arXiv subjects

Huihuo Zheng

Publications and source records attributed to Huihuo Zheng.

17 recordsLinked to original sources

arXiv2026

ECAS: An Edge-Controlled Agentic System for Validation-Gated Scientific Application Execution

Scientific applications increasingly rely on high-performance computing (HPC), yet translating a scientist's high-level goal into a correct target-scale execution remains brittle and labor-intensive. Large language model (LLM) agents promise to automate this, but two obstacles remain: granting a cloud-hosted model direct HPC access exposes credentials and execution authority, while withholding it demands continuous human supervision; and one-shot generation cannot adapt when generated artifacts fail in a site-specific HPC environment. We present \textsc{ECAS}, an \textbf{E}dge-\textbf{C}ontrolled \textbf{A}gentic \textbf{S}ystem for closed-loop execution of scientific computing campaigns with limited human intervention. \textsc{ECAS} separates \emph{reasoning}, \emph{control}, and \emph{execution}: a cloud-hosted LLM proposes plans, artifacts, and repairs; a user-controlled edge agent retains credentials, workflow state, and execution authority while enforcing policy and resource constraints; and the HPC system computes. Its core mechanism is \emph{validation-gated execution}: generated artifacts pass static checks and small-scale validation, failures trigger repairs from sanitized execution feedback, and target-scale execution is permitted only after validation and policy checks pass. \textsc{ECAS} also draws on an edge-resident library of expert-distilled, site-specific skills that is never disclosed to the cloud. In preliminary experiments with three scientific applications on two production ALCF systems under six injected fault types, closed-loop repair improves application success from 0/6 to 6/6 over one-shot generation, validation gating prevents all three observed target-scale failures, and skill conditioning improves success from 4/6 to 6/6. These results show the feasibility of delegating adaptive reasoning to the cloud while retaining execution control at the edge.

cs.DC↗

arXiv2026

LLMs Can Predict Failure Risk, But Struggle to Predict Which Collaboration Protocol Pays Off: Cost-Aware Protocol Routing Across Reasoning Tasks

Multi-agent large language model (LLM) systems can improve reasoning by spending more computation, but deployment requires deciding when extra collaboration is worth its cost. We isolate this decision by running every problem under four protocols while holding the solver fixed within each setting: direct solving (Baseline), iterative self-correction (Single), planner-executor-reviewer collaboration (PER), and multi-agent deliberation (Broadcast). The primary benchmark comprises 4,181 competition-level math problems; paired robustness checks cover four benchmarks spanning competition math, biology, and broader science with two solver families. Across fixed policies, trained routers, and frozen LLM routers, conservative policies under-escalate, whereas higher-solve frozen routers often over-escalate. A post-answer, pre-collaboration gpt-oss-120b probe ranks Baseline failures with 0.8847 AUROC (4,151 parseable cases; 95% CI [0.8732, 0.8955]). The same score remains informative for predicting whether any collaboration helps (0.7683 AUPRC), but is much weaker for identifying PER- or Broadcast-specific value (0.1674 and 0.1041 AUPRC). Separately, the pre-answer self-confidence gate reaches 78.0% solve at 45K tokens, compared with 73.8% at 71.3K for a frozen gpt-oss-120b router and 92.4% for a retrospective fixed-order oracle. Across 10 paired model-condition settings, the oracle adds 23.2-58.3 points of retrospective coverage over Baseline, but protocol profiles vary by task. In the six settings with held-out router evaluations, oracle gaps remain 18.5-28.9 points. Confidence can therefore support initial escalation, while protocol-specific cost-aware routing remains unresolved.

cs.AI↗

arXiv2026

Precise but Uncoupled: Reviewer Precision Does Not Guarantee Critique Uptake in Multi-Agent Math Reasoning

Many math- and science-oriented agent systems use hierarchical designs with specialized reviewer roles, assuming that a dedicated review stage should help turn wrong candidates into correct ones. We test this assumption on 4,181 verifier-grounded Omni-MATH problems using matched gpt-oss-120b actors. Collaboration adds little on the easiest tiers, but from tier 4 onward the gains open sharply; in this harder regime, broadcast-style peer discussion reaches higher final accuracy than a planner-executor-reviewer pipeline (PER). We ask whether this gap is explained by reviewer quality or by whether critique changes the next answer the protocol carries forward. It is not explained by reviewer precision alone: PER's reviewer is more precise than broadcast's (0.861 vs. 0.644), yet evaluator-verified useful critique is much less likely to change the next candidate and produces lower reviewer-guided repair. These results show that reviewer detection quality and critique uptake are empirically separable. Within matched PER interventions, forcing explicit acknowledgment lowers final accuracy, while embedding reviewer guidance directly in the solver's working context partially improves follow-through without closing the gap. Overall, reviewer-centric evaluation can overstate system quality: a protocol may spot errors well yet still fail to solve more problems if it does not act on those critiques.

