BZPEERReading, trust and collaboration for research
Menu

Discover

GraphFollow connections around a paper, topic or saved view.

Workspaces

HomePick up where your research flow left off.InboxHandle requests, replies and notifications from one place.CollectionsCurate reading lists, evidence sets and shareable dossiers.ResearchSee your private provenance graph, trail and imports.CollaborationMove from discovery into focused discussion rooms.PublishDraft, preview and publish full-text research articles.

Network

ExpertsFind specialists worth following or asking for help.CommunitiesJoin research circles organized around shared work.PartnersSee external organizations active around the same topics.

Opportunities

ListingsBrowse roles, calls and collaborations with clearer fit signals.

Account

Log inReturn to your reading and collaboration workspace.Create accountStart saving work, following people and joining teams.
Log inCreate account

Researcher profile

Mingu Kang

Mingu Kang contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate
Machine LearningArtificial Intelligencecond-mat.str-elHardware ArchitectureMultiagent Systemsphysics.chem-phphysics.opticsComputation and LanguageComputer Visioncond-mat.mtrl-scicond-mat.supr-con

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author
10works
0followers
11topics
4close collaborators

Actions

Decide how to stay connected

Follow researcher0

Identity and collaboration

How to connect with this researcher

Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.

Log in to claim

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

Research graph

See the researcher in context

Open full explorer

Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.

Building this graph slice

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

10 published item(s)

preprint2026arXiv

Interaction-Breaking Adversarial Learning Framework for Robust Multi-Agent Reinforcement Learning

Cooperation is central to multi-agent reinforcement learning (MARL), yet learned coordination can be fragile when external perturbations disrupt inter-agent interactions. Prior robust MARL methods have primarily considered value-oriented attacks, leaving a gap in robustness when interaction structures themselves are corrupted. In this paper, we propose an interaction-breaking adversarial learning (IBAL) framework that takes an information-theoretic view to construct attacks that impede coordination by perturbing agents' observations and actions, and trains agents to perform reliably under such disruptions. Empirically, our approach improves robustness over existing robust MARL baselines across diverse attack settings and yields stronger performance even under agent-missing scenarios.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

Shaping Zero-Shot Coordination via State Blocking

Zero-shot coordination (ZSC) aims to enable agents to cooperate with independently trained partners without prior interaction, a key requirement for real-world multi-agent systems and human-AI collaboration. Existing approaches have largely emphasized increasing partner diversity during training, yet such strategies often fall short of achieving reliable generalization to unseen partners. We introduce State-Blocked Coordination (SBC), a simple yet effective framework that improves ZSC by inducing diverse interaction scenarios without direct environment modification. Specifically, SBC generates a family of virtual environments through state blocking, allowing agents to experience a wide range of suboptimal partner policies. Across multiple benchmarks, SBC demonstrates superior performance in zero-shot coordination, including strong generalization to human partners.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Accelerating Attention through Gradient-Based Learned Runtime Pruning

Self-attention is a key enabler of state-of-art accuracy for various transformer-based Natural Language Processing models. This attention mechanism calculates a correlation score for each word with respect to the other words in a sentence. Commonly, only a small subset of words highly correlates with the word under attention, which is only determined at runtime. As such, a significant amount of computation is inconsequential due to low attention scores and can potentially be pruned. The main challenge is finding the threshold for the scores below which subsequent computation will be inconsequential. Although such a threshold is discrete, this paper formulates its search through a soft differentiable regularizer integrated into the loss function of the training. This formulation piggy backs on the back-propagation training to analytically co-optimize the threshold and the weights simultaneously, striking a formally optimal balance between accuracy and computation pruning. To best utilize this mathematical innovation, we devise a bit-serial architecture, dubbed LeOPArd, for transformer language models with bit-level early termination microarchitectural mechanism. We evaluate our design across 43 back-end tasks for MemN2N, BERT, ALBERT, GPT-2, and Vision transformer models. Post-layout results show that, on average, LeOPArd yields 1.9x and 3.9x speedup and energy reduction, respectively, while keeping the average accuracy virtually intact (<0.2% degradation)

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Conformational heterogeneity of molecules physisorbed on a gold surface at room temperature

