Krishna Bajpai
Krishna Bajpai has a total of 2 co-authors. Here are the top co-authors by number of shared publications:
Vedanshi Gupta
1 shared publication- Vedanshi Gupta1 shared publication
Latest Publications
- Open Access
- 3 March 2026
- Published by Elsevier in SSRN Electronic Journal
EcoLLM: A Joint Optimization Framework for Ultra-low Power, Mixed-precision LLM Inference on Resource-constrained Edge Systems
Deployment of Large Language Models (LLMs) on resource-constrained edge devices, such as IoT systems and mobile AI platforms, is significantly hampered by excessive energy consumption, large memory footprints, and high inference latency. Current methods relying on the sequential application of model compression techniques—namely quantization and pruning—often result in sub-optimal and hardware-unaware configurations. This paper introduces EcoLLM, a novel joint optimization framework designed for achieving ultra-low power LLM inference. EcoLLM systematically determines the optimal layer-wise configurations of mixed-precision bit-widths (bw) and structured sparsity rates (s) guided by a predictive, hardwareaware energy consumption model. The primary technical contributions include: (1) The E-Model, a calibrated analytical Energy Cost Model that rigorously quantifies layer-wise energy consumption by explicitly incorporating the effects of data movement, precision, and sparsity. (2) The E-Metric (Energy Efficiency Metric), a holistic performance index that quantifies the energyaccuracy Pareto frontier by normalizing efficiency (Tokens/Joule) against model degradation (Perplexity loss). (3) The Adaptive Allocation Algorithm (A-Algorithm), an iterative greedy search approach designed to minimize the E-Metric across the entire network while simultaneously mitigating the cumulative effects of reconstruction error propagation. Implemented within the opensource TinyEdgeLLM Library and validated on a reference edge NPU/CPU platform, EcoLLM achieved a 6.4x compression ratio (from 487MB to 76MB) on the GPT-2 architecture with < 2.0% perplexity degradation, resulting in a 45% reduction in power consumption (from 3.8W to 2.1W) on a reference Raspberry Pi 4 platform.1) KeywordsLarge Language Models, Energy Optimization, Mixed-Precision, Quantization, Pruning, Perplexity, Raspberry Pi 4 platform
- 3 March 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessEcoLLM: A Joint Optimization Framework for Ultra-low Power, Mixed-precision LLM Inference on Resource-constrained Edge Systems
Deployment of Large Language Models (LLMs) on resource-constrained edge devices, such as IoT systems and mobile AI platforms, is significantly hampered by excessive energy consumption, large memory footprints, and high inference latency. Current methods relying on the sequential application of model compression techniques—namely quantization and pruning—often result in sub-optimal and hardware-unaware configurations. This paper introduces EcoLLM, a novel joint optimization framework designed for achieving ultra-low power LLM inference. EcoLLM systematically determines the optimal layer-wise configurations of mixed-precision bit-widths (bw) and structured sparsity rates (s) guided by a predictive, hardwareaware energy consumption model. The primary technical contributions include: (1) The E-Model, a calibrated analytical Energy Cost Model that rigorously quantifies layer-wise energy consumption by explicitly incorporating the effects of data movement, precision, and sparsity. (2) The E-Metric (Energy Efficiency Metric), a holistic performance index that quantifies the energyaccuracy Pareto frontier by normalizing efficiency (Tokens/Joule) against model degradation (Perplexity loss). (3) The Adaptive Allocation Algorithm (A-Algorithm), an iterative greedy search approach designed to minimize the E-Metric across the entire network while simultaneously mitigating the cumulative effects of reconstruction error propagation. Implemented within the opensource TinyEdgeLLM Library and validated on a reference edge NPU/CPU platform, EcoLLM achieved a 6.4x compression ratio (from 487MB to 76MB) on the GPT-2 architecture with < 2.0% perplexity degradation, resulting in a 45% reduction in power consumption (from 3.8W to 2.1W) on a reference Raspberry Pi 4 platform.1) KeywordsLarge Language Models, Energy Optimization, Mixed-Precision, Quantization, Pruning, Perplexity, Raspberry Pi 4 platform
- Open Access
- 26 December 2025
- Published by MDPI AG in Preprints.org
HyperFabric Interconnect (HFI): A Unified, Scalable Communication Fabric for HPC, AI, Quantum, and Neuromorphic Workloads
The evolution of high-performance computing (HPC) interconnects has produced specialized fabrics such as InfiniBand[1], Intel Omni-Path, and NVIDIA NVLink[2], each optimized for distinct workloads. However, the increasing convergence of HPC, AI/ML, quantum, and neuromorphic computing requires a unified communication substrate capable of supporting diverse requirements including ultra-low latency, high bandwidth, collective operations, and adaptive routing. We present HyperFabric Interconnect (HFI), a novel design that combines the strengths of existing interconnects while addressing their scalability and workload-fragmentation limitations. Our evaluation on simulated clusters demonstrates HFI’s ability to reduce job completion time (JCT) by up to 30%, improve tail latency consistency by 45% under mixed loads and 4× better jitter control in latency-sensitive applications., and sustain efficient scaling across heterogeneous workloads. Beyond simulation, we provide an analytical model and deployment roadmap that highlight HFI’s role as a converged interconnect for the exascale and post-exascale era.
