Arnab Raha

Aradhana Mohan Parvathy, Soumendu Kumar Ghosh, Shamik Kundu et al.

The rapid growth in sizes of Large language models (LLMs) results in high compute and memory costs during inference. Quantization has been a significant pathway to addressing this challenge. In the quest to push the limits of quantization, weights, which are static, can often be quantized aggressively (e.g. 4 bits), while activations often require higher precision (e.g., 8 bits) to preserve accuracy, forcing hardware to operate with higher-precision datapaths. We leverage the statistical property that a significant fraction of activations are concentrated around zero, resulting in sparsity in the higher-order bits. Our proposal, SPARQLe, is a hardware-software co-design framework that exploits this sub-precision redundancy in any given quantized model. SPARQLe represents each 2k-bit activation tensor as a dense k-bit LSB tensor and a sparse k-bit MSB tensor compressed with a precision bitmap, and proposes a lightweight algorithm to increase MSB sparsity. SPARQLe reduces activation memory traffic and enables efficient computation on k-bit datapaths while preserving 2k-bit activation accuracy. SPARQLe includes an accelerator that operates directly on this hybrid format with minimal control overheads. Across the BitNet 3B, Llama2 7B, and Llama3 8B models, SPARQLe reduces prefill latency by 16-24.3% and decode latency by 13.5-23.4%, with 17-26.7% and 6.5-14.2% lower prefill and decode energy, respectively. SPARQLe demonstrates that sub-precision activation sparsity offers an effective and complementary pathway towards efficient LLM inference.

Hoseok Kim, Arghadip Das, Soumendu Ghosh et al.

Modern edge System-on-Chips (SoCs) integrate heterogeneous processing units (PUs) such as CPUs, GPUs, and NPUs, yet current inference stacks map entire models to a single PU, leaving significant performance and energy efficiency on the table. This is exacerbated by emerging architectures such as state-space models (SSMs), Kolmogorov-Arnold networks (KANs), and multi-stage vision-language-action (VLA) pipelines, whose diverse operator characteristics are not uniformly suited to any single PU. We present BIDENT, a unified operator-level orchestration framework for heterogeneous edge inference that maps individual operators to the most suitable PU based on profiled execution characteristics. BIDENT formulates operator-to-PU assignment as a shortest-path problem over a weighted execution graph, enabling efficient and optimal scheduling under the cost model for both latency- and energy-minimization objectives. Unlike prior work relying on model-specific heuristics or coarse-grained partitioning, BIDENT is model-agnostic and jointly supports sequential execution, intra-model parallelism across independent operators, and multi-model concurrent scheduling in a single formulation. We implement BIDENT on an Intel Core Ultra SoC and evaluate it across 10 model families spanning CNNs, Transformers, SSMs, KANs, spiking networks, and multi-stage pipelines. BIDENT achieves up to 1.60x speedup via intra-model parallelism and a 3.42x geometric mean speedup across 190 multi-model combinations by utilizing otherwise idle compute. Sequential heterogeneous mapping yields more modest gains (up to 1.58x, 1.09x geometric mean), while energy-aware scheduling reduces energy consumption by 48.2% on average in concurrent settings. These results show that operator-level orchestration, not model-level mapping, is the key abstraction for fully exploiting heterogeneity in next-generation edge AI.

Arghadip Das, Hoseok Kim, Soomin Lee et al.

