arXiv subjects

Sunghyun Choi

Publications and source records attributed to Sunghyun Choi.

5 recordsLinked to original sources

arXiv2026

Long Live the Librarian! A Persistent Search Sub-Agent for Energy-Efficient Multi-Agent Software Engineering Systems

Multi-agent systems (MAS) have substantially advanced autonomous software engineering (SWE), but their growing inference energy demands raise sustainability concerns. In this paper, we demonstrate that this cost is concentrated in an overlooked source: redundant output tokens generated across agents. Two empirical findings ground this claim. First, our per-token energy attribution for MAS reveals a sharp asymmetry: an output token consumes 30 to 1,000 times more energy than an input or cached token. Second, MAS inflate per-episode output because agents repeatedly re-explore overlapping repository regions. To address this inefficiency, we propose Librarian, a persistent search sub-agent that tracks repository-search history and suppresses redundant exploration actions across agents. By returning short references to file regions instead of full file excerpts, Librarian further reduces output-token volume. On SWE-Bench Verified and SWE-Bench-Live, Librarian reduces per-episode GPU energy consumption of existing multi-agent SWE systems by 11 to 30% while preserving task performance. Our code is available at https://github.com/ml-postech/Librarian.

cs.MA↗

arXiv2025

The Oversmoothing Fallacy: A Misguided Narrative in GNN Research

Oversmoothing has been recognized as a main obstacle to building deep Graph Neural Networks (GNNs), limiting the performance. This position paper argues that the influence of oversmoothing has been overstated and advocates for a further exploration of deep GNN architectures. Given the three core operations of GNNs, aggregation, linear transformation, and non-linear activation, we show that prior studies have mistakenly confused oversmoothing with the vanishing gradient, caused by transformation and activation rather than aggregation. Our finding challenges prior beliefs about oversmoothing being unique to GNNs. Furthermore, we demonstrate that classical solutions such as skip connections and normalization enable the successful stacking of deep GNN layers without performance degradation. Our results clarify misconceptions about oversmoothing and shed new light on the potential of deep GNNs.

cs.LG↗

arXiv2022

RAPID: Contention Resolution-based Random Access using Context ID for IoT

With the increasing number of Internet of Things (IoT) devices, Machine Type Communication (MTC) has become an important use case of the Fifth Generation (5G) communication systems. Since MTC devices are mostly disconnected from Base Station (BS) for power saving, random access procedure is required for devices to transmit data. If many devices try random access simultaneously, preamble collision problem occurs, thus causing latency increase. In an environment where delay-sensitive and delay-tolerant devices coexist, the contention-based random access procedure cannot satisfy latency requirements of delay-sensitive devices. Therefore, we propose RAPID, a novel random access procedure, which is completed through two message exchanges for the delay-sensitive devices. We also develop Access Pattern Analyzer (APA), which estimates traffic characteristics of MTC devices. When UEs, performing RAPID and contention-based random access, coexist, it is important to determine a value which is the number of preambles for RAPID to reduce random access load. Thus, we analyze random access load using a Markov chain model to obtain the optimal number of preambles for RAPID. Simulation results show RAPID achieves 99.999% reliability with 80.8% shorter uplink latency, and also decreases random access load by 30.5% compared with state-of-the-art techniques.

cs.NI↗

arXiv2019

Ultrareliable and Low-Latency Communication Techniques for Tactile Internet Services

This paper presents novel ultrareliable and low-latency communication (URLLC) techniques for URLLC services, such as Tactile Internet services. Among typical use-cases of URLLC services are tele-operation, immersive virtual reality, cooperative automated driving, and so on. In such URLLC services, new kinds of traffic such as haptic information including kinesthetic information and tactile information need to be delivered in addition to high-quality video and audio traffic in traditional multimedia services. Further, such a variety of traffic has various characteristics in terms of packet sizes and data rates with a variety of requirements of latency and reliability. Furthermore, some traffic may occur in a sporadic manner but require reliable delivery of packets of medium to large sizes within a low latency, which is not supported by current state-of-the-art wireless communication systems and is very challenging for future wireless communication systems. Thus, to meet such a variety of tight traffic requirements in a wireless communication system, novel technologies from the physical layer to the network layer need to be devised. In this paper, some novel physical layer technologies such as waveform multiplexing, multiple access scheme, channel code design, synchronization, and full-duplex transmission for spectrally-efficient URLLC are introduced. In addition, a novel performance evaluation approach, which combines a ray-tracing tool and system-level simulation, is suggested for evaluating the performance of the proposed schemes. Simulation results show the feasibility of the proposed schemes providing realistic URLLC services in realistic geographical environments, which encourages further efforts to substantiate the proposed work.

cs.IT↗

arXiv2014

Load Balancing in Two-Tier Cellular Networks with Open and Hybrid Access Femtocells

Femtocell base station (BS) is a low-power, low-price BS based on cellular communication technology. It is expected to become a cost-effective solution for improving the communication performance of indoor users, whose traffic demands are large in general. There are mainly three access strategies for femtocell, i.e., closed access, open access and hybrid access strategies. While it has been generally known that open/hybrid access femtocells contribute more to enhancing the system-wide performance than closed access femtocells, the operating parameters of both macro and femtocells should be carefully chosen according to the mobile operator's policy, consumer's requirements, and so on. We propose long-term parameter optimization schemes, which maximize the average throughput of macrocell users while guaranteeing some degree of benefits to femtocell owners. To achieve this goal, we jointly optimize the ratio of dedicated resources for femtocells as well as the femtocell service area in open access femtocell networks through the numerical analysis. It is proved that the optimal parameter selection of open access femtocell is a convex optimization problem in typical environments. Then, we extend our algorithm to hybrid access femtocells where some intra-femtocell resources are dedicated only for femtocell owners while remaining resources are shared with foreign macrocell users. Our evaluation results show that the proposed parameter optimization schemes significantly enhance the performance of macrocell users thanks to the large offloading gain. The benefits provided to femtocell users are also adaptively maintained according to the femtocell users' requirements. The results in this paper provide insights about the situations where femtocell deployment on dedicated channels is preferred to the co-channel deployment.

cs.NI↗