Researcher profile

Linfeng Zhang

Linfeng Zhang contributes to research discovery and scholarly infrastructure.

ResearcherAffiliation not importedOpen to collaborate

physics.comp-ph Computer Vision Machine Learning Artificial Intelligence physics.chem-ph cond-mat.mtrl-sci Computation and Language cond-mat.dis-nn cond-mat.stat-mech math.NA Numerical Analysis cond-mat.quant-gas cond-mat.soft cond-mat.str-el Distributed, Parallel, and Cluster Computing Graphics Neural and Evolutionary Computing physics.app-ph Software Engineering

Trust snapshot

Quick read

Trust 21 - EmergingVerification L1Unclaimed author

48works

0followers

19topics

4close collaborators

Actions

Decide how to stay connected

Follow researcher0

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

Direct collaboration

Open a focused conversation when the fit is right

Claim this author entity first to unlock direct invitations.

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.

preprint2026arXiv

AgentSlimming: Towards Efficient and Cost-Aware Multi-Agent Systems

Large Language Model-based Multi-Agent Systems (MAS) have demonstrated remarkable capabilities in complex tasks. However, manually designing optimal communication topologies is labor-intensive, while automated expansion methods often result in bloated structures with redundant agents, leading to excessive token consumption. To address this problem, we introduce \textbf{AgentSlimming}, a plug-and-play compression framework for graph-structured multi-agent workflows. Motivated by pruning and quantization in neural networks, AgentSlimming compresses workflows by first estimating the importance score of each agent with a hybrid mechanism, and then removes redundant agents or replaces them with low-cost ones, where each operation is validated using a baseline-anchored acceptance rule to prevent performance collapse. Experiments show that AgentSlimming reduces average token cost by up to 78.9\% with negligible performance degradation, and sometimes even improves accuracy, achieving a strong Pareto-optimal trade-off between cost and quality. \textit{Our code is publicly available at https://github.com/CitrusYL/AgentSlimming

preprint2026arXiv

Attributing Emergence in Million-Agent Systems

Large language models (LLMs) can simulate human-like reasoning and decision-making in individual agents. LLM-powered multi-agent systems (MAS) combine such agents to simulate population-scale social phenomena such as polarization, information cascades, and market panics. Such studies require attributing macro emergence to individual agents, but existing axiomatic methods scale combinatorially in $N$ and have been confined to $N \lesssim 10^3$, while the phenomena they explain occur at $N \geq 10^6$. We address this gap by adapting Aumann--Shapley path-integral attribution to LLM-powered MAS at million-agent scale; the resulting method satisfies all four axioms, runs four to five orders of magnitude faster than sampled Shapley on the same hardware. We use this method to test the scale gap empirically: across 14 days of public Bluesky data ($1{,}671{,}587$ active users), we compute the attribution at both full scale and the visibility-biased $N = 10^2$ convenience sample used by small-scale studies, and the two disagree structurally. At full scale the long tail and middle tier jointly carry the majority; the biased small panel attributes almost everything to a few high-follower accounts. We then prove that under any nonlinear macro indicator the disagreement cannot be reduced by post-hoc rescaling: an Attribution Scaling Bias theorem shows that no global rescaling factor can reconcile small-scale and full-scale attribution. Full-scale attribution is therefore not a methodological choice but a theoretical requirement for any nonlinear macro indicator.

preprint2026arXiv

Beyond Fixed Formulas: Data-Driven Linear Predictor for Efficient Diffusion Models

To address the high sampling cost of Diffusion Transformers (DiTs), feature caching offers a training-free acceleration method. However, existing methods rely on hand-crafted forecasting formulas that fail under aggressive skipping. We propose L2P (Learnable Linear Predictor), a simple data-driven caching framework that replaces fixed coefficients with learnable per-timestep weights. Rapidly trained in ~20 seconds on a single GPU, L2P accurately reconstructs current features from past trajectories. L2P significantly outperforms existing baselines: it achieves a 4.55x FLOPs reduction and 4.15x latency speedup on FLUX.1-dev, and maintains high visual fidelity under up to 7.18x acceleration on Qwen-Image models, where prior methods show noticeable quality degradation. Our results show learning linear predictors is highly effective for efficient DiT inference. Code is available at https://github.com/Aredstone/L2P-Cache.

preprint2026arXiv

Controllable Video Generation: A Survey

With the rapid development of AI-generated content (AIGC), video generation has emerged as one of its most dynamic and impactful subfields. In particular, the advancement of video generation foundation models has led to growing demand for controllable video generation methods that can more accurately reflect user intent. Most existing foundation models are designed for text-to-video generation, where text prompts alone are often insufficient to express complex, multi-modal, and fine-grained user requirements. This limitation makes it challenging for users to generate videos with precise control using current models. To address this issue, recent research has explored the integration of additional non-textual conditions, such as camera motion, depth maps, and human pose, to extend pretrained video generation models and enable more controllable video synthesis. These approaches aim to enhance the flexibility and practical applicability of AIGC-driven video generation systems. In this survey, we provide a systematic review of controllable video generation, covering both theoretical foundations and recent advances in the field. We begin by introducing the key concepts and commonly used open-source video generation models. We then focus on control mechanisms in video diffusion models, analyzing how different types of conditions can be incorporated into the denoising process to guide generation. Finally, we categorize existing methods based on the types of control signals they leverage, including single-condition generation, multi-condition generation, and universal controllable generation. For a complete list of the literature on controllable video generation reviewed, please visit our curated repository at https://github.com/mayuelala/Awesome-Controllable-Video-Generation.

preprint2026arXiv

DataMaster: Data-Centric Autonomous AI Research

As model families, training recipes, and compute budgets become increasingly standardized, further gains in machine learning systems depend increasingly on data. Yet data engineering remains largely manual and ad hoc: practitioners repeatedly search for external datasets, adapt them to existing pipelines, validate candidate data through downstream training, and carry forward lessons from prior attempts. We study task-conditioned autonomous data engineering, where an autonomous agent improves a fixed learning algorithm by optimizing only the data side, including external data discovery, data selection and composition, cleaning and transformation. The goal is to obtain a stronger downstream solution while leaving the learning algorithm unchanged. To address the open-ended search space, branch-dependent refinement, and delayed validation inherent in autonomous data engineering, we propose DataMaster, a data-agent framework that integrates tree-structured search, shared candidate data, and cumulative memory. DataMaster consists of three key components: a DataTree that organizes alternative data-engineering branches, a shared Data Pool that stores discovered external data sources for reuse, and a Global Memory that records node outcomes, artifacts, and reusable findings. Together, these components allow the agent to discover candidate data, construct executable training inputs, evaluate them through downstream feedback, and carry useful evidence across branches. We evaluate DataMaster on two types of benchmarks, MLE-Bench Lite and PostTrainBench. On MLE-Bench Lite, it improves medal rate by 32.27% over the initial score; on PostTrainBench, it surpasses the instruct model on GPQA (31.02% vs 30.35%).

preprint2026arXiv

Deploy-Master: Automating the Deployment of 50,000+ Agent-Ready Scientific Tools in One Day

