and observe rewards to minimize regret. In the full-information setting, we achieve a regret of $\tilde{O}(d T / \sqrt{m})$, while in the bandit setting we achieve $\tilde{O}(d T \sqrt{K / m})$, where $m \ll T$ is a budget on feedback. Our experiments show that we efficiently learn high-quality routing strategies across diverse LLMs from limited feedback.

cs.LG↗
arXiv2026

An Answer is just the Start: Related Insight Generation for Open-Ended Document-Grounded QA

Answering open-ended questions remains challenging for AI systems because it requires synthesis, judgment, and exploration beyond factual retrieval, and users often refine answers through multiple iterations rather than accepting a single response. Existing QA benchmarks do not explicitly support this refinement process. To address this gap, we introduce a new task, document-grounded related insight generation, where the goal is to generate additional insights from a document collection that help improve, extend, or rethink an initial answer to an open-ended question, ultimately supporting richer user interaction and a better overall question answering experience. We curate and release SCOpE-QA (Scientific Collections for Open-Ended QA), a dataset of 3,000 open-ended questions across 20 research collections. We present InsightGen, a two-stage approach that first constructs a thematic representation of the document collection using clustering, and then selects related context based on neighborhood selection from the thematic graph to generate diverse and relevant insights using LLMs. Extensive evaluation on 3,000 questions using two generation models and two evaluation settings shows that InsightGen consistently produces useful, relevant, and actionable insights, establishing a strong baseline for this new task.

cs.CL↗
arXiv2026

Masking or Mitigating? Deconstructing the Impact of Query Rewriting on Retriever Biases in RAG

Dense retrievers in retrieval-augmented generation (RAG) systems exhibit systematic biases -- including brevity, position, literal matching, and repetition biases -- that can compromise retrieval quality. Query rewriting techniques are now standard in RAG pipelines, yet their impact on these biases remains unexplored. We present the first systematic study of how query enhancement techniques affect dense retrieval biases, evaluating five methods across six retrievers. Our findings reveal that simple LLM-based rewriting achieves the strongest aggregate bias reduction (54\%), yet fails under adversarial conditions where multiple biases combine. Mechanistic analysis uncovers two distinct mechanisms: simple rewriting reduces bias through increased score variance, while pseudo-document generation methods achieve reduction through genuine decorrelation from bias-inducing features. However, no technique uniformly addresses all biases, and effects vary substantially across retrievers. Our results provide practical guidance for selecting query enhancement strategies based on specific bias vulnerabilities. More broadly, we establish a taxonomy distinguishing query-document interaction biases from document encoding biases, clarifying the limits of query-side interventions for debiasing RAG systems.

cs.IR↗
arXiv2026

ContextFocus: Activation Steering for Contextual Faithfulness in Large Language Models

Large Language Models (LLMs) encode vast amounts of parametric knowledge during pre-training. As world knowledge evolves, effective deployment increasingly depends on their ability to faithfully follow externally retrieved context. When such evidence conflicts with the model's internal knowledge, LLMs often default to memorized facts, producing unfaithful outputs. In this work, we introduce ContextFocus, a lightweight activation steering approach that improves context faithfulness in such knowledge-conflict settings while preserving fluency and efficiency. Unlike prior approaches, our solution requires no model finetuning and incurs minimal inference-time overhead, making it highly efficient. We evaluate ContextFocus on the ConFiQA benchmark, comparing it against strong baselines including ContextDPO, COIECD, and prompting-based methods. Furthermore, we show that our method is complementary to prompting strategies and remains effective on larger models. Extensive experiments show that ContextFocus significantly improves contextual-faithfulness. Our results highlight the effectiveness, robustness, and efficiency of ContextFocus in improving contextual-faithfulness of LLM outputs.

cs.CL↗
arXiv2025

From Selection to Generation: A Survey of LLM-based Active Learning

Active Learning (AL) has been a powerful paradigm for improving model efficiency and performance by selecting the most informative data points for labeling and training. In recent active learning frameworks, Large Language Models (LLMs) have been employed not only for selection but also for generating entirely new data instances and providing more cost-effective annotations. Motivated by the increasing importance of high-quality data and efficient model training in the era of LLMs, we present a comprehensive survey on LLM-based Active Learning. We introduce an intuitive taxonomy that categorizes these techniques and discuss the transformative roles LLMs can play in the active learning loop. We further examine the impact of AL on LLM learning paradigms and its applications across various domains. Finally, we identify open challenges and propose future research directions. This survey aims to serve as an up-to-date resource for researchers and practitioners seeking to gain an intuitive understanding of LLM-based AL techniques and deploy them to new applications.