A quantitative single-molecule tip-enhanced Raman spectroscopy (TERS) study at room temperature remained a challenge due to the rapid structural dynamics of molecules exposed to air. Here, we demonstrate the hyperspectral TERS imaging of single or a few brilliant cresyl blue (BCB) molecules at room temperature, along with quantitative spectral analyses. Robust chemical imaging is enabled by the freeze-frame approach using a thin Al$_{2}$O$_{3}$ capping layer, which suppresses spectral diffusions and inhibits chemical reactions and contaminations in air. For the molecules resolved spatially in the TERS image, a clear Raman peak variation up to 7.5 cm$^{-1}$ is observed, which cannot be found in molecular ensembles. From density functional theory-based quantitative analyses of the varied TERS peaks, we reveal the conformational heterogeneity at the single-molecule level. This work provides a facile way to investigate the single-molecule properties in interacting media, expanding the scope of single-molecule vibrational spectroscopy studies.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Discovery of an electronic crystal in a cuprate Mott insulator

Copper oxide high temperature superconductors universally exhibit multiple forms of electronically ordered phases that break the native translational symmetry of the CuO2 planes. The interplay between these orders and the superconducting ground state, as well as how they arise through doping a Mott insulator, is essential to decode the mechanisms of high-temperature superconductivity. Over the years, various forms of electronic liquid crystal phases including charge/spin stripes and incommensurate charge-density-waves (CDWs) were found to emerge out of a correlated metallic ground state in underdoped cuprates. Early theoretical studies also predicted the emergence of a Coulomb-frustrated &#39;charge crystal&#39; phase in the very lightly-doped, insulating limit of the CuO2 planes. Here, we use resonant X-ray scattering, electron transport, and muon spin rotation measurements to fully resolve the electronic and magnetic ground state and search for signatures of charge order in very lightly hole-doped cuprates from the RBa2Cu3O7-d family (RBCO; R: Y or rare earth). X-ray scattering data from RBCO films reveal a breaking of translational symmetry more pervasive than was previously known, extending down to the Mott limit. The ordering vector of this charge crystal state is linearly connected to the charge-density-waves of underdoped RBCO, suggesting that the former phase is a precursor to the latter as hole doping is increased. Most importantly, the coexistence of charge and spin order in RBCO suggests that this electronic symmetry-breaking state is common to the CuO2 planes in the very lightly-doped regime. These findings bridge the gap between the Mott insulating state and the underdoped metallic state and underscore the prominent role of Coulomb-frustrated electronic phase separation among all cuprates.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Freeze-frame approach for robust single-molecule tip-enhnaced Raman spectroscopy at room temperature

A quantitative single-molecule tip-enhanced Raman spectroscopy (TERS) study at room temperature remained a challenge due to the rapid structural dynamics of molecules exposed to air. Here, we demonstrate the single-molecule level hyperspectral TERS imaging of brilliant cresyl blue (BCB) at room temperature for the first time, along with quantitative spectral analyses. Freeze-frame approach using a thin Al2O3 capping layer, which suppresses spectral diffusions and inhibits chemical reactions and contaminations in air, enabled reliable and robust chemical imaging. For the molecules resolved spatially in the TERS image, a clear Raman peak variation up to 7.5 cm-1 is observed, which cannot be found in molecular ensembles. From density functional theory-based quantitative analyses of the varied TERS peaks, we reveal the conformational heterogeneity at the single-molecule level. This work provides a facile way to investigate the single-molecule properties in interacting media, expanding the scope of single-molecule vibrational spectroscopy.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Sparse Attention Acceleration with Synergistic In-Memory Pruning and On-Chip Recomputation

As its core computation, a self-attention mechanism gauges pairwise correlations across the entire input sequence. Despite favorable performance, calculating pairwise correlations is prohibitively costly. While recent work has shown the benefits of runtime pruning of elements with low attention scores, the quadratic complexity of self-attention mechanisms and their on-chip memory capacity demands are overlooked. This work addresses these constraints by architecting an accelerator, called SPRINT, which leverages the inherent parallelism of ReRAM crossbar arrays to compute attention scores in an approximate manner. Our design prunes the low attention scores using a lightweight analog thresholding circuitry within ReRAM, enabling SPRINT to fetch only a small subset of relevant data to on-chip memory. To mitigate potential negative repercussions for model accuracy, SPRINT re-computes the attention scores for the few fetched data in digital. The combined in-memory pruning and on-chip recompute of the relevant attention scores enables SPRINT to transform quadratic complexity to a merely linear one. In addition, we identify and leverage a dynamic spatial locality between the adjacent attention operations even after pruning, which eliminates costly yet redundant data fetches. We evaluate our proposed technique on a wide range of state-of-the-art transformer models. On average, SPRINT yields 7.5x speedup and 19.6x energy reduction when total 16KB on-chip memory is used, while virtually on par with iso-accuracy of the baseline models (on average 0.36% degradation).