Open-source scientific software is abundant, yet most tools remain difficult to compile, configure, and reuse, sustaining a small-workshop mode of scientific computing. This deployment bottleneck limits reproducibility, large-scale evaluation, and the practical integration of scientific tools into modern AI-for-Science (AI4S) and agentic workflows. We present Deploy-Master, a one-stop agentic workflow for large-scale tool discovery, build specification inference, execution-based validation, and publication. Guided by a taxonomy spanning 90+ scientific and engineering domains, our discovery stage starts from a recall-oriented pool of over 500,000 public repositories and progressively filters it to 52,550 executable tool candidates under license- and quality-aware criteria. Deploy-Master transforms heterogeneous open-source repositories into runnable, containerized capabilities grounded in execution rather than documentation claims. In a single day, we performed 52,550 build attempts and constructed reproducible runtime environments for 50,112 scientific tools. Each successful tool is validated by a minimal executable command and registered in SciencePedia for search and reuse, enabling direct human use and optional agent-based invocation. Beyond delivering runnable tools, we report a deployment trace at the scale of 50,000 tools, characterizing throughput, cost profiles, failure surfaces, and specification uncertainty that become visible only at scale. These results explain why scientific software remains difficult to operationalize and motivate shared, observable execution substrates as a foundation for scalable AI4S and agentic science.

preprint2026arXiv

DISA: Offline Importance Sampling for Distribution-Matching LLM-RL

Modern reasoning agents are increasingly evaluated on their ability to generate multiple valid solution paths, plans, or tool-use traces for a given input. Standard reward-maximizing RL tends to collapse onto the most easily reinforced high-reward mode, whereas distribution-matching RL aims to allocate probability mass across the entire reward-shaped solution set. Achieving this objective requires computing a prompt-dependent partition function over the trajectory space. Because existing distribution-matching methods learn this partition function online alongside the policy, calibration errors in the partition function directly distort policy updates and remain impossible to diagnose independently. We introduce DISA, short for Decoupled Importance-Sampled Anchoring, which moves this calibration problem outside the RL loop. DISA draws proposal trajectories offline, estimates the partition function via importance sampling, and freezes the resulting partition-function estimate before policy optimization begins. This decoupling preserves the distribution-matching objective while strictly separating partition-function estimation from policy learning in data, gradients, loss, and diagnostics. Empirically, on two open-weight backbones across six math and three code benchmarks, DISA matches or exceeds the online-coupled distribution-matching baseline FlowRL, outperforms rewardmaximization baselines GRPO and GSPO on math averages, and exceeds LoRASFT distillation by up to 13.8 Mean@8 points on the same offline trajectories. An LLM-as-judge evaluation further shows that DISA retains substantially more strategy-level diversity than reward-maximization baselines, and sensitivity studies on the proposal strength and inverse temperature follow the bias-variance pattern predicted by the analysis.

preprint2026arXiv

DVD: A Robust Method for Detecting Variant Contamination in Large Language Model Evaluation

Evaluating large language models (LLMs) is increasingly confounded by \emph{variant contamination}: the training corpus contains semantically equivalent yet lexically or syntactically altered versions of test items. Unlike verbatim leakage, these paraphrased or structurally transformed variants evade existing detectors based on sampling consistency or perplexity, thereby inflating benchmark scores via memorization rather than genuine reasoning. We formalize this problem and introduce \textbf{DVD} (\textbf{D}etection via \textbf{V}ariance of generation \textbf{D}istribution), a single-sample detector that models the local output distribution induced by temperature sampling. Our key insight is that contaminated items trigger alternation between a \emph{memory-adherence} state and a \emph{perturbation-drift} state, yielding abnormally high variance in the synthetic difficulty of low-probability tokens; uncontaminated items remain in drift with comparatively smooth variance. We construct the first benchmark for variant contamination across two domains Omni-MATH and SuperGPQA by generating and filtering semantically equivalent variants, and simulate contamination via fine-tuning models of different scales and architectures (Qwen2.5 and Llama3.1). Across datasets and models, \textbf{DVD} consistently outperforms perplexity-based, Min-$k$\%++, edit-distance (CDD), and embedding-similarity baselines, while exhibiting strong robustness to hyperparameters. Our results establish variance of the generation distribution as a principled and practical fingerprint for detecting variant contamination in LLM evaluation.

preprint2026arXiv

Elastic-dLLM: Position Preserving Context Compression and Augmentation of Diffusion LLMs

Unlike autoregressive models, which generate one token at a time, dLLMs denoise a chunk of [MASK] tokens jointly and sample one or more tokens per step; despite enabling parallel decoding, this process incurs substantial computational cost due to the large chunk size of masked tokens. We observe that much of this cost is spent on repeatedly processing the preceding context and many [MASK] tokens with the same feature representations, indicating considerable computational redundancy. In this work, we revisit dLLM's redundancy from the perspective of [MASK] tokens. Through systematic analysis, we verify the redundancy of [MASK] tokens while revealing their critical role in providing structural information. Guided by these findings, we propose position-preserving [MASK] token compression and terminal-aware augmentation. By compressing redundant [MASK] computation, this approach accelerates decoding and further provides a natural extension toward context-folding-like long-context scaling under limited input-length constraints for full-sequence dLLMs such as LLaDA-8B-Instruct and LLaDA-1.5. Moreover, for block dLLMs such as LLaDA2.0-mini, it augments the context with a protected terminal [MASK] token to enhance generation quality with negligible overhead.

preprint2026arXiv

Entropy-Gradient Inversion: Moving Toward Internal Mechanism of Large Reasoning Models

The advancement of Large Reasoning Models (LRMs) has catalyzed a paradigm shift from reactive ``fast thinking'' text generation to systematic, step-by-step ``slow thinking'' reasoning, unlocking state-of-the-art performance in complex mathematical and logical tasks. However, the field faces \textit{the fundamental gap between token-level behavioral analysis and internal reasoning mechanisms, and the instability of reinforcement learning (RL) for reasoning optimization relying on costly external verifiers}. We identify and formally define \textbf{Entropy-Gradient Inversion}, a robust negative correlation between token entropy and logit gradients that acts as a definitive geometric fingerprint for LRM reasoning capability. Building on this, we propose \textbf{Correlation-Regularized Group Policy Optimization (CorR-PO)}, which embeds this inversion signature into RL reward regularization. Extensive experiments on various reasoning benchmarks across multiple model scales show CorR-PO consistently outperforms state-of-the-art baselines, confirming that stronger inversion directly correlates with superior reasoning performance.