cs.AI↗

arXiv2025

AERIS: Argonne Earth Systems Model for Reliable and Skillful Predictions

Generative machine learning offers new opportunities to better understand complex Earth system dynamics. Recent diffusion-based methods address spectral biases and improve ensemble calibration in weather forecasting compared to deterministic methods, yet have so far proven difficult to scale stably at high resolutions. We introduce AERIS, a 1.3 to 80B parameter pixel-level Swin diffusion transformer to address this gap, and SWiPe, a generalizable technique that composes window parallelism with sequence and pipeline parallelism to shard window-based transformers without added communication cost or increased global batch size. On Aurora (10,080 nodes), AERIS sustains 10.21 ExaFLOPS (mixed precision) and a peak performance of 11.21 ExaFLOPS with $1 \times 1$ patch size on the 0.25° ERA5 dataset, achieving 95.5% weak scaling efficiency, and 81.6% strong scaling efficiency. AERIS outperforms the IFS ENS and remains stable on seasonal scales to 90 days, highlighting the potential of billion-parameter diffusion models for weather and climate prediction.

cs.LG↗

arXiv2025

HiPerRAG: High-Performance Retrieval Augmented Generation for Scientific Insights

The volume of scientific literature is growing exponentially, leading to underutilized discoveries, duplicated efforts, and limited cross-disciplinary collaboration. Retrieval Augmented Generation (RAG) offers a way to assist scientists by improving the factuality of Large Language Models (LLMs) in processing this influx of information. However, scaling RAG to handle millions of articles introduces significant challenges, including the high computational costs associated with parsing documents and embedding scientific knowledge, as well as the algorithmic complexity of aligning these representations with the nuanced semantics of scientific content. To address these issues, we introduce HiPerRAG, a RAG workflow powered by high performance computing (HPC) to index and retrieve knowledge from more than 3.6 million scientific articles. At its core are Oreo, a high-throughput model for multimodal document parsing, and ColTrast, a query-aware encoder fine-tuning algorithm that enhances retrieval accuracy by using contrastive learning and late-interaction techniques. HiPerRAG delivers robust performance on existing scientific question answering benchmarks and two new benchmarks introduced in this work, achieving 90% accuracy on SciQ and 76% on PubMedQA-outperforming both domain-specific models like PubMedGPT and commercial LLMs such as GPT-4. Scaling to thousands of GPUs on the Polaris, Sunspot, and Frontier supercomputers, HiPerRAG delivers million document-scale RAG workflows for unifying scientific knowledge and fostering interdisciplinary innovation.

cs.IR↗

arXiv2024

Imaging 3D Chemistry at 1 nm Resolution with Fused Multi-Modal Electron Tomography

Measuring the three-dimensional (3D) distribution of chemistry in nanoscale matter is a longstanding challenge for metrological science. The inelastic scattering events required for 3D chemical imaging are too rare, requiring high beam exposure that destroys the specimen before an experiment completes. Even larger doses are required to achieve high resolution. Thus, chemical mapping in 3D has been unachievable except at lower resolution with the most radiation-hard materials. Here, high-resolution 3D chemical imaging is achieved near or below one nanometer resolution in a Au-Fe$_3$O$_4$ metamaterial, Co$_3$O$_4$ - Mn$_3$O$_4$ core-shell nanocrystals, and ZnS-Cu$_{0.64}$S$_{0.36}$ nanomaterial using fused multi-modal electron tomography. Multi-modal data fusion enables high-resolution chemical tomography often with 99\% less dose by linking information encoded within both elastic (HAADF) and inelastic (EDX / EELS) signals. Now sub-nanometer 3D resolution of chemistry is measurable for a broad class of geometrically and compositionally complex materials.

physics.comp-ph↗

arXiv2024

Physics-inspired spatiotemporal-graph AI ensemble for the detection of higher order wave mode signals of spinning binary black hole mergers

