Searcharxiv⌕ Search

arXiv subjects

Tianyu Zhu

Publications and source records attributed to Tianyu Zhu.

At least 19 recordsLinked to original sources

Surface-bulk magnetic dichotomy in intercalated transition metal dichalcogenides

Magnetically intercalated transition metal dichalcogenides are layered materials with highly tunable magnetic orders and unconventional transport properties. Here we report a surface-bulk magnetic dichotomy in $\mathrm{M}_{1/3}\mathrm{T}\mathrm{X}_{2}$, in which the surface and bulk favor distinct magnetic orders. In $\mathrm{Fe}_{1/3}\mathrm{Nb}\mathrm{S}_{2}$, angle-resolved photoemission spectroscopy (ARPES) results are reproduced by an A-type antiferromagnetic electronic structure, while neutron scattering and bulk DFT+$U$ calculations identify stripe order as the bulk ground state. DFT+$U$ slab calculations resolve this contradiction by showing that relaxation of an Fe-terminated surface reverses the magnetic preference from stripe to A-type order. The same dichotomy recovers the $K$-centered Fermi-surface pockets that are observed in $\mathrm{Co}_{1/3}\mathrm{Nb}\mathrm{S}_{2}$ and $\mathrm{Co}_{1/3}\mathrm{Ta}\mathrm{S}_{2}$ but absent from previous bulk calculations. Notably, the surface-stabilized magnetic phase of $\mathrm{Fe}_{1/3}\mathrm{Nb}\mathrm{S}_{2}$ carries nonrelativistic spin splitting and can exhibit altermagnetic character. These results caution against the general association of ARPES signals with bulk magnetic states in these systems, and suggest surface-bulk magnetic dichotomy as a promising route to realize magnetic phases, including altermagnetism, that are not energetically favored in the bulk.

cond-mat.mtrl-sci↗

ACE-CaP-0: Real-World Code-as-Policies in the Era of Experience

Recent advances in frontier models have pushed the boundaries of agentic robot control, enabling agents to perform complex robotic tasks through reasoning and interaction. However, repeated calls to these models during execution can substantially increase task completion time. Code as Policies (CaP) offers a route for coding agents to acquire physical skills through experience and retain them in executable programs, allowing the reasoning invested in development to be reused across runs. Developing reliable code policies while limiting the cost of physical experimentation remains a central challenge. We investigate human guidance as a source of priors for feasible actions and simulation as an auxiliary environment for exploring and testing candidate policies. Our findings favor robot demonstrations available on request and simulation used selectively during real-world policy refinement. These findings inform ACE-CaP-0, an agentic recipe that couples policy revision with experience acquisition through Demonstrations on Request and Simulation on Demand. Across eight real-world manipulation tasks spanning alignment, multi-object handling, and bimanual coordination, the resulting frozen programs achieve 92.1\% success (221/240 trials). Mean execution times across tasks range from 0.97 to 2.27 minutes. The programs execute with runtime perception and control, without a human or coding agent in the loop. These results highlight CaP's potential to turn experience into reliable, reusable physical skills.

cs.RO↗

Outcome-Routed Distillation and Reinforcement Learning for Diffusion Language Models

On-policy distillation provides dense supervision on the student's own samples, but its targets still come from the teacher. Verifier reinforcement learning (RL) instead uses an outcome signal, but it provides no direct constraint on capabilities outside the verifier. Combining the two is difficult for diffusion language models (dLLMs), because bidirectional denoising makes the policy likelihoods required by ratio-based RL unreliable. We develop a likelihood-free pipeline for a mixture-of-experts dLLM. Each rollout group is routed by verifier outcome. Mixed groups use ratio-free, group-centred REINFORCE. If every student rollout fails but a teacher succeeds, the model receives dense KL supervision on the student's denoising canvases. All other groups are skipped. A KL term anchored to the distilled checkpoint limits changes outside the verifier reward. The tool-calling teacher succeeds on only 18.6% of the training prompts. At 24 denoising steps, distillation exceeds the teacher by 6.77 percentage points, and routing improves the distilled checkpoint by a further 18.31 points. Routing lowers reasoning-trace accuracy by 0.71 points but raises direct-answer accuracy on the same items by 13.21 points. The anchored arm restores trace accuracy. Its tool-calling score differs by -0.16 points, with a confidence interval that includes zero, although this comparison also removes a small fallback distillation route. A domain-conditional recipe retains a 16.37-point tool-calling gain; a second seed gives 18.23 points and reproduces both preservation results, but the math result is seed-dependent. We pre-registered every arm and evaluated each comparison on paired items in a single sweep on one machine.

