Ayan Sengupta

Tanmoy Chakraborty, Ayan Sengupta, Suparna Bhattacharya et al.

Large language models (LLMs) frequently achieve impressive scores on standardized benchmarks, yet accuracy alone offers a limited view of their capabilities. Evaluating open-source LLMs through leaderboards faces persistent issues like data contamination, narrow task scope, and weak alignment with real-world reliability. Benchmark-based evaluations such as MMLU PRO, BBH, or IFEval primarily capture \textit{what} a model outputs on fixed test sets, not \textit{how} it processes information, calibrates uncertainty, or structures internal knowledge. In this article, we advocate for a shift from benchmark-centric evaluation toward a complementary, \textit{state-centered intrinsic assessment} of LLMs. To this end, we introduce \textbf{Latent Performance Profiling (LPP)} -- a framework that derives task-agnostic diagnostics from hidden activations and output distributions. LPP defines a set of scalar metrics on a model's latent representations and dynamics, revealing scale-independent traits that enable interpretable comparisons and uncover hidden vulnerabilities. Unlike static accuracy scores, LPP provides stable, architecture-sensitive signatures across models of similar size. With extensive empirical analyses across eight LLMs, spanning a size range of 0.5B-14B, we demonstrate that models with similar benchmark scores can exhibit contrasting latent profiles, such as differences in entropy or adaptability. Guided by these insights, we design synthetic probes for uncertainty and symbolic reasoning that align with intrinsic metrics while decoupling from leaderboard bias. We recommend that reporting LPP alongside benchmarks provides a deeper, interpretable understanding of model behavior, enabling more reliable model selection, safety assessment, and evaluation beyond surface-level accuracy.

Ayan Sengupta, Tharun Suresh, Md Shad Akhtar et al.

Being a popular mode of text-based communication in multilingual communities, code-mixing in online social media has became an important subject to study. Learning the semantics and morphology of code-mixed language remains a key challenge, due to scarcity of data and unavailability of robust and language-invariant representation learning technique. Any morphologically-rich language can benefit from character, subword, and word-level embeddings, aiding in learning meaningful correlations. In this paper, we explore a hierarchical transformer-based architecture (HIT) to learn the semantics of code-mixed languages. HIT consists of multi-headed self-attention and outer product attention components to simultaneously comprehend the semantic and syntactic structures of code-mixed texts. We evaluate the proposed method across 6 Indian languages (Bengali, Gujarati, Hindi, Tamil, Telugu and Malayalam) and Spanish for 9 NLP tasks on 17 datasets. The HIT model outperforms state-of-the-art code-mixed representation learning and multilingual language models in all tasks. We further demonstrate the generalizability of the HIT architecture using masked language modeling-based pre-training, zero-shot learning, and transfer learning approaches. Our empirical results show that the pre-training objectives significantly improve the performance on downstream tasks.

Ayan Sengupta, Md Shad Akhtar, Tanmoy Chakraborty

Code-mixing and script-mixing are prevalent across online social networks and multilingual societies. However, a user's preference toward code-mixing depends on the socioeconomic status, demographics of the user, and the local context, which existing generative models mostly ignore while generating code-mixed texts. In this work, we make a pioneering attempt to develop a persona-aware generative model to generate texts resembling real-life code-mixed texts of individuals. We propose a Persona-aware Generative Model for Code-mixed Generation, PARADOX, a novel Transformer-based encoder-decoder model that encodes an utterance conditioned on a user's persona and generates code-mixed texts without monolingual reference data. We propose an alignment module that re-calibrates the generated sequence to resemble real-life code-mixed texts. PARADOX generates code-mixed texts that are semantically more meaningful and linguistically more valid. To evaluate the personification capabilities of PARADOX, we propose four new metrics -- CM BLEU, CM Rouge-1, CM Rouge-L and CM KS. On average, PARADOX achieves 1.6 points better CM BLEU, 47% better perplexity and 32% better semantic coherence than the non-persona-based counterparts.