- 26 December 2025
- Published by MDPI AG in Preprints.org
Open AccessHyperFabric Interconnect (HFI): A Unified, Scalable Communication Fabric for HPC, AI, Quantum, and Neuromorphic Workloads
The evolution of high-performance computing (HPC) interconnects has produced specialized fabrics such as InfiniBand[1], Intel Omni-Path, and NVIDIA NVLink[2], each optimized for distinct workloads. However, the increasing convergence of HPC, AI/ML, quantum, and neuromorphic computing requires a unified communication substrate capable of supporting diverse requirements including ultra-low latency, high bandwidth, collective operations, and adaptive routing. We present HyperFabric Interconnect (HFI), a novel design that combines the strengths of existing interconnects while addressing their scalability and workload-fragmentation limitations. Our evaluation on simulated clusters demonstrates HFI’s ability to reduce job completion time (JCT) by up to 30%, improve tail latency consistency by 45% under mixed loads and 4× better jitter control in latency-sensitive applications., and sustain efficient scaling across heterogeneous workloads. Beyond simulation, we provide an analytical model and deployment roadmap that highlight HFI’s role as a converged interconnect for the exascale and post-exascale era.
- Open Access
- 25 December 2025
- Published by MDPI AG in Preprints.org
Strategic Ignorance in Financial Markets: When Not Knowing Improves Alpha and Stability
The contemporary financial landscape is characterized by a paradox of plenty: while the volume and velocity of market data have increased exponentially, the decision-making quality of human and algorithmic agents often deteriorates under the weight of information overload. This research investigates the concept of strategic ignorance—the deliberate decision to filter, delay, or ignore specific information streams—as a mechanism to enhance alpha generation and market stability. By integrating theories of rational inattention, information design, and multi-agent reinforcement learning, this paper demonstrates that agents operating under finite Shannon capacity constraints achieve superior riskadjusted returns by effectively masking microstructure noise. Through high-fidelity simulations of limit order markets, we compare full-information agents with those utilizing selective ignorance filters. Our results indicate that strategic ignorance reduces the propensity for overreaction and herding, leading to higher Sharpe ratios and lower maximum drawdowns, particularly in high-volatility regimes. We find that reinforcement learning agents, trained to dynamically mask noisy features through a Kalman-enhanced framework, learn to prioritize persistent fundamental signals over transient price fluctuations. The findings suggest that “less is more” in financial decision-making; strategic blindness serves not only as a protective heuristic against cognitive and computational overload but also as a source of orthogonal alpha in fragmented, high-frequency environments. This study provides a comprehensive system architecture for implementing selective ignorance in institutional trading and offers a novel perspective on the informational foundations of market stability.
- 25 December 2025
- Published by MDPI AG in Preprints.org
Open AccessStrategic Ignorance in Financial Markets: When Not Knowing Improves Alpha and Stability
The contemporary financial landscape is characterized by a paradox of plenty: while the volume and velocity of market data have increased exponentially, the decision-making quality of human and algorithmic agents often deteriorates under the weight of information overload. This research investigates the concept of strategic ignorance—the deliberate decision to filter, delay, or ignore specific information streams—as a mechanism to enhance alpha generation and market stability. By integrating theories of rational inattention, information design, and multi-agent reinforcement learning, this paper demonstrates that agents operating under finite Shannon capacity constraints achieve superior riskadjusted returns by effectively masking microstructure noise. Through high-fidelity simulations of limit order markets, we compare full-information agents with those utilizing selective ignorance filters. Our results indicate that strategic ignorance reduces the propensity for overreaction and herding, leading to higher Sharpe ratios and lower maximum drawdowns, particularly in high-volatility regimes. We find that reinforcement learning agents, trained to dynamically mask noisy features through a Kalman-enhanced framework, learn to prioritize persistent fundamental signals over transient price fluctuations. The findings suggest that “less is more” in financial decision-making; strategic blindness serves not only as a protective heuristic against cognitive and computational overload but also as a source of orthogonal alpha in fragmented, high-frequency environments. This study provides a comprehensive system architecture for implementing selective ignorance in institutional trading and offers a novel perspective on the informational foundations of market stability.