We present a new class of AI models for the detection of quasi-circular, spinning, non-precessing binary black hole mergers whose waveforms include the higher order gravitational wave modes $(l, |m|)=\{(2, 2), (2, 1), (3, 3), (3, 2), (4, 4)\}$, and mode mixing effects in the $l = 3, |m| = 2$ harmonics. These AI models combine hybrid dilated convolution neural networks to accurately model both short- and long-range temporal sequential information of gravitational waves; and graph neural networks to capture spatial correlations among gravitational wave observatories to consistently describe and identify the presence of a signal in a three detector network encompassing the Advanced LIGO and Virgo detectors. We first trained these spatiotemporal-graph AI models using synthetic noise, using 1.2 million modeled waveforms to densely sample this signal manifold, within 1.7 hours using 256 A100 GPUs in the Polaris supercomputer at the ALCF. Our distributed training approach had optimal performance, and strong scaling up to 512 A100 GPUs. With these AI ensembles we processed data from a three detector network, and found that an ensemble of 4 AI models achieves state-of-the-art performance for signal detection, and reports two misclassifications for every decade of searched data. We distributed AI inference over 128 GPUs in the Polaris supercomputer and 128 nodes in the Theta supercomputer, and completed the processing of a decade of gravitational wave data from a three detector network within 3.5 hours. Finally, we fine-tuned these AI ensembles to process the entire month of February 2020, which is part of the O3b LIGO/Virgo observation run, and found 6 gravitational waves, concurrently identified in Advanced LIGO and Advanced Virgo data, and zero false positives. This analysis was completed in one hour using one A100 GPU.

astro-ph.IM↗

arXiv2023

AI ensemble for signal detection of higher order gravitational wave modes of quasi-circular, spinning, non-precessing binary black hole mergers

We introduce spatiotemporal-graph models that concurrently process data from the twin advanced LIGO detectors and the advanced Virgo detector. We trained these AI classifiers with 2.4 million IMRPhenomXPHM waveforms that describe quasi-circular, spinning, non-precessing binary black hole mergers with component masses $m_{\{1,2\}}\in[3M_\odot, 50 M_\odot]$, and individual spins $s^z_{\{1,2\}}\in[-0.9, 0.9]$; and which include the $(\ell, |m|) = \{(2, 2), (2, 1), (3, 3), (3, 2), (4, 4)\}$ modes, and mode mixing effects in the $\ell = 3, |m| = 2$ harmonics. We trained these AI classifiers within 22 hours using distributed training over 96 NVIDIA V100 GPUs in the Summit supercomputer. We then used transfer learning to create AI predictors that estimate the total mass of potential binary black holes identified by all AI classifiers in the ensemble. We used this ensemble, 3 classifiers for signal detection and 2 total mass predictors, to process a year-long test set in which we injected 300,000 signals. This year-long test set was processed within 5.19 minutes using 1024 NVIDIA A100 GPUs in the Polaris supercomputer (for AI inference) and 128 CPU nodes in the ThetaKNL supercomputer (for post-processing of noise triggers), housed at the Argonne Leadership Computing Facility. These studies indicate that our AI ensemble provides state-of-the-art signal detection accuracy, and reports 2 misclassifications for every year of searched data. This is the first AI ensemble designed to search for and find higher order gravitational wave mode signals.

astro-ph.IM↗

arXiv2022

Inference-optimized AI and high performance computing for gravitational wave detection at scale

We introduce an ensemble of artificial intelligence models for gravitational wave detection that we trained in the Summit supercomputer using 32 nodes, equivalent to 192 NVIDIA V100 GPUs, within 2 hours. Once fully trained, we optimized these models for accelerated inference using NVIDIA TensorRT. We deployed our inference-optimized AI ensemble in the ThetaGPU supercomputer at Argonne Leadership Computer Facility to conduct distributed inference. Using the entire ThetaGPU supercomputer, consisting of 20 nodes each of which has 8 NVIDIA A100 Tensor Core GPUs and 2 AMD Rome CPUs, our NVIDIA TensorRT-optimized AI ensemble processed an entire month of advanced LIGO data (including Hanford and Livingston data streams) within 50 seconds. Our inference-optimized AI ensemble retains the same sensitivity of traditional AI models, namely, it identifies all known binary black hole mergers previously identified in this advanced LIGO dataset and reports no misclassifications, while also providing a 3X inference speedup compared to traditional artificial intelligence models. We used time slides to quantify the performance of our AI ensemble to process up to 5 years worth of advanced LIGO data. In this synthetically enhanced dataset, our AI ensemble reports an average of one misclassification for every month of searched advanced LIGO data. We also present the receiver operating characteristic curve of our AI ensemble using this 5 year long advanced LIGO dataset. This approach provides the required tools to conduct accelerated, AI-driven gravitational wave detection at scale.