preprint2026arXiv

EvoStreaming: Your Offline Video Model Is a Natively Streaming Assistant

Streaming video understanding demands more than watching longer videos: assistants must decide when to speak in real time, balancing responsiveness against verbosity. Yet most video-language models (VideoLLMs) are trained for offline inference, and existing streaming benchmarks externalize this timing decision to the evaluator. We address this gap with RealStreamEval, a frame-level multi-turn evaluation protocol that exposes models to sequential observations and penalizes unnecessary responses. Under this protocol, we observed that strong offline VideoLLMs retain useful visual understanding but lack an interaction policy for deciding when to respond. Motivated by this observation, we propose EvoStreaming, a self-evolved streaming adaptation framework in which the base model itself acts as data generator, relevance annotator, and roll-out policy to synthesize streaming trajectories without external supervision. With only $1{,}000$ self-generated samples ($139\times$ less than the leading streaming instruction-tuning approach) and no architectural changes, EvoStreaming consistently improves the overall RealStreamEval score by up to $10.8$ points across five open VideoLLM backbones (Qwen2/2.5/3-VL, InternVL-3.5, MiniCPM-V4.5) while largely preserving offline video performance. These results suggest that data-efficient interaction tuning is a practical path for adapting existing VideoLLMs to streaming assistants.

preprint2026arXiv

ExtraVAR: Stage-Aware RoPE Remapping for Resolution Extrapolation in Visual Autoregressive Models

Visual Autoregressive (VAR) models have emerged as a strong alternative to diffusion for image synthesis, yet their fixed training resolution prevents direct generation at higher resolutions. Naively transferring training-free extrapolation methods from LLMs or diffusion models to VAR yields three characteristic failure modes: global repetition, local repetition, and detail degradation. We trace them to a unified band-stage mismatch: VAR generates images in a coarse-to-fine, scale-wise process where each stage is driven by a distinct dominant RoPE frequency band, and each failure mode emerges when the dominant band of a particular stage is disrupted. Building on this insight, we propose Stage-Aware RoPE Remapping, a training-free strategy that assigns each frequency band a stage-specific remapping rule, jointly suppressing all three failure modes. We further observe that attention becomes systematically dispersed as the image resolution increases. Existing methods typically depend on predefined attention scaling factors, which are neither adaptive to the target resolution nor capable of faithfully capturing the actual extent of attention dispersion. We therefore propose Entropy-Driven Adaptive Attention Calibration, which quantifies dispersion via a resolution-invariant normalized entropy and yields a closed-form per-head scaling factor that realigns the extrapolated-resolution attention entropy with its training-resolution counterpart. Extensive experiments show that our method consistently outperforms prior resolution-extrapolation methods in both structural coherence and fine-detail fidelity. Our code is available at https://github.com/feihongyan1/ExtraVAR.

preprint2026arXiv

FlexDraft: Flexible Speculative Decoding via Attention Tuning and Bonus-Guided Calibration

Speculative decoding accelerates memory-bound LLM inference without quality degradation by using a fast drafter to propose multiple candidate tokens and the target model to verify them in parallel. However, conventional sequential speculative decoding suffers from mutual waiting between drafting and verification, and repeated exchange of intermediate states further increases memory access overhead. Parallel speculative decoding addresses this limitation by performing drafting and verification within a single target forward pass, allowing future drafts to be prepared while current candidates are being verified. Although effective at small batch sizes, existing parallel speculative decoding methods either require costly continual pretraining with quality degradation or suffer from low acceptance rates. More importantly, this paradigm inherently suffers from uncertainty in both the bonus token and the accepted length, leading to draft verification mismatch and causing throughput gains to collapse at large batch sizes. To address these limitations, we introduce FlexDraft, a lossless speculative decoding framework that flexibly adapts to varying batch sizes through three key designs. (1) Attention Tuning enables block diffusion drafting by tuning only the attention projectors of the final few layers on mask tokens, while keeping the autoregressive path frozen to preserve the target distribution and produce high quality drafts with minimal trainable parameters. (2) Bonus-guided Calibration uses a lightweight MLP conditioned on the resolved bonus token to calibrate draft logits, mitigating draft verification mismatch caused by bonus token uncertainty. (3) Flex Decoding dynamically switches between parallel draft and verify at small batch sizes and sequential draft then verify at large batch sizes, and adjusts verification length based on draft confidence to eliminate redundant computation.

preprint2026arXiv

Focused Forcing: Content-Aware Per-Frame KV Selection for Efficient Autoregressive Video Diffusion

Recent advances in autoregressive video diffusion have enabled sequential and streaming video generation. However, long-horizon generation requires increasingly large KV caches, making efficient compression without sacrificing quality challenging. Existing methods mostly select historical frames based on attention scores, but their context decisions remain coarse. When multiple frames are generated in the same chunk, these methods often apply a shared history selection to the whole chunk, score historical frames solely by attention, and assign head-wise budgets either uniformly or by attention-pattern heuristics rather than explicit head-importance estimation. We show that frames within the same generated chunk can depend on distinct historical frames, that the same historical frame can receive different attention scores as its relative temporal distance to the current frames changes, and that masking different heads induces unequal generation degradation. Motivated by these findings, we propose \textbf{Focused Forcing}, a training-free KV selection method that focuses cached history along both generated-frame and head dimensions. For each generated frame, Focused Forcing preserves the most relevant and distinctive historical frames by combining attention scores with diversity scores of historical frames, while assigning larger budgets to heads with higher estimated importance. Across multiple autoregressive generation paradigms, Focused Forcing achieves up to $\textbf{1.48}\times$ end-to-end acceleration without training, while \textbf{improving visual quality and text alignment}. \textit{Our code will be released on GitHub.}

preprint2026arXiv

Forecast the Principal, Stabilize the Residual: Subspace-Aware Feature Caching for Efficient Diffusion Transformers

Diffusion Transformer (DiT) models have achieved unprecedented quality in image and video generation, yet their iterative sampling process remains computationally prohibitive. To accelerate inference, feature caching methods have emerged by reusing intermediate representations across timesteps. However, existing caching approaches treat all feature components uniformly. We reveal that DiT feature spaces contain distinct principal and residual subspaces with divergent temporal behavior: the principal subspace evolves smoothly and predictably, while the residual subspace exhibits volatile, low-energy oscillations that resist accurate prediction. Building on this insight, we propose SVD-Cache, a subspace-aware caching framework that decomposes diffusion features via Singular Value Decomposition (SVD), applies exponential moving average (EMA) prediction to the dominant low-rank components, and directly reuses the residual subspace. Extensive experiments demonstrate that SVD-Cache achieves near-lossless across diverse models and methods, including 5.55$\times$ speedup on FLUX and HunyuanVideo, and compatibility with model acceleration techniques including distillation, quantization and sparse attention. Our code is in supplementary material and will be released on Github.

preprint2026arXiv

From Sketch to Fresco: Efficient Diffusion Transformer with Progressive Resolution

Diffusion Transformers achieve impressive generative quality but remain computationally expensive due to iterative sampling. Recently, dynamic resolution sampling has emerged as a promising acceleration technique by reducing the resolution of early sampling steps. However, existing methods rely on heuristic re-noising at every resolution transition, injecting noise that breaks cross-stage consistency and forces the model to relearn global structure. In addition, these methods indiscriminately upsample the entire latent space at once without checking which regions have actually converged, causing accumulated errors, and visible artifacts. Therefore, we propose \textbf{Fresco}, a dynamic resolution framework that unifies re-noise and global structure across stages with progressive upsampling, preserving both the efficiency of low-resolution drafting and the fidelity of high-resolution refinement, with all stages aligned toward the same final target. Fresco achieves near-lossless acceleration across diverse domains and models, including 10$\times$ speedup on FLUX, and 5$\times$ on HunyuanVideo, while remaining orthogonal to distillation, quantization and feature caching, reaching 22$\times$ speedup when combined with distilled models. Our code is in supplementary material and will be released on Github.