- Open Access
- 25 December 2025
- Published by MDPI AG in Preprints.org
The Market Control Illusion: A Separated Control and Controllability Framework for Non-Monotonic Prediction–Profit Dynamics in Financial Markets
Modern quantitative trading systems implicitly assume that predictive accuracy implies economic control. This paper proves that assumption is false. We formalize a fundamental separation between predictability—an observer-centric statistical property—and controllability—an actuator-centric dynamical property of trading under feedback, frictions, and competition. Using tools from stochastic control, market microstructure, and performative economics, we establish a negative result: high predictive power is neither sufficient for nor monotone in profitability and can, under realistic regimes of convex execution costs and feedback, strictly accelerate P&L erosion. We model markets as closed-loop dynamical systems in which agents’ actions alter the state conditional on which predictions are made. We characterize exploitable alpha via a controllability condition and show that prediction-optimal policies are generically antioptimal under feedback. Finally, we propose a control-theoretic replacement for prediction-centric trading could specify that this replaces MSE-centric optimization to emphasize the practical takeaway.
- 25 December 2025
- Published by MDPI AG in Preprints.org
Open AccessThe Market Control Illusion: A Separated Control and Controllability Framework for Non-Monotonic Prediction–Profit Dynamics in Financial Markets
Modern quantitative trading systems implicitly assume that predictive accuracy implies economic control. This paper proves that assumption is false. We formalize a fundamental separation between predictability—an observer-centric statistical property—and controllability—an actuator-centric dynamical property of trading under feedback, frictions, and competition. Using tools from stochastic control, market microstructure, and performative economics, we establish a negative result: high predictive power is neither sufficient for nor monotone in profitability and can, under realistic regimes of convex execution costs and feedback, strictly accelerate P&L erosion. We model markets as closed-loop dynamical systems in which agents’ actions alter the state conditional on which predictions are made. We characterize exploitable alpha via a controllability condition and show that prediction-optimal policies are generically antioptimal under feedback. Finally, we propose a control-theoretic replacement for prediction-centric trading could specify that this replaces MSE-centric optimization to emphasize the practical takeaway.
- Open Access
- 23 December 2025
- Published by Springer Nature in Research Square
Markets as Adversarial Control Systems: Stability, Metastability, and Feedback-Induced Instability
Modern electronic markets have undergone a structural transition from human-mediated exchange mechanisms to high-frequency adversarial control systems governed by competing algorithmic feedback laws. While traditional economic theory attributes market dislocations to behavioral biases or exogenous shocks, such explanations fail to account for instabilities arising in environments populated by rational, mathematically optimized agents. This paper reframes financial markets as dynamical plants controlled by heterogeneous, interacting controllers operating under latency, information, and impact constraints. We show that instability is not anomalous but emerges endogenously from optimal feedback operating in nonnormal, metastable state spaces. Introducing the concepts of metastable markets, non-normal transient amplification, and a quantitative Control Fragility Index (CFI), we demonstrate how markets can appear statistically stable while remaining critically sensitive to marginal changes in control gain, synchronization, or delay. We further identify a six-part taxonomy of feedbackinduced instabilities—including reflexive resonance, information gradient collapse, and liquidity phase transitions—that arise without manipulation, deception, or irrational behavior. Finally, we argue that market stability must be treated as a firstclass design constraint and propose design-level regulatory interventions grounded in control theory rather than behavioral restriction. Our results suggest that systemic instability is the natural byproduct of local optimality in adversarial feedback environments and that equilibrium existence alone is insufficient to guarantee dynamical safety.