gr-qc↗

arXiv2022

Interpretable AI forecasting for numerical relativity waveforms of quasi-circular, spinning, non-precessing binary black hole mergers

We present a deep-learning artificial intelligence model that is capable of learning and forecasting the late-inspiral, merger and ringdown of numerical relativity waveforms that describe quasi-circular, spinning, non-precessing binary black hole mergers. We used the NRHybSur3dq8 surrogate model to produce train, validation and test sets of $\ell=|m|=2$ waveforms that cover the parameter space of binary black hole mergers with mass-ratios $q\leq8$ and individual spins $|s^z_{\{1,2\}}| \leq 0.8$. These waveforms cover the time range $t\in[-5000\textrm{M}, 130\textrm{M}]$, where $t=0M$ marks the merger event, defined as the maximum value of the waveform amplitude. We harnessed the ThetaGPU supercomputer at the Argonne Leadership Computing Facility to train our AI model using a training set of 1.5 million waveforms. We used 16 NVIDIA DGX A100 nodes, each consisting of 8 NVIDIA A100 Tensor Core GPUs and 2 AMD Rome CPUs, to fully train our model within 3.5 hours. Our findings show that artificial intelligence can accurately forecast the dynamical evolution of numerical relativity waveforms in the time range $t\in[-100\textrm{M}, 130\textrm{M}]$. Sampling a test set of 190,000 waveforms, we find that the average overlap between target and predicted waveforms is $\gtrsim99\%$ over the entire parameter space under consideration. We also combined scientific visualization and accelerated computing to identify what components of our model take in knowledge from the early and late-time waveform evolution to accurately forecast the latter part of numerical relativity waveforms. This work aims to accelerate the creation of scalable, computationally efficient and interpretable artificial intelligence models for gravitational wave astrophysics.

gr-qc↗

arXiv2020

Dynamic Compressed Sensing for Real-Time Tomographic Reconstruction

Electron tomography has achieved higher resolution and quality at reduced doses with recent advances in compressed sensing. Compressed sensing (CS) theory exploits the inherent sparse signal structure to efficiently reconstruct three-dimensional (3D) volumes at the nanoscale from undersampled measurements. However, the process bottlenecks 3D reconstruction with computation times that run from hours to days. Here we demonstrate a framework for dynamic compressed sensing that produces a 3D specimen structure that updates in real-time as new specimen projections are collected. Researchers can begin interpreting 3D specimens as data is collected to facilitate high-throughput and interactive analysis. Using scanning transmission electron microscopy (STEM), we show that dynamic compressed sensing accelerates the convergence speed by 3-fold while also reducing its error by 27% for an Au/SrTiO3 nanoparticle specimen. Before a tomography experiment is completed, the 3D tomogram has interpretable structure within 33% of completion and fine details are visible as early as 66%. Upon completion of an experiment, a high-fidelity 3D visualization is produced without further delay. Additionally, reconstruction parameters that tune data fidelity can be manipulated throughout the computation without rerunning the entire process.

physics.comp-ph↗

arXiv2019

Deep Learning at Scale for the Construction of Galaxy Catalogs in the Dark Energy Survey

The scale of ongoing and future electromagnetic surveys pose formidable challenges to classify astronomical objects. Pioneering efforts on this front include citizen science campaigns adopted by the Sloan Digital Sky Survey (SDSS). SDSS datasets have been recently used to train neural network models to classify galaxies in the Dark Energy Survey (DES) that overlap the footprint of both surveys. Herein, we demonstrate that knowledge from deep learning algorithms, pre-trained with real-object images, can be transferred to classify galaxies that overlap both SDSS and DES surveys, achieving state-of-the-art accuracy $\gtrsim99.6\%$. We demonstrate that this process can be completed within just eight minutes using distributed training. While this represents a significant step towards the classification of DES galaxies that overlap previous surveys, we need to initiate the characterization of unlabelled DES galaxies in new regions of parameter space. To accelerate this program, we use our neural network classifier to label over ten thousand unlabelled DES galaxies, which do not overlap previous surveys. Furthermore, we use our neural network model as a feature extractor for unsupervised clustering and find that unlabeled DES images can be grouped together in two distinct galaxy classes based on their morphology, which provides a heuristic check that the learning is successfully transferred to the classification of unlabelled DES images. We conclude by showing that these newly labeled datasets can be combined with unsupervised recursive training to create large-scale DES galaxy catalogs in preparation for the Large Synoptic Survey Telescope era.