AI model architectures are diversifying rapidly. Although dense matrix multiplication underlies today's CNNs and transformers, emerging architectures (state-space models, long convolutions via the fast Fourier transform (FFT), Kolmogorov-Arnold networks, and spiking networks) are not multiply-accumulate (MAC) dominated; they spend much of their computation on vector and non-MAC primitives that homogeneous, MAC-centric neural processing units (NPUs) serve poorly. This has motivated heterogeneous NPUs (HPUs) built from non-identical tiles. Prior heterogeneous designs vary only one or two coarse knobs (typically MAC precision or array size) and are evaluated on narrow workloads; no existing framework supports fine-grained HPU design, where tiles differ across many architectural dimensions at once. We present MOSAIC, an analytical simulator and design-space-exploration (DSE) framework for HPU microarchitecture design. MOSAIC searches the joint space of tile-level heterogeneity: beyond array size and precision, it varies tile-type composition (large Big, small Little, and non-MAC Special-Function tiles), dataflow, sparsity mode, MAC engine type, and special-function units for non-MAC operators (FFT, spiking-integrate, polynomial). Unlike prior simulators that model a single homogeneous tile type, MOSAIC models non-MAC tiles with their own energy, area, and timing models and maps operators across a mix of tiles with a heterogeneity-aware compiler. A multi-seed pipeline pairing a stratified sweep with genetic-algorithm refinement returns Pareto-optimal designs, with cost models calibrated to a 7 nm node and cross-validated against NVIDIA's Deep Learning Accelerator (NVDLA). Across a 20-workload suite, the best general-purpose HPU found by MOSAIC (~200 mm^2 Big+Little+Special-Function) achieves +46.91% mean iso-area energy savings over the best iso-area homogeneous baseline.

Anand Menon, Samit S Miftah, Shamik Kundu et al.

Hardware verification is crucial in modern SoC design, consuming around 70% of development time. SystemVerilog assertions ensure correct functionality. However, existing industrial practices rely on manual efforts for assertion generation, which becomes increasingly untenable as hardware systems become complex. Recent research shows that Large Language Models (LLMs) can automate this process. However, proprietary SOTA models like GPT-4o often generate inaccurate assertions and require expensive licenses, while smaller open-source LLMs need fine-tuning to manage HDL code complexities. To address these issues, we introduce **VERT**, an open-source dataset designed to enhance SystemVerilog assertion generation using LLMs. VERT enables researchers in academia and industry to fine-tune open-source models, outperforming larger proprietary ones in both accuracy and efficiency while ensuring data privacy through local fine-tuning and eliminating costly licenses. The dataset is curated by systematically augmenting variables from open-source HDL repositories to generate synthetic code snippets paired with corresponding assertions. Experimental results demonstrate that fine-tuned models like Deepseek Coder 6.7B and Llama 3.1 8B outperform GPT-4o, achieving up to 96.88% improvement over base models and 24.14% over GPT-4o on platforms including OpenTitan, CVA6, OpenPiton and Pulpissimo. VERT is available at https://github.com/AnandMenon12/VERT.

Akshat Ramachandran, Souvik Kundu, Arnab Raha et al.

Large language model (LLM) pruning with fixed N:M structured sparsity significantly limits the expressivity of the sparse model, yielding sub-optimal performance. In contrast, supporting multiple N:M patterns to provide sparse representational freedom introduces costly overhead in hardware. To address these challenges for LLMs, we first present a flexible layer-wise outlier-density-aware N:M sparsity (FLOW) selection method. FLOW enables the identification of optimal layer-wise N and M values (from a given range) by simultaneously accounting for the presence and distribution of outliers, allowing a higher degree of representational freedom. To deploy sparse models with such N:M flexibility, we then introduce a flexible, low-overhead digital compute-in-memory architecture (FlexCiM). FlexCiM supports diverse sparsity patterns by partitioning a digital CiM (DCiM) macro into smaller sub-macros, which are adaptively aggregated and disaggregated through distribution and merging mechanisms for different N and M values. Extensive experiments on both transformer-based and recurrence-based state space foundation models (SSMs) demonstrate that FLOW outperforms existing alternatives with an accuracy improvement of up to 36%, while FlexCiM achieves up to 1.75x lower inference latency and 1.5x lower energy consumption compared to existing sparse accelerators. Code is available at: https://github.com/FLOW-open-project/FLOW

Arghadip Das, Arnab Raha, Shamik Kundu et al.