- 23 December 2025
- Published by Springer Nature in Research Square
Open AccessMarkets as Adversarial Control Systems: Stability, Metastability, and Feedback-Induced Instability
Modern electronic markets have undergone a structural transition from human-mediated exchange mechanisms to high-frequency adversarial control systems governed by competing algorithmic feedback laws. While traditional economic theory attributes market dislocations to behavioral biases or exogenous shocks, such explanations fail to account for instabilities arising in environments populated by rational, mathematically optimized agents. This paper reframes financial markets as dynamical plants controlled by heterogeneous, interacting controllers operating under latency, information, and impact constraints. We show that instability is not anomalous but emerges endogenously from optimal feedback operating in nonnormal, metastable state spaces. Introducing the concepts of metastable markets, non-normal transient amplification, and a quantitative Control Fragility Index (CFI), we demonstrate how markets can appear statistically stable while remaining critically sensitive to marginal changes in control gain, synchronization, or delay. We further identify a six-part taxonomy of feedbackinduced instabilities—including reflexive resonance, information gradient collapse, and liquidity phase transitions—that arise without manipulation, deception, or irrational behavior. Finally, we argue that market stability must be treated as a firstclass design constraint and propose design-level regulatory interventions grounded in control theory rather than behavioral restriction. Our results suggest that systemic instability is the natural byproduct of local optimality in adversarial feedback environments and that equilibrium existence alone is insufficient to guarantee dynamical safety.
- Open Access
- 22 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
EcoLLM: A Joint Optimization Framework for Ultra-Low Power, Mixed-Precision LLM Inference on Resource-Constrained Edge Systems
Deployment of Large Language Models (LLMs) on resource-constrained edge devices, such as IoT systems and mobile AI platforms, is significantly hampered by excessive energy consumption, large memory footprints, and high inference latency. Current methods relying on the sequential application of model compression techniques-namely quantization and pruning-often result in sub-optimal and hardware-unaware configurations. This paper introduces EcoLLM, a novel joint optimization framework designed for achieving ultra-low power LLM inference. EcoLLM systematically determines the optimal layer-wise configurations of mixed-precision bit-widths (bw) and structured sparsity rates (s) guided by a predictive, hardwareaware energy consumption model. The primary technical contributions include: (1) The E-Model, a calibrated analytical Energy Cost Model that rigorously quantifies layer-wise energy consumption by explicitly incorporating the effects of data movement, precision, and sparsity. (2) The E-Metric (Energy Efficiency Metric), a holistic performance index that quantifies the energyaccuracy Pareto frontier by normalizing efficiency (Tokens/Joule) against model degradation (Perplexity loss). (3) The Adaptive Allocation Algorithm (A-Algorithm), an iterative greedy search approach designed to minimize the E-Metric across the entire network while simultaneously mitigating the cumulative effects of reconstruction error propagation. Implemented within the opensource TinyEdgeLLM Library and validated on a reference edge NPU/CPU platform, EcoLLM achieved a 6.4x compression ratio (from 487MB to 76MB) on the GPT-2 architecture with < 2.0% perplexity degradation, resulting in a 45% reduction in power consumption (from 3.8W to 2.1W) on a reference Raspberry Pi 4 platform.
- 22 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessEcoLLM: A Joint Optimization Framework for Ultra-Low Power, Mixed-Precision LLM Inference on Resource-Constrained Edge Systems
Deployment of Large Language Models (LLMs) on resource-constrained edge devices, such as IoT systems and mobile AI platforms, is significantly hampered by excessive energy consumption, large memory footprints, and high inference latency. Current methods relying on the sequential application of model compression techniques-namely quantization and pruning-often result in sub-optimal and hardware-unaware configurations. This paper introduces EcoLLM, a novel joint optimization framework designed for achieving ultra-low power LLM inference. EcoLLM systematically determines the optimal layer-wise configurations of mixed-precision bit-widths (bw) and structured sparsity rates (s) guided by a predictive, hardwareaware energy consumption model. The primary technical contributions include: (1) The E-Model, a calibrated analytical Energy Cost Model that rigorously quantifies layer-wise energy consumption by explicitly incorporating the effects of data movement, precision, and sparsity. (2) The E-Metric (Energy Efficiency Metric), a holistic performance index that quantifies the energyaccuracy Pareto frontier by normalizing efficiency (Tokens/Joule) against model degradation (Perplexity loss). (3) The Adaptive Allocation Algorithm (A-Algorithm), an iterative greedy search approach designed to minimize the E-Metric across the entire network while simultaneously mitigating the cumulative effects of reconstruction error propagation. Implemented within the opensource TinyEdgeLLM Library and validated on a reference edge NPU/CPU platform, EcoLLM achieved a 6.4x compression ratio (from 487MB to 76MB) on the GPT-2 architecture with < 2.0% perplexity degradation, resulting in a 45% reduction in power consumption (from 3.8W to 2.1W) on a reference Raspberry Pi 4 platform.