preprint2026arXiv

Inverse Knowledge Search over Verifiable Reasoning: Synthesizing a Scientific Encyclopedia from a Long Chains-of-Thought Knowledge Base

Most scientific materials compress reasoning, presenting conclusions while omitting the derivational chains that justify them. This compression hinders verification by lacking explicit, step-wise justifications and inhibits cross-domain links by collapsing the very pathways that establish the logical and causal connections between concepts. We introduce a scalable framework that decompresses scientific reasoning, constructing a verifiable Long Chain-of-Thought (LCoT) knowledge base and projecting it into an emergent encyclopedia, SciencePedia. Our pipeline operationalizes an endpoint-driven, reductionist strategy: a Socratic agent, guided by a curriculum of around 200 courses, generates approximately 3 million first-principles questions. To ensure high fidelity, multiple independent solver models generate LCoTs, which are then rigorously filtered by prompt sanitization and cross-model answer consensus, retaining only those with verifiable endpoints. This verified corpus powers the Brainstorm Search Engine, which performs inverse knowledge search -- retrieving diverse, first-principles derivations that culminate in a target concept. This engine, in turn, feeds the Plato synthesizer, which narrates these verified chains into coherent articles. The initial SciencePedia comprises approximately 200,000 fine-grained entries spanning mathematics, physics, chemistry, biology, engineering, and computation. In evaluations across six disciplines, Plato-synthesized articles (conditioned on retrieved LCoTs) exhibit substantially higher knowledge-point density and significantly lower factual error rates than an equally-prompted baseline without retrieval (as judged by an external LLM). Built on this verifiable LCoT knowledge base, this reasoning-centric approach enables trustworthy, cross-domain scientific synthesis at scale and establishes the foundation for an ever-expanding encyclopedia.

preprint2026arXiv

Multi-Task Fine-Tuning Enables Robust Out-of-Distribution Generalization in Atomistic Models

Accurate de novo molecular and materials design requires structure-property models that generalize beyond known regimes. Although pretrained atomistic models achieve strong in-distribution accuracy after fine-tuning, their reliability under out-of-distribution (OOD) conditions remains unclear. We identify a critical failure mode in downstream adaptation: standard fine-tuning induces representation collapse, erasing pretrained chemical and structural priors and severely degrading OOD performance. To address this limitation, we propose multi-task fine-tuning (MFT), which jointly optimizes downstream property prediction with a physically grounded force-field objective inherited from pretraining. This approach preserves essential chemical priors while enabling task-specific adaptation. Across molecular and materials benchmarks, MFT consistently improves OOD generalization, approaching the theoretical limit set by in-distribution accuracy, while outperforming standard fine-tuning, training from scratch, and state-of-the-art task-specific models. These results establish safe adaptation as a central requirement for large atomistic models and position MFT as a practical and data-efficient pathway toward robust molecular and materials discovery.

preprint2026arXiv

Reasoning emerges from constrained inference manifolds in large language models

Reasoning in large language models is predominantly evaluated through labeled benchmarks, conflating task performance with the quality of internal inference. Here we study reasoning as an intrinsic dynamical process by examining the evolution of internal representations during inference. We find that inference-time dynamics consistently self-organize into low-dimensional manifolds embedded within high-dimensional representation spaces. we find that such geometric compression, although pervasive, is not sufficient for stable or reliable reasoning. Instead, effective reasoning dynamics emerge within a constrained structural regime characterized by three conditions: adequate representational expressivity, spontaneous manifold compression, and preservation of non-degenerate information volume within the compressed subspace. Models outside this regime exhibit characteristic pathological inference dynamics. Based on these insights, we introduce a unified, label-free diagnostic computed solely from internal dynamics. These findings suggest that reasoning in LLMs is fundamentally governed by geometric and informational constraints, offering a complementary framework to benchmark-centric assessment.

preprint2026arXiv

Respecting Self-Uncertainty in On-Policy Self-Distillation for Efficient LLM Reasoning

On-policy self-distillation trains a reasoning model on its own rollouts while a teacher, often the same model conditioned on privileged context, provides dense token-level supervision. Existing objectives typically weight the teacher's token-level signal uniformly across a chain-of-thought sequence, despite substantial variation in the entropy of the teacher's predictive distribution. We propose EGRSD (Entropy-Guided Reinforced Self-Distillation), which unifies token-level updates through three signals: a reward-grounded direction, a teacher-student likelihood-ratio magnitude, and the proposed teacher-entropy confidence gate that down-weights high-entropy token positions while maintaining a nonzero lower bound on every token weight. We further introduce CL-EGRSD, a causal-lookahead variant that distinguishes sustained high-entropy spans from transient high-entropy positions whose following context rapidly becomes low entropy. Experiments with Qwen3-4B and Qwen3-8B in thinking mode show that EGRSD and CL-EGRSD advance the accuracy-length frontier among the compared trainable methods.

preprint2026arXiv

SpecEdit: Training-Free Acceleration for Diffusion based Image Editing via Semantic Locking

Diffusion-based image editing offers strong semantic controllability, but remains computationally expensive due to iterative high-resolution denoising over all spatial tokens. Dynamic-resolution sampling reduces this cost by performing early steps at reduced resolution. However, existing approaches prioritize upsampling using low-level heuristics such as edge detection or channel variance, which are weakly aligned with editing semantics and may lead to structural inconsistency. Moreover, spatial regions are often upsampled without verifying whether semantic modification is actually required, resulting in redundant high-resolution computation and accumulated errors. Therefore, we propose SpecEdit, a training-free dynamic-resolution framework tailored for diffusion-based image editing. SpecEdit follows a draft-and-verify scheme: a low-resolution draft first estimates the semantic outcome, after which token-level discrepancies are used to identify edit-relevant tokens for high-resolution denoising, while the remaining tokens stay at a coarse resolution. Experiments on Qwen-Image-Edit and FLUX.1-Kontext-dev demonstrate up to 10x and 7x acceleration, while maintaining strong quality. SpecEdit is complementary to step distillation and other acceleration techniques, achieving up to 13x speedup when combined with existing methods. Our code is in supplementary material and will be released on GitHub.

