Searcharxiv⌕ Search

arXiv subjects

Ezgi Sarikayak

Publications and source records attributed to Ezgi Sarikayak.

3 recordsLinked to original sources

Re-Reading Is Not Free: Token Traces and Cache-Aware Energy Estimation for LLM Agents in Software Engineering

As large language models (LLMs) evolve from conversational assistants into autonomous agents embedded in software development workflows, estimating their environmental footprint requires moving beyond conventional token-based accounting. Existing energy estimators focus on isolated request-response interactions, overlooking the repeated processing of accumulated context, cache reuse, and work delegated to subagents. We analyze nearly one billion tokens collected across 42 sessions and 752 task episodes over ten weeks of real-world AI-assisted software development. Cached-context reads make up 94.8% of all API-reported tokens, while uncached input accounts for only 0.017%; for every generated token, approximately 260 cached tokens are read. Context grows 5.3-fold over a session compared with 1.4-fold within individual tasks. To quantify the energy implications, we extend an existing analytical inference-energy model to account for cached-context processing, yielding a cache-aware hybrid estimator. It estimates approximately 250 kWh for the workload, compared with approximately 10 kWh from the output-based estimator EcoLogits, a 25-fold difference largely driven by cached-context reads. Neither estimator has been validated on long cached contexts, but no modeling assumption we varied closes the gap. Starting each task from a fresh context would lower our estimate by about 60%. We also find that parent-session logs report only a subagent's final internal call, undercounting delegated tokens by a factor of ten; for this workload the correction changes the energy estimate by about 5%, but token-volume studies are fully exposed to it. Our findings expose a substantial gap between how current energy estimators and agentic workloads treat context, highlighting the need for workflow-level energy accounting that captures context accumulation, cache reuse, and delegated execution.

cs.SE↗

Measuring and Optimizing Energy Efficiency in ICT: A Multi-vocal Literature Review on Data Centers, Cloud Computing, and LLM-Driven Code Optimization

Rapid growth in Information and Communication Technology (ICT) increases electricity use and environmental impact. Addressing this challenge requires a systematic, software-centric view of how energy and related sustainability attributes are measured, reported, and improved. We present a comprehensive Multi-vocal Literature Review (MLR) integrating peer-reviewed and gray literature to map current practice in measuring and optimizing energy use in ICT software, with emphasis on data centers and cloud computing environments (DCCCEs). First, we synthesize how sustainability is measured across ICT by proposing a taxonomy organized along metric domains (power/energy, energy efficiency, environmental impact) and component granularity (from applications, to containers/VMs, and system-of-systems levels). We summarize measurement approaches and the tooling landscape that implements them. Second, we review optimization strategies at the software and system levels, covering dynamic power and resource management, VM/container consolidation, energy-aware software and system design, and cross-layer, system-wide methods relevant to DCCCEs. Third, we examine how large language models (LLMs) are applied to code generation and transformation for resource-efficient software, noting that most studies target code generation (predominantly in Python) and evaluate runtime and memory, with fewer works directly reporting energy. Hardware- and network-level telemetry are treated as data sources for software-level analysis. We identify gaps in vendor-neutral telemetry, consistent measurement and reporting practices, network-level metrics, and unified, cross-layer optimization frameworks, and outline directions for improving reproducibility and comparability. We screened 635 records (2010-2025) and included 181 sources (122 peer-reviewed, 59 gray).

cs.SE↗

Agent4cs: A Multi-agent System for Code Summarization in Large Hierarchical Codebases

Understanding large, complex codebases, especially those with obfuscated structures and incomplete documentation, remains a significant challenge. Existing code summarization solutions often rely on a single language model or coding assistant like Claude Code, and treat source code as flat text, underutilizing the rich interdependencies and hierarchical information within a repository. To address these shortcomings, we propose Agent4cs - a multi-agent framework that summarizes large codebases in a bottom-up fashion, where a summarization agent focuses on producing robust summaries; a keyword-extraction agent proactively identifies critical information from subfolders; and a quality-assurance agent iteratively refines the outputs for readability, coherence, and completeness. Evaluated on 7 frontier models, Agent4cs improves semantic consistency across all folder levels by average 8% compared to two structured prompting baselines with code segments. Furthermore, extensive evaluation on real-world datasets demonstrates up to 38% gains in normalized keyword coverage rate over the same baselines.

cs.AI↗