Aradhye Agarwal, Ayan Sengupta, Tanmoy Chakraborty

Test-time scaling (TTS) -- the dynamic allocation of compute during inference -- is a promising direction for improving reasoning in large language models (LLMs). However, a systematic comparison of well-known TTS strategies under identical conditions is missing, and the influence of model type and problem difficulty on performance remains unclear. To address these gaps, we conduct the first large-scale study of TTS, spanning over thirty billion tokens generated using eight open-source LLMs (7B to 235B parameters), across four reasoning datasets. We observe three consistent trends: (1) no single TTS strategy universally dominates; (2) reasoning models exhibit distinct trace-quality patterns across problem difficulty and trace length, forming short-horizon and long-horizon categories; and (3) for a given model type, the optimal TTS performance scales monotonically with compute budget. Based on these insights, we provide a practical recipe for selecting the best TTS strategy, considering problem difficulty, model type, and compute budget, providing a practical guide to effective inference-time scaling.

Samhruth Ananthanarayanan, Ayan Sengupta, Tanmoy Chakraborty

As context windows in LLMs scale to 100K+ tokens, the key-value (KV) cache becomes the dominant memory bottleneck, with recent methods claiming 80-90% savings and minimal benchmark degradation. We argue these evaluations miss a structural issue: attention is not just storage but routing, and retaining KV pairs does not guarantee semantic accessibility. We propose a physics-inspired view of KV compression as a controlled perturbation of token-level routing, distinguishing retention, accessibility, and utilization. Using synthetic tasks probing multi-entity tracking, disambiguation, coreference, and multi-hop reasoning, we find that moderate compression degrades internal representations with little accuracy loss, revealing redundancy; all models exhibit a sharp hallucination safety cliff near 90% compression, correlated with spikes in Global Eviction Ratio (GER), suggesting a phase transition in semantic reachability; and architectures differ in routing dynamics, with LLaMA showing early consensus and late diversification, and Qwen showing funnel-like late convergence, leading to distinct resilience profiles. Beyond erasure, we identify representational rigidity, where excessive head-level consensus collapses routing flexibility despite token survival. These results suggest sparse token-route structures govern compression tolerance, reframing KV compression as a structural probe of attention geometry and linking long-context scalability to sparsity and the lottery ticket hypothesis in self-attention.

Aradhye Agarwal, Suhas K Ramesh, Ayan Sengupta et al.

Fine-tuning large language models (LLMs) on downstream tasks requires substantial computational resources. Selective PEFT, a class of parameter-efficient fine-tuning (PEFT) methodologies, aims to mitigate these computational challenges by selectively fine-tuning only a small fraction of the model parameters. Although parameter-efficient, these techniques often fail to match the performance of fully fine-tuned models, primarily due to inherent biases introduced during parameter selection. Traditional selective PEFT techniques use a fixed set of parameters selected using different importance heuristics, failing to capture parameter importance dynamically and often leading to suboptimal performance. We introduce $\text{ID}^3$, a novel selective PEFT method that calculates parameter importance continually, and dynamically unmasks parameters by balancing exploration and exploitation in parameter selection. Our empirical study on 16 tasks spanning natural language understanding, mathematical reasoning and summarization demonstrates the effectiveness of our method compared to fixed-masking selective PEFT techniques. We analytically show that $\text{ID}^3$ reduces the number of gradient updates by a factor of two, enhancing computational efficiency. Since $\text{ID}^3$ is robust to random initialization of neurons and operates directly on the optimization process, it is highly flexible and can be integrated with existing additive and reparametrization-based PEFT techniques such as adapters and LoRA respectively.

Ayan Sengupta, Shantanu Dixit, Md Shad Akhtar et al.