- Open Access
- 19 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
The Dynamic Hierarchical Cooperative Swarm (DHCS) Algorithm: A Comprehensive Analysis for High-Dimensional Optimization
This study presents the Dynamic Hierarchical Cooperative Swarm (DHCS) algorithm, a novel bio-inspired metaheuristic designed to tackle the challenges of complex, highdimensional optimization problems. Unlike conventional optimization methods, DHCS integrates dynamic clustering, adaptive role assignment, and periodic synchronization to maintain an effective balance between exploration and exploitation. Each agent in the swarm operates autonomously while collaborating with others through hierarchical structures, enabling emergent collective intelligence. To evaluate its performance, DHCS was tested on the 1000dimensional Ackley benchmark function and compared against traditional algorithms such as Particle Swarm Optimization (PSO) and Genetic Algorithms (GA). The results demonstrate that DHCS achieves superior convergence characteristics, robustness against local optima, and reduced computational overhead despite the complexity of its internal mechanisms. Furthermore, DHCS exhibits scalability to ultra-high-dimensional problems and shows promising adaptability to dynamic environments. This paper details the problem domain, formalizes the DHCS algorithmic framework, presents a comprehensive experimental evaluation, and discusses potential applications in industrial optimization, including engineering design, supply chain logistics, machine learning, and financial modeling.
- 19 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessThe Dynamic Hierarchical Cooperative Swarm (DHCS) Algorithm: A Comprehensive Analysis for High-Dimensional Optimization
This study presents the Dynamic Hierarchical Cooperative Swarm (DHCS) algorithm, a novel bio-inspired metaheuristic designed to tackle the challenges of complex, highdimensional optimization problems. Unlike conventional optimization methods, DHCS integrates dynamic clustering, adaptive role assignment, and periodic synchronization to maintain an effective balance between exploration and exploitation. Each agent in the swarm operates autonomously while collaborating with others through hierarchical structures, enabling emergent collective intelligence. To evaluate its performance, DHCS was tested on the 1000dimensional Ackley benchmark function and compared against traditional algorithms such as Particle Swarm Optimization (PSO) and Genetic Algorithms (GA). The results demonstrate that DHCS achieves superior convergence characteristics, robustness against local optima, and reduced computational overhead despite the complexity of its internal mechanisms. Furthermore, DHCS exhibits scalability to ultra-high-dimensional problems and shows promising adaptability to dynamic environments. This paper details the problem domain, formalizes the DHCS algorithmic framework, presents a comprehensive experimental evaluation, and discusses potential applications in industrial optimization, including engineering design, supply chain logistics, machine learning, and financial modeling.
- Open Access
- 15 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
HyperFabric Interconnect (HFI): A Unified, Scalable Communication Fabric for HPC, AI, Quantum, and Neuromorphic Workloads
The evolution of high-performance computing (HPC) interconnects has produced specialized fabrics such as InfiniBand [1], Intel Omni-Path, and NVIDIA NVLink [2], each optimized for distinct workloads. However, the increasing convergence of HPC, AI/ML, quantum, and neuromorphic computing requires a unified communication substrate capable of supporting diverse requirements including ultra-low latency, high bandwidth, collective operations, and adaptive routing. We present HyperFabric Interconnect (HFI), a novel design that combines the strengths of existing interconnects while addressing their scalability and workload-fragmentation limitations. Our evaluation on simulated clusters demonstrates HFI's ability to reduce job completion time (JCT) by up to 30%, improve tail latency consistency by 45%, and sustain efficient scaling across heterogeneous workloads. Beyond simulation, we provide an analytical model and deployment roadmap that highlight HFI's role as a converged interconnect for the exascale and postexascale era.
- 15 September 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in TechRxiv
Open AccessHyperFabric Interconnect (HFI): A Unified, Scalable Communication Fabric for HPC, AI, Quantum, and Neuromorphic Workloads
The evolution of high-performance computing (HPC) interconnects has produced specialized fabrics such as InfiniBand [1], Intel Omni-Path, and NVIDIA NVLink [2], each optimized for distinct workloads. However, the increasing convergence of HPC, AI/ML, quantum, and neuromorphic computing requires a unified communication substrate capable of supporting diverse requirements including ultra-low latency, high bandwidth, collective operations, and adaptive routing. We present HyperFabric Interconnect (HFI), a novel design that combines the strengths of existing interconnects while addressing their scalability and workload-fragmentation limitations. Our evaluation on simulated clusters demonstrates HFI's ability to reduce job completion time (JCT) by up to 30%, improve tail latency consistency by 45%, and sustain efficient scaling across heterogeneous workloads. Beyond simulation, we provide an analytical model and deployment roadmap that highlight HFI's role as a converged interconnect for the exascale and postexascale era.