astro-ph.IM↗

arXiv2019

Dielectric dependent hybrid functionals for heterogeneous materials

We derive a dielectric-dependent hybrid functional which accurately describes the electronic properties of heterogeneous interfaces and surfaces, as well as those of three- and two-dimensional bulk solids. The functional, which does not contain any adjustable parameter, is a generalization of self-consistent hybrid functionals introduced for homogeneous solids, where the screened Coulomb interaction is defined using a spatially varying, local dielectric function. The latter is determined self-consistently using density functional calculations in finite electric fields. We present results for the band gaps and dielectric constants of 3D and 2D bulk materials, and band offsets for interfaces, showing an accuracy comparable to that of GW calculations.

cond-mat.mtrl-sci↗

arXiv2017

From real materials to model Hamiltonians with density matrix downfolding

Due to advances in computer hardware and new algorithms, it is now possible to perform highly accurate many-body simulations of realistic materials with all their intrinsic complications. The success of these simulations leaves us with a conundrum: how do we extract useful physical models and insight from these simulations? In this article, we present a formal theory of downfolding--extracting an effective Hamiltonian from first-principles calculations. The theory maps the downfolding problem into fitting information derived from wave functions sampled from a low-energy subspace of the full Hilbert space. Since this fitting process most commonly uses reduced density matrices, we term it density matrix downfolding (DMD).

cond-mat.str-el↗

arXiv2017

The importance of sigma bonding electrons for the accurate description of electron correlation in graphene

Electron correlation in graphene is unique because of the interplay of the Dirac cone dispersion of $π$ electrons with long range Coulomb interaction. The random phase approximation predicts no metallic screening at long distance and low energy because of the zero density of states at Fermi level, so one might expect that graphene should be a poorly screened system. However, empirically graphene is a weakly interacting semimetal, which leads to the question of how electron correlations take place in graphene at different length scales. We address this question by computing the equal time and dynamic structure factor $S(\vec q)$ and $S(\vec q, ω)$ of freestanding graphene using {\it ab-initio} fixed-node diffusion Monte Carlo and the random phase approximation. We find that the $σ$ electrons contribute strongly to $S(\vec q,ω)$ for relevant experimental values of $ω$ even at distances up to around 80 Å. These findings illustrate how the emergent physics from underlying Coulomb interactions results in the observed weakly correlated semimetal.

cond-mat.mtrl-sci↗

arXiv2015

Density-matrix based determination of low-energy model Hamiltonians from ab initio wavefunctions

We propose a way of obtaining effective low energy Hubbard-like model Hamiltonians from ab initio Quantum Monte Carlo calculations for molecular and extended systems. The Hamiltonian parameters are fit to best match the ab initio two-body density matrices and energies of the ground and excited states, and thus we refer to the method as ab initio density matrix based downfolding (AIDMD). For benzene (a finite system), we find good agreement with experimentally available energy gaps without using any experimental inputs. For graphene, a two dimensional solid (extended system) with periodic boundary conditions, we find the effective on-site Hubbard $U^{*}/t$ to be $1.3 \pm 0.2$, comparable to a recent estimate based on the constrained random phase approximation. For molecules, such parameterizations enable calculation of excited states that are usually not accessible within ground state approaches. For solids, the effective Hamiltonian enables large scale calculations using techniques designed for lattice models.

cond-mat.str-el↗

arXiv2015

Computation of the correlated metal-insulator transition in vanadium dioxide from first principles

Vanadium dioxide(VO$_2$) is a paradigmatic example of a strongly correlated system that undergoes a metal-insulator transition at a structural phase transition. To date, this transition has necessitated significant post-hoc adjustments to theory in order to be described properly. Here we report standard state-of-the-art first principles quantum Monte Carlo (QMC) calculations of the structural dependence of the properties of VO$_2$. Using this technique, we simulate the interactions between electrons explicitly, which allows for the metal-insulator transition to naturally emerge, importantly without ad-hoc adjustments. The QMC calculations show that the structural transition directly causes the metal-insulator transition and a change in the coupling of vanadium spins. This change in the spin coupling results in a prediction of a momentum-independent magnetic excitation in the insulating state. While two-body correlations are important to set the stage for this transition, they do not change significantly when VO$_2$ becomes an insulator. These results show that it is now possible to account for electron correlations in a quantitatively accurate way that is also specific to materials.

cond-mat.str-el↗