preprint2022arXiv

A deep potential model with long-range electrostatic interactions

Machine learning models for the potential energy of multi-atomic systems, such as the deep potential (DP) model, make possible molecular simulations with the accuracy of quantum mechanical density functional theory, at a cost only moderately higher than that of empirical force fields. However, the majority of these models lack explicit long-range interactions and fail to describe properties that derive from the Coulombic tail of the forces. To overcome this limitation we extend the DP model by approximating the long-range electrostatic interaction between ions (nuclei+core electrons) and valence electrons with that of distributions of spherical Gaussian charges located at ionic and electronic sites. The latter are rigorously defined in terms of the centers of the maximally localized Wannier distributions, whose dependence on the local atomic environment is modeled accurately by a deep neural network. In the deep potential long-range (DPLR) model, the electrostatic energy of the Gaussian charge system is added to short-range interactions that are represented as in the standard DP model. The resulting potential energy surface is smooth and possesses analytical forces and virial. Missing effects in the standard DP scheme are recovered, improving on accuracy and predictive power. By including long-range electrostatics, DPLR correctly extrapolates to large systems the potential energy surface learned from quantum mechanical calculations on smaller systems. We illustrate the approach with three examples, the potential energy profile of the water dimer, the free energy of interaction of a water molecule with a liquid water slab, and the phonon dispersion curves of the NaCl crystal.

preprint2022arXiv

Ab-initio study of interacting fermions at finite temperature with neural canonical transformation

We present a variational density matrix approach to the thermal properties of interacting fermions in the continuum. The variational density matrix is parametrized by a permutation equivariant many-body unitary transformation together with a discrete probabilistic model. The unitary transformation is implemented as a quantum counterpart of neural canonical transformation, which incorporates correlation effects via a flow of fermion coordinates. As the first application, we study electrons in a two-dimensional quantum dot with an interaction-induced crossover from Fermi liquid to Wigner molecule. The present approach provides accurate results in the low-temperature regime, where conventional quantum Monte Carlo methods face severe difficulties due to the fermion sign problem. The approach is general and flexible for further extensions, thus holds the promise to deliver new physical results on strongly correlated fermions in the context of ultracold quantum gases, condensed matter, and warm dense matter physics.

preprint2022arXiv

Contrastive Deep Supervision

The success of deep learning is usually accompanied by the growth in neural network depth. However, the traditional training method only supervises the neural network at its last layer and propagates the supervision layer-by-layer, which leads to hardship in optimizing the intermediate layers. Recently, deep supervision has been proposed to add auxiliary classifiers to the intermediate layers of deep neural networks. By optimizing these auxiliary classifiers with the supervised task loss, the supervision can be applied to the shallow layers directly. However, deep supervision conflicts with the well-known observation that the shallow layers learn low-level features instead of task-biased high-level semantic features. To address this issue, this paper proposes a novel training framework named Contrastive Deep Supervision, which supervises the intermediate layers with augmentation-based contrastive learning. Experimental results on nine popular datasets with eleven models demonstrate its effects on general image classification, fine-grained image classification and object detection in supervised learning, semi-supervised learning and knowledge distillation. Codes have been released in Github.

preprint2022arXiv

DeePKS+ABACUS as a Bridge between Expensive Quantum Mechanical Models and Machine Learning Potentials

Recently, the development of machine learning (ML) potentials has made it possible to perform large-scale and long-time molecular simulations with the accuracy of quantum mechanical (QM) models. However, for high-level QM methods, such as density functional theory (DFT) at the meta-GGA level and/or with exact exchange, quantum Monte Carlo, etc., generating a sufficient amount of data for training a ML potential has remained computationally challenging due to their high cost. In this work, we demonstrate that this issue can be largely alleviated with Deep Kohn-Sham (DeePKS), a ML-based DFT model. DeePKS employs a computationally efficient neural network-based functional model to construct a correction term added upon a cheap DFT model. Upon training, DeePKS offers closely-matched energies and forces compared with high-level QM method, but the number of training data required is orders of magnitude less than that required for training a reliable ML potential. As such, DeePKS can serve as a bridge between expensive QM models and ML potentials: one can generate a decent amount of high-accuracy QM data to train a DeePKS model, and then use the DeePKS model to label a much larger amount of configurations to train a ML potential. This scheme for periodic systems is implemented in a DFT package ABACUS, which is open-source and ready for use in various applications.

preprint2022arXiv

DP Compress: a Model Compression Scheme for Generating Efficient Deep Potential Models

Machine-learning-based interatomic potential energy surface (PES) models are revolutionizing the field of molecular modeling. However, although much faster than electronic structure schemes, these models suffer from costly computations via deep neural networks to predict the energy and atomic forces, resulting in lower running efficiency as compared to the typical empirical force fields. Herein, we report a model compression scheme for boosting the performance of the Deep Potential (DP) model, a deep learning based PES model. This scheme, we call DP Compress, is an efficient post-processing step after the training of DP models (DP Train). DP Compress combines several DP-specific compression techniques, which typically speed up DP-based molecular dynamics simulations by an order of magnitude faster, and consume an order of magnitude less memory. We demonstrate that DP Compress is sufficiently accurate by testing a variety of physical properties of Cu, H2O, and Al-Cu-Mg systems. DP Compress applies to both CPU and GPU machines and is publicly available online.

preprint2022arXiv

Extending the limit of molecular dynamics with ab initio accuracy to 10 billion atoms

High-performance computing, together with a neural network model trained from data generated with first-principles methods, has greatly boosted applications of \textit{ab initio} molecular dynamics in terms of spatial and temporal scales on modern supercomputers. Previous state-of-the-art can achieve $1-2$ nanoseconds molecular dynamics simulation per day for 100-million atoms on the entire Summit supercomputer. In this paper, we have significantly reduced the memory footprint and computational time by a comprehensive approach with both algorithmic and system innovations. The neural network model is compressed by model tabulation, kernel fusion, and redundancy removal. Then optimizations such as acceleration of customized kernel, tabulation of activation function, MPI+OpenMP parallelization are implemented on GPU and ARM architectures. Testing results of the copper system show that the optimized code can scale up to the entire machine of both Fugaku and Summit, and the corresponding system size can be extended by a factor of $134$ to an unprecedented $17$ billion atoms. The strong scaling of a $13.5$-million atom copper system shows that the time-to-solution can be 7 times faster, reaching $11.2$ nanoseconds per day. This work opens the door for unprecedentedly large-scale molecular dynamics simulations based on {\it ab initio} accuracy and can be potentially utilized in studying more realistic applications such as mechanical properties of metals, semiconductor devices, batteries, etc. The optimization techniques detailed in this paper also provide insight for relevant high-performance computing applications.

preprint2022arXiv

Finding the Task-Optimal Low-Bit Sub-Distribution in Deep Neural Networks

Quantized neural networks typically require smaller memory footprints and lower computation complexity, which is crucial for efficient deployment. However, quantization inevitably leads to a distribution divergence from the original network, which generally degrades the performance. To tackle this issue, massive efforts have been made, but most existing approaches lack statistical considerations and depend on several manual configurations. In this paper, we present an adaptive-mapping quantization method to learn an optimal latent sub-distribution that is inherent within models and smoothly approximated with a concrete Gaussian Mixture (GM). In particular, the network weights are projected in compliance with the GM-approximated sub-distribution. This sub-distribution evolves along with the weight update in a co-tuning schema guided by the direct task-objective optimization. Sufficient experiments on image classification and object detection over various modern architectures demonstrate the effectiveness, generalization property, and transferability of the proposed method. Besides, an efficient deployment flow for the mobile CPU is developed, achieving up to 7.46$\times$ inference acceleration on an octa-core ARM CPU. Our codes have been publicly released at \url{https://github.com/RunpeiDong/DGMS}.