State-Space Models (SSMs) have emerged as efficient alternatives to transformers for sequential data tasks, offering linear or near-linear scalability with sequence length, making them ideal for long-sequence applications in NLP, vision, and edge AI, including real-time transcription, translation, and contextual search. These applications require lightweight, high-performance models for deployment on resource-constrained devices like laptops and PCs. Designing specialized accelerators for every emerging neural network is costly and impractical; instead, optimizing models for existing NPUs in AI PCs provides a scalable solution. To this end, we propose XAMBA, the first framework to enable and optimize SSMs on commercial off-the-shelf (COTS) state-of-the-art (SOTA) NPUs. XAMBA follows a three-step methodology: (1) enabling SSMs on NPUs, (2) optimizing performance to meet KPI requirements, and (3) trading accuracy for additional performance gains. After enabling SSMs on NPUs, XAMBA mitigates key bottlenecks using CumBA and ReduBA, replacing sequential CumSum and ReduceSum operations with matrix-based computations, significantly improving execution speed and memory efficiency. Additionally, ActiBA enhances performance by approximating expensive activation functions (e.g., Swish, Softplus) using piecewise linear mappings, reducing latency with minimal accuracy loss. Evaluations on an Intel Core Ultra Series 2 AI PC show that XAMBA achieves up to 4.8X speed-up over the baseline. Our implementation is available at https://github.com/arghadippurdue/XAMBA.

Sanjay Das, Swastik Bhattacharya, Souvik Kundu et al.

Large Language Models (LLMs) have revolutionized natural language processing (NLP), excelling in tasks like text generation and summarization. However, their increasing adoption in mission-critical applications raises concerns about hardware-based threats, particularly bit-flip attacks (BFAs). BFAs, enabled by fault injection methods such as Rowhammer, target model parameters in memory, compromising both integrity and performance. Identifying critical parameters for BFAs in the vast parameter space of LLMs poses significant challenges. While prior research suggests transformer-based architectures are inherently more robust to BFAs compared to traditional deep neural networks, we challenge this assumption. For the first time, we demonstrate that as few as three bit-flips can cause catastrophic performance degradation in an LLM with billions of parameters. Current BFA techniques are inadequate for exploiting this vulnerability due to the difficulty of efficiently identifying critical parameters within the immense parameter space. To address this, we propose AttentionBreaker, a novel framework tailored for LLMs that enables efficient traversal of the parameter space to identify critical parameters. Additionally, we introduce GenBFA, an evolutionary optimization strategy designed to refine the search further, isolating the most critical bits for an efficient and effective attack. Empirical results reveal the profound vulnerability of LLMs to AttentionBreaker. For example, merely three bit-flips (4.129 x 10^-9% of total parameters) in the LLaMA3-8B-Instruct 8-bit quantized (W8) model result in a complete performance collapse: accuracy on MMLU tasks drops from 67.3% to 0%, and Wikitext perplexity skyrockets from 12.6 to 4.72 x 10^5. These findings underscore the effectiveness of AttentionBreaker in uncovering and exploiting critical vulnerabilities within LLM architectures.

Mahadev Sunil Kumar, Arnab Raha, Debayan Das et al.

Distributed deep neural networks (DNNs) have become central to modern computer vision, yet their deployment on resource-constrained edge devices remains hindered by substantial parameter counts, computational demands, and the probability of device failure. Here, we present an approach to the efficient deployment of distributed DNNs that jointly respect hardware limitations, preserve task performance, and remain robust to partial system failures. Our method integrates structured model pruning with a multi-objective optimization framework to tailor network capacity for heterogeneous device constraints, while explicitly accounting for device availability and failure probability during deployment. We demonstrate this framework using Multi-View Convolutional Neural Networks (MVCNN), a state-of-the-art architecture for 3D object recognition, by quantifying the contribution of individual views to classification accuracy and allocating pruning budgets accordingly. Experimental results show that the resulting models satisfy user-specified bounds on accuracy and memory footprint, even under multiple simultaneous device failures. The inference time is reduced by factors up to 4.7x across diverse simulated device configurations. These findings suggest that performance-aware, view-adaptive, and failure-resilient compression provides a viable pathway for deploying complex vision models in distributed edge environments.

Sanjay Das, Swastik Bhattacharya, Shamik Kundu et al.