cs.LG↗
arXiv2024

PromptRefine: Enhancing Few-Shot Performance on Low-Resource Indic Languages with Example Selection from Related Example Banks

Large Language Models (LLMs) have recently demonstrated impressive few-shot learning capabilities through in-context learning (ICL). However, ICL performance is highly dependent on the choice of few-shot demonstrations, making the selection of the most optimal examples a persistent research challenge. This issue is further amplified in low-resource Indic languages, where the scarcity of ground-truth data complicates the selection process. In this work, we propose PromptRefine, a novel Alternating Minimization approach for example selection that improves ICL performance on low-resource Indic languages. PromptRefine leverages auxiliary example banks from related high-resource Indic languages and employs multi-task learning techniques to align language-specific retrievers, enabling effective cross-language retrieval. Additionally, we incorporate diversity in the selected examples to enhance generalization and reduce bias. Through comprehensive evaluations on four text generation tasks -- Cross-Lingual Question Answering, Multilingual Question Answering, Machine Translation, and Cross-Lingual Summarization using state-of-the-art LLMs such as LLAMA-3.1-8B, LLAMA-2-7B, Qwen-2-7B, and Qwen-2.5-7B, we demonstrate that PromptRefine significantly outperforms existing frameworks for retrieving examples.

cs.CL↗
arXiv2024

FiRST: Finetuning Router-Selective Transformers for Input-Adaptive Latency Reduction

Auto-regressive Large Language Models (LLMs) demonstrate remarkable performance across different domains such as vision and language processing. However, due to sequential processing through a stack of transformer layers, autoregressive decoding faces significant computation/latency challenges, particularly in resource-constrained environments like mobile and edge devices. Existing approaches in literature that aim to improve latency via skipping layers have two distinct flavors - 1) Early exit, and 2) Input-agnostic heuristics where tokens exit at pre-determined layers irrespective of input sequence. Both the above strategies have limitations - the former cannot be applied to handle KV Caching necessary for speed-ups in modern framework and the latter does not capture the variation in layer importance across tasks or more generally, across input sequences. To address both limitations, we propose FiRST, an algorithm that reduces inference latency by using layer-specific routers to select a subset of transformer layers adaptively for each input sequence - the prompt (during the prefill stage) decides which layers will be skipped during decoding. FiRST preserves compatibility with KV caching enabling faster inference while being quality-aware. FiRST is model-agnostic and can be easily enabled on any pre-trained LLM. Our approach reveals that input adaptivity is critical - indeed, different task-specific middle layers play a crucial role in evolving hidden representations depending on tasks. Extensive experiments show that FiRST significantly reduces latency while outperforming other layer selection strategies in quality metics. It retains competitive performance to base model (without layer skipping) and in some cases, even improves upon it. FiRST is thus a promising and efficient solution for LLM deployment in low-resource environments.

cs.CL↗
arXiv2024

To Store or Not to Store: a graph theoretical approach for Dataset Versioning

In this work, we study the cost efficient data versioning problem, where the goal is to optimize the storage and reconstruction (retrieval) costs of data versions, given a graph of datasets as nodes and edges capturing edit/delta information. One central variant we study is MinSum Retrieval (MSR) where the goal is to minimize the total retrieval costs, while keeping the storage costs bounded. This problem (along with its variants) was introduced by Bhattacherjee et al. [VLDB'15]. While such problems are frequently encountered in collaborative tools (e.g., version control systems and data analysis pipelines), to the best of our knowledge, no existing research studies the theoretical aspects of these problems. We establish that the currently best-known heuristic, LMG, can perform arbitrarily badly in a simple worst case. Moreover, we show that it is hard to get $o(n)$-approximation for MSR on general graphs even if we relax the storage constraints by an $O(\log n)$ factor. Similar hardness results are shown for other variants. Meanwhile, we propose poly-time approximation schemes for tree-like graphs, motivated by the fact that the graphs arising in practice from typical edit operations are often not arbitrary. As version graphs typically have low treewidth, we further develop new algorithms for bounded treewidth graphs. Furthermore, we propose two new heuristics and evaluate them empirically. First, we extend LMG by considering more potential ``moves'', to propose a new heuristic LMG-All. LMG-All consistently outperforms LMG while having comparable run time on a wide variety of datasets, i.e., version graphs. Secondly, we apply our tree algorithms on the minimum-storage arborescence of an instance, yielding algorithms that are qualitatively better than all previous heuristics for MSR, as well as for another variant BoundedMin Retrieval (BMR).