preprint2022arXiv

Machine-learning interatomic potential for molecular dynamics simulation of ferroelectric KNbO3 perovskite

Ferroelectric perovskites have been ubiquitously applied in piezoelectric devices for decades, among which, eco-friendly lead-free (K,Na)NbO3-based materials have been recently demonstrated to be an excellent candidate for sustainable development. Molecular dynamics is a versatile theoretical calculation approach for the investigation of the dynamical properties of ferroelectric perovskites. However, molecular dynamics simulation of ferroelectric perovskites has been limited to simple systems, since the conventional construction of interatomic potential is rather difficult and inefficient. In the present study, we construct a machine-learning interatomic potential of KNbO3 (as a representative system of (K,Na)NbO3) by using a deep neural network model. Including first-principles calculation data into the training dataset ensures the quantum-mechanics accuracy of the interatomic potential. The molecular dynamics based on machine-learning interatomic potential shows good agreement with the first-principles calculations, which can accurately predict multiple fundamental properties, e.g., atomic force, energy, elastic properties, and phonon dispersion. In addition, the interatomic potential exhibits satisfactory performance in the simulation of domain wall and temperature-dependent phase transition. The construction of interatomic potential based on machine learning could potentially be transferred to other ferroelectric perovskites and consequently benefits the theoretical study of ferroelectrics.

preprint2022arXiv

Neural network representation of electronic structure from $ab$ $initio$ molecular dynamics

Despite their rich information content, electronic structure data amassed at high volumes in $ab$ $initio$ molecular dynamics simulations are generally under-utilized. We introduce a transferable high-fidelity neural network representation of such data in the form of tight-binding Hamiltonians for crystalline materials. This predictive representation of $ab$ $initio$ electronic structure, combined with machine-learning boosted molecular dynamics, enables efficient and accurate electronic evolution and sampling. When applied to a one-dimension charge-density wave material, carbyne, we are able to compute the spectral function and optical conductivity in the canonical ensemble. The spectral functions evaluated during soliton-antisoliton pair annihilation process reveal significant renormalization of low-energy edge modes due to retarded electron-lattice coupling beyond the Born-Oppenheimer limit. The availability of an efficient and reusable surrogate model for the electronic structure dynamical system will enable calculating many interesting physical properties, paving way to previously inaccessible or challenging avenues in materials modeling.

preprint2022arXiv

PointDistiller: Structured Knowledge Distillation Towards Efficient and Compact 3D Detection

The remarkable breakthroughs in point cloud representation learning have boosted their usage in real-world applications such as self-driving cars and virtual reality. However, these applications usually have an urgent requirement for not only accurate but also efficient 3D object detection. Recently, knowledge distillation has been proposed as an effective model compression technique, which transfers the knowledge from an over-parameterized teacher to a lightweight student and achieves consistent effectiveness in 2D vision. However, due to point clouds' sparsity and irregularity, directly applying previous image-based knowledge distillation methods to point cloud detectors usually leads to unsatisfactory performance. To fill the gap, this paper proposes PointDistiller, a structured knowledge distillation framework for point clouds-based 3D detection. Concretely, PointDistiller includes local distillation which extracts and distills the local geometric structure of point clouds with dynamic graph convolution and reweighted learning strategy, which highlights student learning on the crucial points or voxels to improve knowledge distillation efficiency. Extensive experiments on both voxels-based and raw points-based detectors have demonstrated the effectiveness of our method over seven previous knowledge distillation methods. For instance, our 4X compressed PointPillars student achieves 2.8 and 3.4 mAP improvements on BEV and 3D object detection, outperforming its teacher by 0.9 and 1.8 mAP, respectively. Codes have been released at https://github.com/RunpeiDong/PointDistiller.

preprint2022arXiv

Region-aware Knowledge Distillation for Efficient Image-to-Image Translation

Recent progress in image-to-image translation has witnessed the success of generative adversarial networks (GANs). However, GANs usually contain a huge number of parameters, which lead to intolerant memory and computation consumption and limit their deployment on edge devices. To address this issue, knowledge distillation is proposed to transfer the knowledge from a cumbersome teacher model to an efficient student model. However, most previous knowledge distillation methods are designed for image classification and lead to limited performance in image-to-image translation. In this paper, we propose Region-aware Knowledge Distillation ReKo to compress image-to-image translation models. Firstly, ReKo adaptively finds the crucial regions in the images with an attention module. Then, patch-wise contrastive learning is adopted to maximize the mutual information between students and teachers in these crucial regions. Experiments with eight comparison methods on nine datasets demonstrate the substantial effectiveness of ReKo on both paired and unpaired image-to-image translation. For instance, our 7.08X compressed and 6.80X accelerated CycleGAN student outperforms its teacher by 1.33 and 1.04 FID scores on Horse to Zebra and Zebra to Horse, respectively. Codes will be released on GitHub.

preprint2022arXiv

Structural Phase Transitions in SrTiO3 from Deep Potential Molecular Dynamics

Strontium titanate (SrTiO3) is regarded as an essential material for oxide electronics. One of its many remarkable features is subtle structural phase transition, driven by antiferrodistortive lattice mode, from a high-temperature cubic phase to a low-temperature tetragonal phase. Classical molecular dynamics (MD) simulation is an efficient technique to reveal atomistic features of phase transition, but its application is often limited by the accuracy of empirical interatomic potentials. Here, we develop an accurate deep potential (DP) model of SrTiO3 based on a machine learning method using data from first-principles density functional theory (DFT) calculations. The DP model has DFT-level accuracy, capable of performing efficient MD simulations and accurate property predictions. Using the DP model, we investigate the temperature-driven cubic-to-tetragonal phase transition and construct the in-plane biaxial strain-temperature phase diagram of SrTiO3. The simulations demonstrate that strain-induced ferroelectric phase is characterized by two order parameters, ferroelectric distortion and antiferrodistortion, and the ferroelectric phase transition has both displacive and order-disorder characters. This works lays the foundation for the development of accurate DP models of other complex perovskite materials.

preprint2022arXiv

Universal approximation of symmetric and anti-symmetric functions

We consider universal approximations of symmetric and anti-symmetric functions, which are important for applications in quantum physics, as well as other scientific and engineering computations. We give constructive approximations with explicit bounds on the number of parameters with respect to the dimension and the target accuracy $ε$. While the approximation still suffers from the curse of dimensionality, to the best of our knowledge, these are the first results in the literature with explicit error bounds for functions with symmetry or anti-symmetry constraints.

preprint2022arXiv

Viscosity in water from first-principles and deep-neural-network simulations

We report on an extensive study of the viscosity of liquid water at near-ambient conditions, performed within the Green-Kubo theory of linear response and equilibrium ab initio molecular dynamics (AIMD), based on density-functional theory (DFT). In order to cope with the long simulation times necessary to achieve an acceptable statistical accuracy, our ab initio approach is enhanced with deep-neural-network potentials (NNP). This approach is first validated against AIMD results, obtained by using the Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional and paying careful attention to crucial, yet often overlooked, aspects of the statistical data analysis. Then, we train a second NNP to a dataset generated from the Strongly Constrained and Appropriately Normed (SCAN) functional. Once the error resulting from the imperfect prediction of the melting line is offset by referring the simulated temperature to the theoretical melting one, our SCAN predictions of the shear viscosity of water are in very good agreement with experiments.

