13 papers · 1 filter
CyberWorld: World Models for Sample-Efficient Autonomous Cyber Defense
Ryozo Masukawa, Sanggeon Yun, Raheeb Hassan +3
Deep reinforcement learning has become a prominent approach to autonomous cyber defense. Existing methods are predominantly model-free and consequently require extensive environmen…
Asymmetric Capacity Allocation in Self-Refinement Pipelines
Zhuoyi Yang, Ian G. Harris, Salar Hashemitaheri +7
Self-refinement, typically structured as generation, critique, and revision, is a widely adopted paradigm for improving LLM generation and serves as a core mechanism in many LLM ag…
Vector Symbolic Policy Gradient
Ryozo Masukawa, Sanggeon Yun, SungHeon Jeong +6
We answer this question with Vector-Symbolic Policy Gradient (VSPG), a discrete-action actor that represents each action by a unit-norm hypervector and scores it by similarity to t…
PolyQ: Codesigning End-to-End Quantization Framework for Scalable Edge CPU LLM Inference
Hyunwoo Oh, Suyeon Jang, Hanning Chen +4
CPUs are the most universal target for on-device LLM inference, but existing low-bit quantization methods offer either coarse operating points or fine-grained mixed precision that…
Qubit-Efficient Quantum Search for Hyperdimensional Decomposition via Logarithmic Encoding
Sanggeon Yun, Hyunwoo Oh, Ryozo Masukawa +2
Hyperdimensional Computing (HDC) represents symbols using high-dimensional hypervectors of dimension . In hypervector decomposition, the objective is to recover constituent…
FusionSense: Tri-Stage Near-Sensor Learning for Runtime-Adaptive Multimodal Edge Intelligence
Sanggeon Yun, Ryozo Masukawa, Minhyoung Na +5
Autonomous systems and smart-industry deployments increasingly split computation across near-sensor, edge, and cloud resources, where tight energy, latency, and reliability budgets…
-Musketeers: Reinforcement Learning Shapes Collaboration Among Language Models
Ryozo Masukawa, Sanggeon Yun, Hyunwoo Oh +8
Recent progress in reinforcement learning with verifiable rewards (RLVR) shows that small, specialized language models (SLMs) can exhibit structured reasoning without relying on la…
HopFormer: Sparse Graph Transformers with Explicit Receptive Field Control
Sanggeon Yun, Raheeb Hassan, Ryozo Masukawa +2
Graph Transformers typically rely on explicit positional or structural encodings and dense global attention to incorporate graph topology. In this work, we show that neither is ess…
Internal Flow Signatures for Self-Checking and Refinement in LLMs
Sungheon Jeong, Sanggeon Yun, Ryozo Masukawa +3
Large language models can generate fluent answers that are unfaithful to the provided context, while many safeguards rely on external verification or a separate judge after generat…
LogHD: Robust Compression of Hyperdimensional Classifiers via Logarithmic Class-Axis Reduction
Sanggeon Yun, Hyunwoo Oh, Ryozo Masukawa +3
Hyperdimensional computing (HDC) suits memory, energy, and reliability-constrained systems, yet the standard "one prototype per class" design requires memory (with clas…
DecoHD: Decomposed Hyperdimensional Classification under Extreme Memory Budgets
Sanggeon Yun, Hyunwoo Oh, Ryozo Masukawa +1
Decomposition is a proven way to shrink deep networks without changing input-output dimensionality or interface semantics. We bring this idea to hyperdimensional computing (HDC), w…
MissionHD: Hyperdimensional Refinement of Distribution-Deficient Reasoning Graphs for Video Anomaly Detection
Sanggeon Yun, Raheeb Hassan, Ryozo Masukawa +2
LLM-generated reasoning graphs, referred to as mission-specific graphs (MSGs), are increasingly used for video anomaly detection (VAD) and recognition (VAR). However, they are typi…
A Few Large Shifts: Layer-Inconsistency Based Minimal Overhead Adversarial Example Detection
Sanggeon Yun, Ryozo Masukawa, Hyunwoo Oh +2
Deep neural networks (DNNs) are highly susceptible to adversarial examples---small, malicious perturbations that can cause incorrect predictions. We introduce a lightweight, plug-i…