cs.LG↗

Context-Tower Conversion Preserves Generation While Freezing Retains Knowledge: Low-Budget AR-to-Diffusion Conversion of MoE LLMs

Converting a pretrained autoregressive (AR) model to a diffusion language model (dLLM) enables parallel generation without pretraining a new model. Published conversion methods differ by roughly three orders of magnitude in training data and have not been compared under a common protocol. We compare two conversions of the same 30B Mixture-of-Experts (MoE) parent, holding the corpus, supervised-token budget, trainable parameter set and evaluation harness fixed, each under its own training recipe. The in-place model updates a subset of the parent's weights using denoising and representation-alignment losses; the frozen-tower model instead conditions through cross-attention on a frozen causal copy of the parent. With 1B training tokens, the frozen-tower model scores 71.60 on HumanEval pass@10 against 6.19 for the in-place model, an 11.6x improvement. At the same budget it also keeps 95% of the parent's GSM8K score and 99% of its MMLU-Pro score. A dense-parent experiment reproduces the HumanEval separation. Within the two-tower design at about 500M tokens, freezing the context tower retains substantially more MMLU-Pro performance than training it, while both give similar observed HumanEval scores. Our theoretical analysis establishes that both conversion classes contain an exact sampler for the AR parent under a hard attention mask and left-to-right commitment of one position per round. Under a shared loss, freezing removes the gradient contribution through the context states. Furthermore, evaluation protocol substantially affects a published 500B-token conversion's scores in both directions across tasks, while its AR parent's scores vary by less than three points, so comparing dLLMs needs a common protocol. These results show that, in the tested low-budget regime, the frozen-tower configuration retains substantially more of the parent's generation performance than in-place conversion.

cs.LG↗

Mitigating Popularity Bias in Recommendation with Global Listwise Learning and Progressive Bi-Weighting

In recommender systems, user feedback typically follows a long-tail distribution, which leads many recommendation algorithms to exacerbate popularity bias by disproportionately favoring popular items. To mitigate this issue, recent studies have employed Inverse Propensity Scoring (IPS) to rebalance training data via reweighting user-item interactions. However, the effectiveness of IPS-based approaches is often constrained by locally unbiased objectives and inaccurate propensity estimation. In this paper, we propose Multinomial Likelihood with Bi-Weighting (Mult-BiW) to address these limitations. First, we introduce a debiasing framework, termed Mult-IPS, which integrates multinomial likelihood with IPS to capture global and unbiased user preferences over the entire item set. Second, we develop a Bi-Weighting (BiW) strategy that jointly leverages propensity scores and a collection model, incorporating a smoothing mechanism to enhance the robustness of propensity estimation. We further provide theoretical analyses that establish an upper bound on the empirical bias and characterize the optimal form of the collection model. Third, to mitigate the adverse effects of aggressive reweighting on representation learning, we design a Progressive Bi-Weighting strategy that gradually transitions from discriminative representation learning to popularity debiasing. Extensive experiments on real-world datasets show that Mult-BiW consistently outperforms state-of-the-art baselines.

cs.IR↗

MolLangData: A Large-Scale Dataset for Molecular Structure-Language Description via a Rule-Regularized Method

Molecular function is largely determined by structure. Accurately aligning molecular structure with natural language is therefore essential for enabling large language models (LLMs) to both understand structure for chemical reasoning and generate molecules from natural-language design intent. However, the substantial cost of human annotation makes it infeasible to construct large-scale, high-quality datasets of structure-grounded descriptions. This work proposes a fully automated annotation framework for generating precise molecular descriptions at scale, such that the original molecule can be unambiguously reconstructed from the description alone. Our approach extends a rule-based chemical nomenclature parser to interpret IUPAC names and construct enriched, XML metadata that explicitly encodes molecular structure. This is then used to guide LLMs in producing accurate natural-language descriptions. Using this framework, we curate MolLangData, a dataset of approximately $163$k molecule--description pairs. A rigorous validation protocol combining expert human and LLM-based reconstruction on a subset of $2,000$ molecules demonstrates $98.6\%$ description precision. Using the curated dataset, we train a $4$B-parameter LLM via large-scale reinforcement learning for language-conditional molecule generation. The proposed framework and dataset provide a reliable foundation for molecule--language alignment, readily beneficial to broader chemical tasks.