cs.DS↗
arXiv2024

Towards Optimizing the Costs of LLM Usage

Generative AI and LLMs in particular are heavily used nowadays for various document processing tasks such as question answering and summarization. However, different LLMs come with different capabilities for different tasks as well as with different costs, tokenization, and latency. In fact, enterprises are already incurring huge costs of operating or using LLMs for their respective use cases. In this work, we propose optimizing the usage costs of LLMs by estimating their output quality (without actually invoking the LLMs), and then solving an optimization routine for the LLM selection to either keep costs under a budget, or minimize the costs, in a quality and latency aware manner. We propose a model to predict the output quality of LLMs on document processing tasks like summarization, followed by an LP rounding algorithm to optimize the selection of LLMs. We study optimization problems trading off the quality and costs, both theoretically and empirically. We further propose a sentence simplification model for reducing the number of tokens in a controlled manner. Additionally, we propose several deterministic heuristics for reducing tokens in a quality aware manner, and study the related optimization problem of applying the heuristics optimizing the quality and cost trade-off. We perform extensive empirical validation of our methods on not only enterprise datasets but also on open-source datasets, annotated by us, and show that we perform much better compared to closest baselines. Our methods reduce costs by 40%- 90% while improving quality by 4%-7%. We will release the annotated open source datasets to the community for further research and exploration.

cs.CL↗
arXiv2023

R2D2: Reducing Redundancy and Duplication in Data Lakes

Enterprise data lakes often suffer from substantial amounts of duplicate and redundant data, with data volumes ranging from terabytes to petabytes. This leads to both increased storage costs and unnecessarily high maintenance costs for these datasets. In this work, we focus on identifying and reducing redundancy in enterprise data lakes by addressing the problem of 'dataset containment'. To the best of our knowledge, this is one of the first works that addresses table-level containment at a large scale. We propose R2D2: a three-step hierarchical pipeline that efficiently identifies almost all instances of containment by progressively reducing the search space in the data lake. It first builds (i) a schema containment graph, followed by (ii) statistical min-max pruning, and finally, (iii) content level pruning. We further propose minimizing the total storage and access costs by optimally identifying redundant datasets that can be deleted (and reconstructed on demand) while respecting latency constraints. We implement our system on Azure Databricks clusters using Apache Spark for enterprise data stored in ADLS Gen2, and on AWS clusters for open-source data. In contrast to existing modified baselines that are inaccurate or take several days to run, our pipeline can process an enterprise customer data lake at the TB scale in approximately 5 hours with high accuracy. We present theoretical results as well as extensive empirical validation on both enterprise (scale of TBs) and open-source datasets (scale of MBs - GBs), which showcase the effectiveness of our pipeline.

cs.DB↗
arXiv2023

Approximate Caching for Efficiently Serving Diffusion Models

Text-to-image generation using diffusion models has seen explosive popularity owing to their ability in producing high quality images adhering to text prompts. However, production-grade diffusion model serving is a resource intensive task that not only require high-end GPUs which are expensive but also incurs considerable latency. In this paper, we introduce a technique called approximate-caching that can reduce such iterative denoising steps for an image generation based on a prompt by reusing intermediate noise states created during a prior image generation for similar prompts. Based on this idea, we present an end to end text-to-image system, Nirvana, that uses the approximate-caching with a novel cache management-policy Least Computationally Beneficial and Frequently Used (LCBFU) to provide % GPU compute savings, 19.8% end-to-end latency reduction and 19% dollar savings, on average, on two real production workloads. We further present an extensive characterization of real production text-to-image prompts from the perspective of caching, popularity and reuse of intermediate states in a large production environment.