State-space models (SSMs), exemplified by the Mamba architecture, have recently emerged as state-of-the-art sequence-modeling frameworks, offering linear-time scalability together with strong performance in long-context settings. Owing to their unique combination of efficiency, scalability, and expressive capacity, SSMs have become compelling alternatives to transformer-based models, which suffer from the quadratic computational and memory costs of attention mechanisms. As SSMs are increasingly deployed in real-world applications, it is critical to assess their susceptibility to both software- and hardware-level threats to ensure secure and reliable operation. Among such threats, hardware-induced bit-flip attacks (BFAs) pose a particularly severe risk by corrupting model parameters through memory faults, thereby undermining model accuracy and functional integrity. To investigate this vulnerability, we introduce RAMBO, the first BFA framework specifically designed to target Mamba-based architectures. Through experiments on the Mamba-1.4b model with LAMBADA benchmark, a cloze-style word-prediction task, we demonstrate that flipping merely a single critical bit can catastrophically reduce accuracy from 74.64% to 0% and increase perplexity from 18.94 to 3.75 x 10^6. These results demonstrate the pronounced fragility of SSMs to adversarial perturbations.

Swastik Bhattacharya, Sanjay Das, Anand Menon et al.

Deep Neural Networks (DNNs) continue to grow in complexity with Large Language Models (LLMs) incorporating vast numbers of parameters. Handling these parameters efficiently in traditional accelerators is limited by data-transmission bottlenecks, motivating Compute-in-Memory (CiM) architectures that integrate computation within or near memory to reduce data movement. Recent work has explored CiM designs using Floating-Point (FP) and Integer (INT) operations. FP computations typically deliver higher output quality due to their wider dynamic range and precision, benefiting precision-sensitive Generative AI applications. These include models such as LLMs, thus driving advancements in FP-CiM accelerators. However, the vulnerability of FP-CiM to hardware faults remains underexplored, posing a major reliability concern in mission-critical settings. To address this gap, we systematically analyze hardware fault effects in FP-CiM by introducing bit-flip faults at key computational stages, including digital multipliers, CiM memory cells, and digital adder trees. Experiments with Convolutional Neural Networks (CNNs) such as AlexNet and state-of-the-art LLMs including LLaMA-3.2-1B and Qwen-0.3B-Base reveal how faults at each stage affect inference accuracy. Notably, a single adder fault can reduce LLM accuracy to 0%. Based on these insights, we propose a fault-resilient design, SafeCiM, that mitigates fault impact far better than a naive FP-CiM with a pre-alignment stage. For example, with 4096 MAC units, SafeCiM reduces accuracy degradation by up to 49x for a single adder fault compared to the baseline FP-CiM architecture.

Arghadip Das, Shamik Kundu, Arnab Raha et al.

Graph Neural Networks (GNNs) are vital for learning from graph-structured data, enabling applications in network analysis, recommendation systems, and speech analytics. Deploying them on edge devices like client PCs and laptops enhances real-time processing, privacy, and cloud independence. GNNs aid Retrieval-Augmented Generation (RAG) for Large Language Models (LLMs) and enable event-based vision tasks. However, irregular memory access, sparsity, and dynamic structures cause high latency and energy overhead on resource-constrained devices. While modern edge processors integrate CPUs, GPUs, and NPUs, NPUs designed for data-parallel tasks struggle with irregular GNN computations. We introduce GraNNite, the first hardware-aware framework optimizing GNN execution on commercial-off-the-shelf (COTS) SOTA DNN accelerators via a structured three-step methodology: (1) enabling NPU execution, (2) optimizing performance, and (3) trading accuracy for efficiency gains. Step 1 employs GraphSplit for workload distribution and StaGr for static aggregation, while GrAd and NodePad handle dynamic graphs. Step 2 boosts performance using EffOp for control-heavy tasks and GraSp for sparsity exploitation. Graph Convolution optimizations PreG, SymG, and CacheG reduce redundancy and memory transfers. Step 3 balances quality versus efficiency, where QuantGr applies INT8 quantization, and GrAx1, GrAx2, and GrAx3 accelerate attention, broadcast-add, and SAGE-max aggregation. On Intel Core Ultra AI PCs, GraNNite achieves 2.6X to 7.6X speedups over default NPU mappings and up to 8.6X energy gains over CPUs and GPUs, delivering 10.8X and 6.7X higher performance than CPUs and GPUs, respectively, across GNN models.

