arXiv subjects

Harikaran Subbaraj

Publications and source records attributed to Harikaran Subbaraj.

3 recordsLinked to original sources

arXiv2025

Understanding and Steering the Cognitive Behaviors of Reasoning Models at Test-Time

Large Language Models (LLMs) often rely on long chain-of-thought (CoT) reasoning to solve complex tasks. While effective, these trajectories are frequently inefficient, leading to high latency from excessive token generation, or unstable reasoning that alternates between underthinking (shallow, inconsistent steps) and overthinking (repetitive, verbose reasoning). In this work, we study the structure of reasoning trajectories and uncover specialized attention heads that correlate with distinct cognitive behaviors such as verification and backtracking. By lightly intervening on these heads at inference time, we can steer the model away from inefficient modes. Building on this insight, we propose CREST, a training-free method for Cognitive REasoning Steering at Test-time. CREST has two components: (1) an offline calibration step that identifies cognitive heads and derives head-specific steering vectors, and (2) an inference-time procedure that rotates hidden representations to suppress components along those vectors. CREST adaptively suppresses unproductive reasoning behaviors, yielding both higher accuracy and lower computational cost. Across diverse reasoning benchmarks and models, CREST improves accuracy by up to 17.5% while reducing token usage by 37.6%, offering a simple and effective pathway to faster, more reliable LLM reasoning.

cs.CL↗

arXiv2020

Optimizing Prediction Serving on Low-Latency Serverless Dataflow

Prediction serving systems are designed to provide large volumes of low-latency inferences machine learning models. These systems mix data processing and computationally intensive model inference and benefit from multiple heterogeneous processors and distributed computing resources. In this paper, we argue that a familiar dataflow API is well-suited to this latency-sensitive task, and amenable to optimization even with unmodified black-box ML models. We present the design of Cloudflow, a system that provides this API and realizes it on an autoscaling serverless backend. Cloudflow transparently implements performance-critical optimizations including operator fusion and competitive execution. Our evaluation shows that Cloudflow's optimizations yield significant performance improvements on synthetic workloads and that Cloudflow outperforms state-of-the-art prediction serving systems by as much as 2x on real-world prediction pipelines, meeting latency goals of demanding applications like real-time video analysis.

cs.DC↗

arXiv2018

Dynamic Space-Time Scheduling for GPU Inference

Serving deep neural networks in latency critical interactive settings often requires GPU acceleration. However, the small batch sizes typical in online inference results in poor GPU utilization, a potential performance gap which GPU resource sharing can address. In this paper, we explore several techniques to leverage both temporal and spatial multiplexing to improve GPU utilization for deep learning inference workloads. We evaluate the performance trade-offs of each approach with respect to resource-efficiency, latency predictability, and isolation when compared with conventional batched inference. Our experimental analysis suggests up to a 5x potential for improved utilization through the exploration of more advanced spatial and temporal multiplexing strategies. Our preliminary prototype of a dynamic space-time scheduler demonstrates a 3.23x floating-point throughput increase over space-only multiplexing and a 7.73x increase over time-only multiplexing for convolutions, while also providing better isolation and latency predictability.

cs.DC↗