cs.CV↗
arXiv2023

Towards Optimizing Storage Costs on the Cloud

We study the problem of optimizing data storage and access costs on the cloud while ensuring that the desired performance or latency is unaffected. We first propose an optimizer that optimizes the data placement tier (on the cloud) and the choice of compression schemes to apply, for given data partitions with temporal access predictions. Secondly, we propose a model to learn the compression performance of multiple algorithms across data partitions in different formats to generate compression performance predictions on the fly, as inputs to the optimizer. Thirdly, we propose to approach the data partitioning problem fundamentally differently than the current default in most data lakes where partitioning is in the form of ingestion batches. We propose access pattern aware data partitioning and formulate an optimization problem that optimizes the size and reading costs of partitions subject to access patterns. We study the various optimization problems theoretically as well as empirically, and provide theoretical bounds as well as hardness results. We propose a unified pipeline of cost minimization, called SCOPe that combines the different modules. We extensively compare the performance of our methods with related baselines from the literature on TPC-H data as well as enterprise datasets (ranging from GB to PB in volume) and show that SCOPe substantially improves over the baselines. We show significant cost savings compared to platform baselines, of the order of 50% to 83% on enterprise Data Lake datasets that range from terabytes to petabytes in volume.

cs.DB↗
arXiv2019

A Simple Dynamic Learning Rate Tuning Algorithm For Automated Training of DNNs

Training neural networks on image datasets generally require extensive experimentation to find the optimal learning rate regime. Especially, for the cases of adversarial training or for training a newly synthesized model, one would not know the best learning rate regime beforehand. We propose an automated algorithm for determining the learning rate trajectory, that works across datasets and models for both natural and adversarial training, without requiring any dataset/model specific tuning. It is a stand-alone, parameterless, adaptive approach with no computational overhead. We theoretically discuss the algorithm's convergence behavior. We empirically validate our algorithm extensively. Our results show that our proposed approach \emph{consistently} achieves top-level accuracy compared to SOTA baselines in the literature in natural as well as adversarial training.

cs.LG↗
arXiv2019

Layer Dynamics of Linearised Neural Nets

Despite the phenomenal success of deep learning in recent years, there remains a gap in understanding the fundamental mechanics of neural nets. More research is focussed on handcrafting complex and larger networks, and the design decisions are often ad-hoc and based on intuition. Some recent research has aimed to demystify the learning dynamics in neural nets by attempting to build a theory from first principles, such as characterising the non-linear dynamics of specialised \textit{linear} deep neural nets (such as orthogonal networks). In this work, we expand and derive properties of learning dynamics respected by general multi-layer linear neural nets. Although an over-parameterisation of a single layer linear network, linear multi-layer neural nets offer interesting insights that explain how learning dynamics proceed in small pockets of the data space. We show in particular that multiple layers in linear nets grow at approximately the same rate, and there are distinct phases of learning with markedly different layer growth. We then apply a linearisation process to a general RelU neural net and show how nonlinearity breaks down the growth symmetry observed in liner neural nets. Overall, our work can be viewed as an initial step in building a theory for understanding the effect of layer design on the learning dynamics from first principles.

cs.LG↗
arXiv2017

Impact of Detour-Aware Policies on Maximizing Profit in Ridesharing

This paper provides efficient solutions to maximize profit for commercial ridesharing services, under a pricing model with detour-based discounts for passengers. We propose greedy heuristics for real-time ride matching that offer different trade-offs between optimality and speed. Simulations on New York City (NYC) taxi trip data show that our heuristics are up to 90% optimal and 10^5 times faster than the (necessarily) exponential-time optimal algorithm. Commercial ridesharing service providers generate significant savings by matching multiple ride requests using heuristic methods. The resulting savings are typically shared between the service provider (in the form of increased profit) and the ridesharing passengers (in the form of discounts). It is not clear a priori how this split should be effected, since higher discounts would encourage more ridesharing, thereby increasing total savings, but the fraction of savings taken as profit is reduced. We simulate a scenario where the decisions of the passengers to opt for ridesharing depend on the discount offered by the service provider. We provide an adaptive learning algorithm IDFLA that learns the optimal profit-maximizing discount factor for the provider. An evaluation over NYC data shows that IDFLA, on average, learns the optimal discount factor in under 16 iterations. Finally, we investigate the impact of imposing a detour-aware routing policy based on sequential individual rationality, a recently proposed concept. Such restricted policies offer a better ride experience, increasing the provider's market share, but at the cost of decreased average per-ride profit due to the reduced number of matched rides. We construct a model that captures these opposing effects, wherein simulations based on NYC data show that a 7% increase in market share would suffice to offset the decreased average per-ride profit.