Knowledge distillation (KD) methods are pivotal in compressing large pre-trained language models into smaller models, ensuring computational efficiency without significantly dropping performance. Traditional KD techniques assume homogeneity in modalities between the teacher (source) and the student (target) models. On the other hand, existing multimodal knowledge distillation methods require modality-specific pre-training of the teacher model, which is computationally infeasible in most cases. In this paper, we introduce ARMADA, an efficient cross-modal knowledge distillation framework designed to transfer knowledge from large vision-language models, including black-box models, to language-only models. Unlike existing KD techniques that rely on the internal structures of multimodal teachers or require computationally expensive pre-training, ARMADA leverages novel alignment techniques to distil knowledge without altering the teacher model, ensuring efficiency and scalability. We empirically validate ARMADA on twelve natural language understanding, eight complex generative reasoning and five instruction-tuning tasks, demonstrating consistent performance improvements in large models such as DeBERTa-v2-1.4B, OPT-1.3B, LLaMA-{3B, 7B, 8B}. ARMADA achieves up to 3.4% improvement on language understanding tasks and 2.6% boost in generative reasoning, all without requiring expensive multimodal pre-training or fine-tuning of the teacher model. Our findings challenge conventional knowledge distillation paradigms by demonstrating that even vision-language models, despite lacking direct textual understanding, can significantly enhance language models when distilled appropriately.

Ayan Sengupta, Md Shad Akhtar, Tanmoy Chakraborty

Multi-head self-attention-based Transformers have shown promise in different learning tasks. Albeit these models exhibit significant improvement in understanding short-term and long-term contexts from sequences, encoders of Transformers and their variants fail to preserve layer-wise contextual information. Transformers usually project tokens onto sparse manifolds and fail to preserve mathematical equivalence among the token representations. In this work, we propose TransJect, an encoder model that guarantees a theoretical bound for layer-wise distance preservation between a pair of tokens. We propose a simple alternative to dot-product attention to ensure Lipschitz continuity. This allows TransJect to learn injective mappings to transform token representations to different manifolds with similar topology and preserve Euclidean distance between every pair of tokens in subsequent layers. Evaluations across multiple benchmark short- and long-sequence classification tasks show maximum improvements of 6.8% and 5.9%, respectively, over the variants of Transformers. Additionally, TransJect displays 79% better performance than Transformer on the language modeling task. We further highlight the shortcomings of multi-head self-attention from the statistical physics viewpoint. Although multi-head self-attention was incepted to learn different abstraction levels within the networks, our empirical analyses suggest that different attention heads learn randomly and unorderly. In contrast, TransJect adapts a mixture of experts for regularization; these experts are more orderly and balanced and learn different sparse representations from the input sequences. TransJect exhibits very low entropy and can be efficiently scaled to larger depths.

Ayan Sengupta, Siddhant Chaudhary, Tanmoy Chakraborty

The ever-increasing size of large language models (LLMs) presents significant challenges for deployment due to their heavy computational and memory requirements. Current model pruning techniques attempt to alleviate these issues by relying heavily on external calibration datasets to determine which parameters to prune or compress, thus limiting their flexibility and scalability across different compression ratios. Moreover, these methods often cause severe performance degradation, particularly in downstream tasks, when subjected to higher compression rates. In this paper, we propose PruneNet, a novel model compression method that addresses these limitations by reformulating model pruning as a policy learning process. PruneNet decouples the pruning process from the model architecture, eliminating the need for calibration datasets. It learns a stochastic pruning policy to assess parameter importance solely based on intrinsic model properties while preserving the spectral structure to minimize information loss. PruneNet can compress the LLaMA-2-7B model in just 15 minutes, achieving over 80% retention of its zero-shot performance with a 30% compression ratio, outperforming existing methods that retain only 75% performance. Furthermore, on complex multitask language understanding tasks, PruneNet demonstrates its robustness by preserving up to 80% performance of the original model, proving itself a superior alternative to conventional structured compression techniques.