cs.CL↗

Binarized High-Efficiency RAW Video Restoration and Beyond

RAW video restoration is fundamental to high-quality low-level perception and serves as the basis for a wide range of downstream vision applications. While binary neural networks (BNNs) enable efficient lightweight deployment for image enhancement, their deficiencies in modeling temporal coherence and activation value distributions hinder their effectiveness when applied to video scenarios. In this paper, we propose BinRVR, a binarized RAW video restoration framework that reduces computation and parameters by approximately 96% while incurring only about 4% performance degradation. Specifically, we present a Binarized Information Interaction Module (BIIM) to jointly model spatial and temporal information in an efficient and unified manner. Moreover, we develop a Distribution-Aware Binarized Convolution (DAB-Conv) that leverages the statistics of full-precision activations to mitigate quantization errors. The proposed framework further supports multi-bit quantization, enabling flexible accuracy-efficiency trade-offs across different hardware constraints. Extensive experiments demonstrate that our BinRVR achieves competitive performance compared with state-of-the-art binarized methods on RAW video restoration tasks, including low-light enhancement, denoising, deblurring, and super-resolution. We further explore the potential of our method on downstream video applications, including object detection and monocular depth estimation.

cs.CV↗

Resolving Finite-Size Errors in EOM-CCSD Band Gaps of Solids with Interacting-Bath Dynamical Embedding Theory

Periodic equation-of-motion coupled-cluster theory with single and double excitations (EOM-CCSD) has shown promise for quantitative calculations of band structures in solids. However, its steep computational scaling has limited calculations to relatively coarse $k$-point meshes, leading to sizable finite-size errors and discrepant estimates of thermodynamic-limit band gaps in recent benchmarks. In this work, we revisit EOM-CCSD band gaps for ten semiconductors and insulators using interacting-bath dynamical embedding theory (ibDET), a systematically improvable Green's function embedding framework that enables dense Brillouin-zone sampling at modest computational cost. By pushing the $k$-point sampling up to $10\times10\times10$, well beyond the system sizes accessible in canonical periodic EOM-CCSD calculations, we significantly reduce finite-size errors and obtain stable thermodynamic-limit extrapolations. We further compare $G_0W_0$@PBE, $G_0W_0$@HF, and EOM-CCSD on an equal footing using the same numerical settings in PySCF. We find that EOM-CCSD yields a mean absolute error of 0.32 eV relative to experimental band gaps for a test set of ten semiconductors and insulators, lower than that of $G_0W_0$@PBE. For ZnO, EOM-CCSD also accurately describes the Zn $3d$-band binding energy, despite overestimating the band gap. These results demonstrate that ibDET offers a practical route to high-accuracy many-body electronic structure calculations in periodic systems.

cond-mat.mtrl-sci↗

Interpolative Separable Density-Fitting for Transcorrelated Hamiltonians

The transcorrelated (TC) method dramatically accelerates the convergence of correlated calculations toward the complete-basis-set (CBS) limit by folding a Jastrow correlator into the Hamiltonian via a similarity transformation, incorporating the electron--electron cusp into the effective interaction. We make the TC framework practical for large systems and flexible, multi-center correlators by compressing the grid-evaluated TC integrals with the interpolative separable density-fitting (ISDF) approximation, combined with the effective two-body (xTC) treatment of the three-body operator. This low-rank representation reduces storage and integration costs by orders of magnitude, and a multi-GPU implementation with automatic differentiation of the correlator makes the construction routine for large basis sets. We demonstrate the resulting ISDF-xTC-CCSD method on the linear hydrogen chain, reaching the joint thermodynamic and CBS limits with basis sets up to cc-pV5Z in agreement with state-of-the-art many-body references to within about 1~mHa/atom, and on the benzene ground-state energy with up to 1200 orbitals (cc-pCV5Z), where the method attains state-of-the-art accuracy at the coupled cluster singles and doubles level and its CBS extrapolation is markedly more robust than that of conventional coupled-cluster methods.