preprint2022arXiv

Wavelet Knowledge Distillation: Towards Efficient Image-to-Image Translation

Remarkable achievements have been attained with Generative Adversarial Networks (GANs) in image-to-image translation. However, due to a tremendous amount of parameters, state-of-the-art GANs usually suffer from low efficiency and bulky memory usage. To tackle this challenge, firstly, this paper investigates GANs performance from a frequency perspective. The results show that GANs, especially small GANs lack the ability to generate high-quality high frequency information. To address this problem, we propose a novel knowledge distillation method referred to as wavelet knowledge distillation. Instead of directly distilling the generated images of teachers, wavelet knowledge distillation first decomposes the images into different frequency bands with discrete wavelet transformation and then only distills the high frequency bands. As a result, the student GAN can pay more attention to its learning on high frequency bands. Experiments demonstrate that our method leads to 7.08 times compression and 6.80 times acceleration on CycleGAN with almost no performance drop. Additionally, we have studied the relation between discriminators and generators which shows that the compression of discriminators can promote the performance of compressed generators.

preprint2020arXiv

Deep Learning of Accurate Force Field of Ferroelectric HfO$_2$

The discovery of ferroelectricity in HfO$_2$-based thin films opens up new opportunities for using this silicon-compatible ferroelectric to realize low-power logic circuits and high-density non-volatile memories. The functional performances of ferroelectrics are intimately related to their dynamic responses to external stimuli such as electric fields at finite temperatures. Molecular dynamics is an ideal technique for investigating dynamical processes on large length and time scales, though its applications to new materials is often hindered by the limited availability and accuracy of classical force fields. Here we present a deep neural network-based interatomic force field of HfO$_2$ learned from {\em ab initio} data using a concurrent learning procedure. The model potential is able to predict structural properties such as elastic constants, equation of states, phonon dispersion relationships, and phase transition barriers of various hafnia polymorphs with accuracy comparable with density functional theory calculations. The validity of this model potential is further confirmed by the reproduction of experimental sequences of temperature-driven ferroelectric-paraelectric phase transitions of HfO$_2$ with isobaric-isothermal ensemble molecular dynamics simulations. We suggest a general approach to extend the model potential of HfO$_2$ to related material systems including dopants and defects.

preprint2020arXiv

Deep neural network for the dielectric response of insulators

We introduce a deep neural network to model in a symmetry preserving way the environmental dependence of the centers of the electronic charge. The model learns from ab-initio density functional theory, wherein the electronic centers are uniquely assigned by the maximally localized Wannier functions. When combined with the Deep Potential model of the atomic potential energy surface, the scheme predicts the dielectric response of insulators for trajectories inaccessible to direct ab-initio simulation. The scheme is non-perturbative and can capture the response of a mutating chemical environment. We demonstrate the approach by calculating the infrared spectra of liquid water at standard conditions, and of ice under extreme pressure, when it transforms from a molecular to an ionic crystal.

preprint2020arXiv

Ground state energy functional with Hartree-Fock efficiency and chemical accuracy

We introduce the Deep Post-Hartree-Fock (DeePHF) method, a machine learning based scheme for constructing accurate and transferable models for the ground-state energy of electronic structure problems. DeePHF predicts the energy difference between results of highly accurate models such as the coupled cluster method and low accuracy models such as the the Hartree-Fock (HF) method, using the ground-state electronic orbitals as the input. It preserves all the symmetries of the original high accuracy model. The added computational cost is less than that of the reference HF or DFT and scales linearly with respect to system size. We examine the performance of DeePHF on organic molecular systems using publicly available datasets and obtain the state-of-art performance, particularly on large datasets.

preprint2020arXiv

Integrating Machine Learning with Physics-Based Modeling

Machine learning is poised as a very powerful tool that can drastically improve our ability to carry out scientific research. However, many issues need to be addressed before this becomes a reality. This article focuses on one particular issue of broad interest: How can we integrate machine learning with physics-based modeling to develop new interpretable and truly reliable physical models? After introducing the general guidelines, we discuss the two most important issues for developing machine learning-based physical models: Imposing physical constraints and obtaining optimal datasets. We also provide a simple and intuitive explanation for the fundamental reasons behind the success of modern machine learning, as well as an introduction to the concurrent machine learning framework needed for integrating machine learning with physics-based modeling. Molecular dynamics and moment closure of kinetic equations are used as examples to illustrate the main issues discussed. We end with a general discussion on where this integration will lead us to, and where the new frontier will be after machine learning is successfully integrated into scientific modeling.

preprint2020arXiv

Neural Canonical Transformation with Symplectic Flows

Canonical transformation plays a fundamental role in simplifying and solving classical Hamiltonian systems. We construct flexible and powerful canonical transformations as generative models using symplectic neural networks. The model transforms physical variables towards a latent representation with an independent harmonic oscillator Hamiltonian. Correspondingly, the phase space density of the physical system flows towards a factorized Gaussian distribution in the latent space. Since the canonical transformation preserves the Hamiltonian evolution, the model captures nonlinear collective modes in the learned latent representation. We present an efficient implementation of symplectic neural coordinate transformations and two ways to train the model. The variational free energy calculation is based on the analytical form of physical Hamiltonian. While the phase space density estimation only requires samples in the coordinate space for separable Hamiltonians. We demonstrate appealing features of neural canonical transformation using toy problems including two-dimensional ring potential and harmonic chain. Finally, we apply the approach to real-world problems such as identifying slow collective modes in alanine dipeptide and conceptual compression of the MNIST dataset.

preprint2020arXiv

Non-Structured DNN Weight Pruning -- Is It Beneficial in Any Platform?