math.OC↗
arXiv2017

Learning to Partition using Score Based Compatibilities

We study the problem of learning to partition users into groups, where one must learn the compatibilities between the users to achieve optimal groupings. We define four natural objectives that optimize for average and worst case compatibilities and propose new algorithms for adaptively learning optimal groupings. When we do not impose any structure on the compatibilities, we show that the group formation objectives considered are $NP$ hard to solve and we either give approximation guarantees or prove inapproximability results. We then introduce an elegant structure, namely that of \textit{intrinsic scores}, that makes many of these problems polynomial time solvable. We explicitly characterize the optimal groupings under this structure and show that the optimal solutions are related to \emph{homophilous} and \emph{heterophilous} partitions, well-studied in the psychology literature. For one of the four objectives, we show $NP$ hardness under the score structure and give a $\frac{1}{2}$ approximation algorithm for which no constant approximation was known thus far. Finally, under the score structure, we propose an online low sample complexity PAC algorithm for learning the optimal partition. We demonstrate the efficacy of the proposed algorithm on synthetic and real world datasets.

cs.LG↗
arXiv2016

LP Rounding and Combinatorial Algorithms for Minimizing Active and Busy Time

We consider fundamental scheduling problems motivated by energy issues. In this framework, we are given a set of jobs, each with a release time, deadline and required processing length. The jobs need to be scheduled on a machine so that at most g jobs are active at any given time. The duration for which a machine is active (i.e., "on") is referred to as its active time. The goal is to find a feasible schedule for all jobs, minimizing the total active time. When preemption is allowed at integer time points, we show that a minimal feasible schedule already yields a 3-approximation (and this bound is tight) and we further improve this to a 2-approximation via LP rounding techniques. Our second contribution is for the non-preemptive version of this problem. However, since even asking if a feasible schedule on one machine exists is NP-hard, we allow for an unbounded number of virtual machines, each having capacity of g. This problem is known as the busy time problem in the literature and a 4-approximation is known for this problem. We develop a new combinatorial algorithm that gives a 3-approximation. Furthermore, we consider the preemptive busy time problem, giving a simple and exact greedy algorithm when unbounded parallelism is allowed, i.e., g is unbounded. For arbitrary g, this yields an algorithm that is 2-approximate.

cs.DS↗
arXiv2016

The Costs and Benefits of Sharing: Sequential Individual Rationality and Sequential Fairness

In designing dynamic shared service systems that incentivize customers to opt for shared rather than exclusive service, the traditional notion of individual rationality may be insufficient, as a customer's estimated utility could fluctuate arbitrarily during their time in the shared system, as long as their realized utility at service completion is not worse than that for exclusive service. In this work, within a model that explicitly considers the "inconvenience costs" incurred by customers due to sharing, we introduce the notion of sequential individual rationality (SIR) that requires that the disutility of existing customers is nonincreasing as the system state changes due to new customer arrivals. Next, under SIR, we observe that cost sharing can also be viewed as benefit sharing, which inspires a natural definition of sequential fairness (SF) - the total incremental benefit due to a new customer is shared among existing customers in proportion to the incremental inconvenience suffered. We demonstrate the effectiveness of these notions by applying them to a ridesharing system, where unexpected detours to pick up subsequent passengers inconvenience the existing passengers. Imposing SIR and SF reveals interesting and surprising results, including: (a) natural limits on the incremental detours permissible, (b) exact characterization of "SIR-feasible" routes, which boast sublinear upper and lower bounds on the fractional detours, (c) exact characterization of sequentially fair cost sharing schemes, which includes a strong requirement that passengers must compensate each other for the detour inconveniences that they cause, and (d) new algorithmic problems related to and motivated by SIR.

cs.GT↗