Ayan Sengupta, Yash Goel, Tanmoy Chakraborty

Neural scaling laws have revolutionized the design and optimization of large-scale AI models by revealing predictable relationships between model size, dataset volume, and computational resources. Early research established power-law relationships in model performance, leading to compute-optimal scaling strategies. However, recent studies highlighted their limitations across architectures, modalities, and deployment contexts. Sparse models, mixture-of-experts, retrieval-augmented learning, and multimodal models often deviate from traditional scaling patterns. Moreover, scaling behaviors vary across domains such as vision, reinforcement learning, and fine-tuning, underscoring the need for more nuanced approaches. In this survey, we synthesize insights from over 50 studies, examining the theoretical foundations, empirical findings, and practical implications of scaling laws. We also explore key challenges, including data efficiency, inference scaling, and architecture-specific constraints, advocating for adaptive scaling strategies tailored to real-world applications. We suggest that while scaling laws provide a useful guide, they do not always generalize across all architectures and training strategies.

Ayan Sengupta, Vaibhav Seth, Arinjay Pathak et al.

Large Language Models (LLMs) are highly resource-intensive to fine-tune due to their enormous size. While low-rank adaptation is a prominent parameter-efficient fine-tuning approach, it suffers from sensitivity to hyperparameter choices, leading to instability in model performance on fine-tuning downstream tasks. This paper highlights the importance of effective parameterization in low-rank fine-tuning to reduce estimator variance and enhance the stability of final model outputs. We propose MonteCLoRA, an efficient fine-tuning technique that employs Monte Carlo estimation to learn an unbiased posterior estimation of low-rank parameters with low expected variance, stabilizing fine-tuned LLMs with only O(r) additional parameters, for a given rank r. MonteCLoRA shows 0.5% and 1.6% improvements in accuracy and robustness over unregularized low-rank adaptation method on natural language understanding tasks with pre-trained RoBERTa-base. Furthermore, in generative tasks with pre-trained LLaMA-1-7B and LLaMA-3.2-3B-Instruct, MonteCLoRA demonstrates robust performance with 50% and 62% lower spreads respectively than the contemporary efficient fine-tuning methods. The theoretical and empirical results presented in the paper underscore how parameterization and hyperpriors balance exploration-exploitation in the low-rank parametric space, therefore leading to more optimal and robust parameter estimation during efficient fine-tuning.

Suhas Kamasetty Ramesh, Ayan Sengupta, Tanmoy Chakraborty

Knowledge distillation (KD) is a key technique for compressing large language models into smaller ones while preserving performance. Despite the recent traction of KD research, its effectiveness for smaller language models (LMs) and the mechanisms driving knowledge transfer remain underexplored. In this work, we present the first large-scale empirical and statistical analysis of KD across models ranging from 0.5B to 7B parameters on 14 complex reasoning tasks in a zero-shot setting. Our findings reveal that KD can improve the average performance of smaller models by up to $10\%$, with a peak task specific gain of $22\%$, while providing only marginal benefits ($\sim 1.3\%$) for larger models. Surprisingly, teacher performance has a minimal impact on student outcomes, while teacher task expertise impacts KD effectiveness. A correlation study indicates that smaller LMs benefit more from KD, whereas larger LMs show diminished gains. Additionally, we uncover a misalignment between improvements in student performance and reasoning fidelity, suggesting that while KD enhances accuracy, it does not always maintain the structured decision-making processes of the teacher. Our ablation study further highlights the importance of teacher signals and logit smoothing in influencing students' performance after distillation. Overall, our study offers a comprehensive empirical and statistical assessment of KD, highlighting both its benefits and trade-offs when distilling knowledge from larger to smaller LMs.