Large deep neural network (DNN) models pose the key challenge to energy efficiency due to the significantly higher energy consumption of off-chip DRAM accesses than arithmetic or SRAM operations. It motivates the intensive research on model compression with two main approaches. Weight pruning leverages the redundancy in the number of weights and can be performed in a non-structured, which has higher flexibility and pruning rate but incurs index accesses due to irregular weights, or structured manner, which preserves the full matrix structure with lower pruning rate. Weight quantization leverages the redundancy in the number of bits in weights. Compared to pruning, quantization is much more hardware-friendly, and has become a "must-do" step for FPGA and ASIC implementations. This paper provides a definitive answer to the question for the first time. First, we build ADMM-NN-S by extending and enhancing ADMM-NN, a recently proposed joint weight pruning and quantization framework. Second, we develop a methodology for fair and fundamental comparison of non-structured and structured pruning in terms of both storage and computation efficiency. Our results show that ADMM-NN-S consistently outperforms the prior art: (i) it achieves 348x, 36x, and 8x overall weight pruning on LeNet-5, AlexNet, and ResNet-50, respectively, with (almost) zero accuracy loss; (ii) we demonstrate the first fully binarized (for all layers) DNNs can be lossless in accuracy in many cases. These results provide a strong baseline and credibility of our study. Based on the proposed comparison framework, with the same accuracy and quantization, the results show that non-structrued pruning is not competitive in terms of both storage and computation efficiency. Thus, we conclude that non-structured pruning is considered harmful. We urge the community not to continue the DNN inference acceleration for non-structured sparsity.

preprint2020arXiv

Pushing the limit of molecular dynamics with ab initio accuracy to 100 million atoms with machine learning

For 35 years, {\it ab initio} molecular dynamics (AIMD) has been the method of choice for modeling complex atomistic phenomena from first principles. However, most AIMD applications are limited by computational cost to systems with thousands of atoms at most. We report that a machine learning-based simulation protocol (Deep Potential Molecular Dynamics), while retaining {\it ab initio} accuracy, can simulate more than 1 nanosecond-long trajectory of over 100 million atoms per day, using a highly optimized code (GPU DeePMD-kit) on the Summit supercomputer. Our code can efficiently scale up to the entire Summit supercomputer, attaining $91$ PFLOPS in double precision ($45.5\%$ of the peak) and {$162$/$275$ PFLOPS in mixed-single/half precision}. The great accomplishment of this work is that it opens the door to simulating unprecedented size and time scales with {\it ab initio} accuracy. It also poses new challenges to the next-generation supercomputer for a better integration of machine learning and physical modeling.

preprint2020arXiv

Raman Spectrum and Polarizability of Liquid Water from Deep Neural Networks

We introduce a scheme based on machine learning and deep neural networks to model the environmental dependence of the electronic polarizability in insulating materials. Application to liquid water shows that training the network with a relatively small number of molecular configurations is sufficient to predict the polarizability of arbitrary liquid configurations in close agreement with ab initio density functional theory calculations. In combination with a neural network representation of the interatomic potential energy surface,the scheme allows us to calculate the Raman spectra along 2-nanosecond classical trajectories at different temperatures for H2O and D2O. The vast gains in efficiency provided by the machine learning approach enable longer trajectories and larger system sizes relative to ab initio methods, reducing the statistical error and improving the resolution of the low-frequency Raman spectra. Decomposing the spectra into intramolecular and intermolecular contributions elucidates the mechanisms behind the temperature dependence of the low-frequency and stretch modes.

preprint2019arXiv

DP-GEN: A concurrent learning platform for the generation of reliable deep learning based potential energy models

In recent years, promising deep learning based interatomic potential energy surface (PES) models have been proposed that can potentially allow us to perform molecular dynamics simulations for large scale systems with quantum accuracy. However, making these models truly reliable and practically useful is still a very non-trivial task. A key component in this task is the generation of datasets used in model training. In this paper, we introduce the Deep Potential GENerator (DP-GEN), an open-source software platform that implements the recently proposed "on-the-fly" learning procedure [Phys. Rev. Materials 3, 023804] and is capable of generating uniformly accurate deep learning based PES models in a way that minimizes human intervention and the computational cost for data generation and model training. DP-GEN automatically and iteratively performs three steps: exploration, labeling, and training. It supports various popular packages for these three steps: LAMMPS for exploration, Quantum Espresso, VASP, CP2K, etc. for labeling, and DeePMD-kit for training. It also allows automatic job submission and result collection on different types of machines, such as high performance clusters and cloud machines, and is adaptive to different job management tools, including Slurm, PBS, and LSF. As a concrete example, we illustrate the details of the process for generating a general-purpose PES model for Cu using DP-GEN.

preprint2018arXiv

End-to-end Symmetry Preserving Inter-atomic Potential Energy Model for Finite and Extended Systems

Machine learning models are changing the paradigm of molecular modeling, which is a fundamental tool for material science, chemistry, and computational biology. Of particular interest is the inter-atomic potential energy surface (PES). Here we develop Deep Potential - Smooth Edition (DeepPot-SE), an end-to-end machine learning-based PES model, which is able to efficiently represent the PES for a wide variety of systems with the accuracy of ab initio quantum mechanics models. By construction, DeepPot-SE is extensive and continuously differentiable, scales linearly with system size, and preserves all the natural symmetries of the system. Further, we show that DeepPot-SE describes finite and extended systems including organic molecules, metals, semiconductors, and insulators with high fidelity.

preprint2018arXiv

Solving Many-Electron Schrödinger Equation Using Deep Neural Networks

We introduce a new family of trial wave-functions based on deep neural networks to solve the many-electron Schrödinger equation. The Pauli exclusion principle is dealt with explicitly to ensure that the trial wave-functions are physical. The optimal trial wave-function is obtained through variational Monte Carlo and the computational cost scales quadratically with the number of electrons. The algorithm does not make use of any prior knowledge such as atomic orbitals. Yet it is able to represent accurately the ground-states of the tested systems, including He, H2, Be, B, LiH, and a chain of 10 hydrogen atoms. This opens up new possibilities for solving large-scale many-electron Schrödinger equation.

preprint2017arXiv

Deep Potential: a general representation of a many-body potential energy surface

We present a simple, yet general, end-to-end deep neural network representation of the potential energy surface for atomic and molecular systems. This methodology, which we call Deep Potential, is "first-principle" based, in the sense that no ad hoc approximations or empirical fitting functions are required. The neural network structure naturally respects the underlying symmetries of the systems. When tested on a wide variety of examples, Deep Potential is able to reproduce the original model, whether empirical or quantum mechanics based, within chemical accuracy. The computational cost of this new model is not substantially larger than that of empirical force fields. In addition, the method has promising scalability properties. This brings us one step closer to being able to carry out molecular simulations with accuracy comparable to that of quantum mechanics models and computational cost comparable to that of empirical potentials.

Institution

Affiliation not imported yet

This author record came from a source that does not expose affiliation metadata. Once the author claims the profile or we enrich the record from another provider, this section will link to the concrete institution.

Topic footprint

Fields this researcher appears in

Source provenance

Where this author record came from

arxivconfidence 95%

external id: arxiv:2605.08142:author:3:linfeng-zhang

Imported May 20, 2026Synced May 21, 2026

arxivconfidence 95%

external id: arxiv:2605.08813:author:6:linfeng-zhang

Imported May 20, 2026Synced May 21, 2026

arxivconfidence 95%

external id: arxiv:2605.10906:author:13:linfeng-zhang

Imported May 20, 2026Synced May 21, 2026

arxivconfidence 95%

external id: arxiv:2605.17295:author:12:linfeng-zhang

Imported May 20, 2026Synced May 21, 2026

arxivconfidence 95%

external id: arxiv:2605.18165:author:4:linfeng-zhang

Imported May 20, 2026Synced May 21, 2026

arxivconfidence 95%

external id: arxiv:2605.20022:author:8:linfeng-zhang

Imported May 20, 2026Synced May 21, 2026

arxivconfidence 95%

external id: arxiv:2604.26365:author:9:linfeng-zhang