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Applying Tensor Decomposition to image for Robustness against Adversarial Attack

Nowadays the deep learning technology is growing faster and shows dramatic performance in computer vision areas. However, it turns out a deep learning based model is highly vulnerable to some small perturbation called an adversarial attack. It can easily fool the deep learning model by adding small perturbations. On the other hand, tensor decomposition method widely uses for compressing the tensor data, including data matrix, image, etc. In this paper, we suggest combining tensor decomposition for defending the model against adversarial example. We verify this idea is simple and effective to resist adversarial attack. In addition, this method rarely degrades the original performance of clean data. We experiment on MNIST, CIFAR10 and ImageNet data and show our method robust on state-of-the-art attack methods.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Radiative control of localized excitons at room temperature with an ultracompact tip-enhanced plasmonic nano-cavity

In atomically thin semiconductors, localized exciton (X$_L$) coupled to light shows single quantum emitting behaviors through radiative relaxation processes providing a new class of optical sources for potential applications in quantum communication. In most studies, however, X$_L$ photoluminescence (PL) from crystal defects has mainly been observed in cryogenic conditions because of their sub-wavelength emission region and low quantum yield at room temperature. Furthermore, engineering the radiative relaxation properties, e.g., emission region, intensity, and energy, remained challenging. Here, we present a plasmonic antenna with a triple-sharp-tips geometry to induce and control the X$_L$ emission of a WSe$_2$ monolayer (ML) at room temperature. By placing a ML crystal on the two sharp Au tips in a bowtie antenna fabricated through cascade domino lithography with a radius of curvature of <1 nm, we effectively induce tensile strain in the nanoscale region to create robust X$_L$ states. An Au tip with tip-enhanced photoluminescence (TEPL) spectroscopy is then added to the strained region to probe and control the X$_L$ emission. With TEPL enhancement of X$_L$ as high as ~10$^6$ in the triple-sharp-tips device, experimental results demonstrate the controllable X$_L$ emission in <30 nm area with a PL energy shift up to 40 meV, resolved by tip-enhanced PL and Raman imaging with <15 nm spatial resolution. Our approach provides a systematic way to control localized quantum light in 2D semiconductors offering new strategies for active quantum nano-optical devices.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Topological flat bands in frustrated kagome lattice CoSn

Electronic flat bands in momentum space, arising from strong localization of electrons in real space, are an ideal stage to realize strong correlation phenomena. In certain lattices with built-in geometrical frustration, electronic confinement and flat bands can naturally arise from the destructive interference of electronic hopping pathways. Such lattice-borne flat bands are often endowed with nontrivial topology if combined with spin-orbit coupling, while their experimental realization in condensed matter system has been elusive so far. Here, we report the direct observation of topological flat bands in the vicinity of the Fermi level in frustrated kagome system CoSn, using angle-resolved photoemission spectroscopy and band structure calculations. The flat band manifests itself as a dispersionless electronic excitation along the G-M high symmetry direction, with an order of magnitude lower bandwidth (below 150 meV) compared to the Dirac bands originating from the same orbitals. The frustration-driven nature of the flat band is directly confirmed by the real-space chiral d-orbital texture of the corresponding effective Wannier wave functions. Spin-orbit coupling opens a large gap of 80 meV at the quadratic band touching point between the Dirac and flat bands, endowing a nonzero Z2 topological invariant to the flat band in the two-dimensional Brillouin zone. Our observation of lattice-driven topological flat band opens a promising route to engineer novel emergent phases of matter at the crossroad between strong correlation physics and electronic topology.

0 citations0 reviewsNo code linkedOpen access