physics.chem-ph↗

ACE-Brain-0.5: A Unified Embodied Foundational Model for Physical Agentic AI

Embodied AI is moving from isolated perception or action modules toward physical agents that understand, plan under goals, act through robot bodies, monitor progress, and improve from experience. Existing systems address this loop only in parts: end-to-end policies generate actions but often lack spatial reasoning, planning, and execution assessment, while robot-agent systems orchestrate tools or specialists but do not learn a shared representation. This fragmentation limits general Physical Agentic AI. We present ACE-Brain-0.5, a unified embodied foundation model that organizes robot intelligence into five coupled functions: spatial perception, decision making, embodied interaction, self-monitoring, and self-improvement. Built on ACE-Brain-0, which established spatial intelligence as a shared scaffold across robot platforms, ACE-Brain-0.5 extends an understanding-centric model into a closed-loop foundation model. A single 8B backbone instantiates the first four functions: grounding objects and affordances, reasoning over 3D and egocentric spatial relations, decomposing instructions into subgoals, generating navigation and manipulation actions, and estimating progress for verification and recovery. To unify these capabilities without cross-task interference, we introduce SSR+, which extends Scaffold-Specialize-Reconcile with a Reactivate stage after task-vector merging. The fifth function, self-improvement, is realized by a companion framework that updates external execution state, including task schemas, spatial memory, and failure-recovery cases, from rollouts. Across fifteen benchmarks, ACE-Brain-0.5 improves over ACE-Brain-0 on 14 of 18 spatial perception and grounding benchmarks, achieves competitive navigation and manipulation performance, and provides strong progress estimation in ID and OOD settings. Together, these results mark an early step toward general Physical Agentic AI.

cs.RO↗

Boosting Text-Driven Video Segmentation via Geometry-Aware Distillation

Text-driven Referring Video Object Segmentation (RVOS) aims to locate and segment target objects in videos given natural language. However, existing models are typically trained on 2D image or video datasets with naive segmentation losses, which overlooks the geometric consistency across frames and leads to weak spatial understanding. In this paper, we propose Geometry-enhanced Language-guided Video segmentation (GeoLaV), a two-stage framework that distills 3D geometric knowledge from images to enhance text-driven video segmentation. In the first stage, we perform monocular geometry pretraining with monocular novel-view synthesis, enabling the model to acquire geometry-consistent visual representations via spatial alignment on large-scale single-image datasets. In the second stage, we introduce geometry-aware distillation and fine-tune the model on video segmentation datasets, transferring 3D structural knowledge from a general 3D prior model. This process reinforces 3D awareness and improves both spatiotemporal coherence and language grounding in segmentation. Extensive experiments show that our method using only image segmentation data already provides notable zero-shot generalization in RVOS. When combined with geometry-aware distillation for fine-tuning on videos, our method achieves state-of-the-art performance across multiple RVOS benchmarks. The code is available at https://github.com/Tony1882880/GeoLaV.

cs.CV↗

Low-Scaling Many-Body Green's Function Calculations for Molecular Systems via Interacting-Bath Dynamical Embedding Theory

We present a molecular extension of our recently proposed Green's function embedding method, interacting-bath dynamical embedding theory (ibDET), for computing charged excitation energies at the $GW$ and EOM-CCSD levels. Starting from atom-centered impurities, we construct bath representations that capture the frequency-dependent entanglement between the impurity and its environment and can be systematically improved via the construction of cluster-specific natural orbitals. Utilizing a $GW$ or coupled-cluster Green's function solver, the self-energy of the full system is assembled from all embedding problems to obtain the interacting Green's function. We show that ibDET provides accurate spectral properties with much reduced cost for a broad range of systems, including conjugated molecules and nanoclusters. Compared with full-system results, the errors in the predicted ionization potentials and electron affinities are around 0.1 eV or smaller, while each embedding problem includes only a small fraction of the total orbital space. This work provides an efficient and scalable framework for computing spectral properties of molecular systems.

physics.chem-ph↗

Transferable Machine Learning of Electronic Hamiltonians with Superposition-of-Atomic-Potentials Features

Machine learning (ML) of electronic Hamiltonians offers a unified route to electronic wave functions and physical observables. We introduce a Hamiltonian learning framework built on electronic features derived from the superposition-of-atomic-potentials (SAP) approximation, an efficient self-consistent-field initial guess that captures essential electron-electron screening. SAP quantities define a symmetry-adapted intrinsic atomic orbital learning basis and provide physics-informed inputs to an orbital-based graph neural network that predicts converged Kohn-Sham Fock matrices. To extend the approach to larger basis sets, we further develop a downfolding scheme that predicts large-basis electronic structure from minimal-basis features. On the QM9 dataset, the model accurately reproduces frontier and core orbital energies, dipole moments, and the full density of states. For organic charge-transport materials, it yields accurate intermolecular transfer integrals for benzene, tetracyanoquinodimethane (TCNQ), and tetrathiafulvalene (TTF) dimers, and transfers to unseen substituted-benzene heterodimers with a mean absolute error of 4.8 meV. These results establish SAP-based ML of electronic Hamiltonians as a transferable and scalable tool for high-throughput electronic-structure prediction.