Aradhye Agarwal, Ayan Sengupta, Tanmoy Chakraborty

Test-time scaling (TTS), which involves dynamic allocation of compute during inference, offers a promising way to improve reasoning in large language models. While existing TTS methods work well, they often rely on long decoding paths or require a large number of samples to be generated, increasing the token usage and inference latency. We observe the surprising fact that for reasoning tasks, shorter traces are much more likely to be correct than longer ones. Motivated by this, we introduce First Finish Search (FFS), a training-free parallel decoding strategy that launches $n$ independent samples and returns as soon as any one completes. We evaluate FFS alongside simple decoding, beam search, majority voting, and budget forcing on four reasoning models (DeepSeek-R1, R1-Distill-Qwen-32B, QwQ-32B and Phi-4-Reasoning-Plus) and across four datasets (AIME24, AIME25-I, AIME25-II and GPQA Diamond). With DeepSeek-R1, FFS achieves $82.23\%$ accuracy on the AIME datasets, a $15\%$ improvement over DeepSeek-R1's standalone accuracy, nearly matching OpenAI's o4-mini performance. Our theoretical analysis explains why stopping at the shortest trace is likely to yield a correct answer and identifies the conditions under which early stopping may be suboptimal. The elegance and simplicity of FFS demonstrate that straightforward TTS strategies can perform remarkably well, revealing the untapped potential of simple approaches at inference time.

Ayan Sengupta, Siddhant Chaudhary, Tanmoy Chakraborty

Key-value (KV) cache compression has emerged as a critical technique for reducing the memory and latency overhead of autoregressive language models during inference. Prior approaches predominantly rely on query-key attention scores to rank and evict cached tokens, assuming that attention intensity correlates with semantic importance. However, this heuristic overlooks the contribution of value vectors, which directly influence the attention output. In this paper, we propose CurDKV, a novel, value-centric KV compression method that selects keys and values based on leverage scores computed from CUR matrix decomposition. Our approach approximates the dominant subspace of the attention output $softmax(QK^T)V$, ensuring that the retained tokens best preserve the model's predictive behavior. Theoretically, we show that attention score approximation does not guarantee output preservation, and demonstrate that CUR-based selection minimizes end-to-end attention reconstruction loss. Empirically, CurDKV achieves up to 9.6% higher accuracy than state-of-the-art methods like SnapKV and ChunkKV under aggressive compression budgets on LLaMA and Mistral, while maintaining compatibility with FlashAttention and Grouped Query Attention. In addition to improved accuracy, CurDKV reduces generation latency by up to 40% at high compression, offering a practical speed-accuracy tradeoff.

Yash Goel, Ayan Sengupta, Tanmoy Chakraborty

We challenge the dominant focus on neural scaling laws and advocate for a paradigm shift toward downscaling in the development of large language models (LLMs). While scaling laws have provided critical insights into performance improvements through increasing model and dataset size, we emphasize the significant limitations of this approach, particularly in terms of computational inefficiency, environmental impact, and deployment constraints. To address these challenges, we propose a holistic framework for downscaling LLMs that seeks to maintain performance while drastically reducing resource demands. This paper outlines practical strategies for transitioning away from traditional scaling paradigms, advocating for a more sustainable, efficient, and accessible approach to LLM development.

Ayan Sengupta, Siddhant Chaudhary, Tanmoy Chakraborty

We introduce compression laws for language language models (LLMs). While recent scaling laws have sought to understand how LLMs scale with respect to model size, pre-training data, and computational resources, we focus on understanding how model compression affects the performance of a pre-trained LLM on downstream tasks. We empirically examine the effects of structured model compression on LLMs through over $1000$ experiments across eight models with sizes ranging from $0.5B$ to $14B$ parameters. Our findings indicate that the test cross-entropy loss increases quadratically with the compression ratio, whereas performance on downstream tasks declines only linearly. Our study emphasizes the importance of recovery fine-tuning in enhancing generation loss, showing that the test loss of compressed LLMs can improve by up to 55% with recovery fine-tuning. At higher compression ratios (up to 90%), compressed LLMs demonstrate a speed increase of 60% during inference compared to their uncompressed counterparts, compensating for the performance degradation at this level. However, for smaller models ($\le 7B$), the computational gains are limited, peaking at just 35%. We conclude that model compression can be highly beneficial for larger models, especially when a smaller model within the same computational budget is not available. These insights provide the practical guidelines for utilizing model compression techniques for adopting LLMs in real-life applications in resource-constrained settings.

Ayan Sengupta, Sourabh Kumar Bhattacharjee, Tanmoy Chakraborty et al.