Ayush Arunachalam, Ian Kintz, Suvadeep Banerjee et al.

Given the widespread use of safety-critical applications in the automotive field, it is crucial to ensure the Functional Safety (FuSa) of circuits and components within automotive systems. The Analog and Mixed-Signal (AMS) circuits prevalent in these systems are more vulnerable to faults induced by parametric perturbations, noise, environmental stress, and other factors, in comparison to their digital counterparts. However, their continuous signal characteristics present an opportunity for early anomaly detection, enabling the implementation of safety mechanisms to prevent system failure. To address this need, we propose a novel framework based on unsupervised machine learning for early anomaly detection in AMS circuits. The proposed approach involves injecting anomalies at various circuit locations and individual components to create a diverse and comprehensive anomaly dataset, followed by the extraction of features from the observed circuit signals. Subsequently, we employ clustering algorithms to facilitate anomaly detection. Finally, we propose a time series framework to enhance and expedite anomaly detection performance. Our approach encompasses a systematic analysis of anomaly abstraction at multiple levels pertaining to the automotive domain, from hardware- to block-level, where anomalies are injected to create diverse fault scenarios. By monitoring the system behavior under these anomalous conditions, we capture the propagation of anomalies and their effects at different abstraction levels, thereby potentially paving the way for the implementation of reliable safety mechanisms to ensure the FuSa of automotive SoCs. Our experimental findings indicate that our approach achieves 100% anomaly detection accuracy and significantly optimizes the associated latency by 5X, underscoring the effectiveness of our devised solution.

Soubhagya Sutar, Arnab Raha, Vijay Raghunathan

Embedded systems play a crucial role in fueling the growth of the Internet-of-Things (IoT) in application domains such as healthcare, home automation, transportation, etc. However, their increasingly network-connected nature, coupled with their ability to access potentially sensitive/confidential information, has given rise to many security and privacy concerns. An additional challenge is the growing number of counterfeit components in these devices, resulting in serious reliability and financial implications. Physically Unclonable Functions (PUFs) are a promising security primitive to help address these concerns. Memory-based PUFs are particularly attractive as they require minimal or no additional hardware for their operation. However, current memory-based PUFs utilize only a single memory technology for constructing the PUF, which has several disadvantages including making them vulnerable to security attacks. In this paper, we propose the design of a new memory-based combination PUF that intelligently combines two memory technologies, SRAM and DRAM, to overcome these shortcomings. The proposed combination PUF exhibits high entropy, supports a large number of challenge-response pairs, and is intrinsically reconfigurable. We have implemented the proposed combination PUF using a Terasic TR4-230 FPGA board and several off-the-shelf SRAMs and DRAMs. Experimental results demonstrate substantial improvements over current memory-based PUFs including the ability to resist various attacks. Extensive authentication tests across a wide temperature range (20 - 60 deg. Celsius) and accelerated aging (12 months) demonstrate the robustness of the proposed design, which achieves a 100% true-positive rate and 0% false-positive rate for authentication across these parameter ranges.

Victor Seal, Arnab Raha, Shovan Maity et al.

This paper presents a forecasting model designed using WSNs (Wireless Sensor Networks) to predict flood in rivers using simple and fast calculations to provide real-time results and save the lives of people who may be affected by the flood. Our prediction model uses multiple variable robust linear regression which is easy to understand and simple and cost effective in implementation, is speed efficient, but has low resource utilization and yet provides real time predictions with reliable accuracy, thus having features which are desirable in any real world algorithm. Our prediction model is independent of the number of parameters, i.e. any number of parameters may be added or removed based on the on-site requirements. When the water level rises, we represent it using a polynomial whose nature is used to determine if the water level may exceed the flood line in the near future. We compare our work with a contemporary algorithm to demonstrate our improvements over it. Then we present our simulation results for the predicted water level compared to the actual water level.