Understanding linguistics and morphology of resource-scarce code-mixed texts remains a key challenge in text processing. Although word embedding comes in handy to support downstream tasks for low-resource languages, there are plenty of scopes in improving the quality of language representation particularly for code-mixed languages. In this paper, we propose HIT, a robust representation learning method for code-mixed texts. HIT is a hierarchical transformer-based framework that captures the semantic relationship among words and hierarchically learns the sentence-level semantics using a fused attention mechanism. HIT incorporates two attention modules, a multi-headed self-attention and an outer product attention module, and computes their weighted sum to obtain the attention weights. Our evaluation of HIT on one European (Spanish) and five Indic (Hindi, Bengali, Tamil, Telugu, and Malayalam) languages across four NLP tasks on eleven datasets suggests significant performance improvement against various state-of-the-art systems. We further show the adaptability of learned representation across tasks in a transfer learning setup (with and without fine-tuning).

Ayan Sengupta, William Scott Paka, Suman Roy et al.

Short text is a popular avenue of sharing feedback, opinions and reviews on social media, e-commerce platforms, etc. Many companies need to extract meaningful information (which may include thematic content as well as semantic polarity) out of such short texts to understand users' behaviour. However, obtaining high quality sentiment-associated and human interpretable themes still remains a challenge for short texts. In this paper we develop ELJST, an embedding enhanced generative joint sentiment-topic model that can discover more coherent and diverse topics from short texts. It uses Markov Random Field Regularizer that can be seen as a generalisation of skip-gram based models. Further, it can leverage higher-order semantic information appearing in word embedding, such as self-attention weights in graphical models. Our results show an average improvement of 10% in topic coherence and 5% in topic diversification over baselines. Finally, ELJST helps understand users' behaviour at more granular levels which can be explained. All these can bring significant values to the service and healthcare industries often dealing with customers.

Ayan Sengupta, Yasser Mohammad, Shinji Nakadai

Despite abundant negotiation strategies in literature, the complexity of automated negotiation forbids a single strategy from being dominant against all others in different negotiation scenarios. To overcome this, one approach is to use mixture of experts, but at the same time, one problem of this method is the selection of experts, as this approach is limited by the competency of the experts selected. Another problem with most negotiation strategies is their incapability of adapting to dynamic variation of the opponent's behaviour within a single negotiation session resulting in poor performance. This work focuses on both, solving the problem of expert selection and adapting to the opponent's behaviour with our Autonomous Negotiating Agent Framework. This framework allows real-time classification of opponent's behaviour and provides a mechanism to select, switch or combine strategies within a single negotiation session. Additionally, our framework has a reviewer component which enables self-enhancement capability by deciding to include new strategies or replace old ones with better strategies periodically. We demonstrate an instance of our framework by implementing maximum entropy reinforcement learning based strategies with a deep learning based opponent classifier. Finally, we evaluate the performance of our agent against state-of-the-art negotiators under varied negotiation scenarios.

Bodhisattwa Prasad Majumder, Ayan Sengupta, Sajal jain et al.

With the advancement of huge data generation and data handling capability, Machine Learning and Probabilistic modelling enables an immense opportunity to employ predictive analytics platform in high security critical industries namely data centers, electricity grids, utilities, airport etc. where downtime minimization is one of the primary objectives. This paper proposes a novel, complete architecture of an intelligent predictive analytics platform, Fault Engine, for huge device network connected with electrical/information flow. Three unique modules, here proposed, seamlessly integrate with available technology stack of data handling and connect with middleware to produce online intelligent prediction in critical failure scenarios. The Markov Failure module predicts the severity of a failure along with survival probability of a device at any given instances. The Root Cause Analysis model indicates probable devices as potential root cause employing Bayesian probability assignment and topological sort. Finally, a community detection algorithm produces correlated clusters of device in terms of failure probability which will further narrow down the search space of finding route cause. The whole Engine has been tested with different size of network with simulated failure environments and shows its potential to be scalable in real-time implementation.