physics.chem-ph↗

Integrating Weather Foundation Model and Satellite to Enable Fine-Grained Solar Irradiance Forecasting

Accurate day-ahead solar irradiance forecasting is essential for integrating solar energy into the power grid. However, it remains challenging due to the pronounced diurnal cycle and inherently complex cloud dynamics. Current methods either lack fine-scale resolution (e.g., numerical weather prediction, weather foundation models) or degrade at longer lead times (e.g., satellite extrapolation). We propose Baguan-solar, a two-stage multimodal framework that fuses forecasts from Baguan, a global weather foundation model, with high-resolution geostationary satellite imagery to produce 24-hour irradiance forecasts at kilometer scale. Its decoupled two-stage design first forecasts day-night continuous intermediates (e.g., cloud cover) and then infers irradiance, while its modality fusion jointly preserves fine-scale cloud structures from satellite and large-scale constraints from Baguan forecasts. Evaluated over East Asia using CLDAS as ground truth, Baguan-solar outperforms strong baselines (including ECMWF IFS, vanilla Baguan, and SolarSeer), reducing RMSE by 16.08% and better resolving cloud-induced transients. An operational deployment of Baguan-solar has supported solar power forecasting in an eastern province in China, since July 2025. Our code is accessible at https://github.com/DAMO-DI-ML/Baguan-solar.git.

cs.LG↗

The Python Simulations of Chemistry Framework: 10 years of an open-source quantum chemistry project

Over the past decade, the Python-based Simulations of Chemistry Framework (PySCF) has developed into a widely used open-source platform for electronic structure theory and quantum chemical method development. This article reviews the major advances since the previous overview in 2020, covering new modules and methodology, infrastructure changes, and performance benchmarks.

physics.chem-ph↗

MolLangBench: A Comprehensive Benchmark for Language-Prompted Molecular Structure Recognition, Editing, and Generation

Precise recognition, editing, and generation of molecules are essential prerequisites for both chemists and AI systems tackling various chemical tasks. We present MolLangBench, a comprehensive benchmark designed to evaluate fundamental molecule-language interface tasks: language-prompted molecular structure recognition, editing, and generation. To ensure high-quality, unambiguous, and deterministic outputs, we construct the recognition tasks using automated cheminformatics tools, and curate editing and generation tasks through rigorous expert annotation and validation. MolLangBench supports the evaluation of models that interface language with different molecular representations, including linear strings, molecular images, and molecular graphs. Evaluations of state-of-the-art models reveal significant limitations: the strongest model (GPT-5) achieves $86.2\%$ and $85.5\%$ accuracy on recognition and editing tasks, which are intuitively simple for humans, and performs even worse on the generation task, reaching only $43.0\%$ accuracy. These results highlight the shortcomings of current AI systems in handling even preliminary molecular recognition and manipulation tasks. We hope MolLangBench will catalyze further research toward more effective and reliable AI systems for chemical applications.The dataset and code can be accessed at https://huggingface.co/datasets/ChemFM/MolLangBench and https://github.com/TheLuoFengLab/MolLangBench, respectively.

cs.CL↗

LibppRPA: An Open-Source Library for Particle-Particle Random Phase Approximation

The accurate description of electron correlation and excitation energies remains a fundamental challenge in quantum chemistry. The particle-particle random phase approximation (ppRPA) has emerged as a promising method for capturing a broad range of excited-state properties. However, the implementation of ppRPA has been largely limited to in-house software, restricting its accessibility and usability. In this work, we present LibppRPA, an open-source and lightweight Python library designed for efficient and flexible ppRPA calculations of (1) electronic excitation energy and its associated analytical gradients and (2) the ground state correlation energy, and its associated analytical gradients. LibppRPA enables seamless integration with existing quantum chemistry packages, such as PySCF, by utilizing occupation numbers, molecular orbital coefficients, and three-center electron repulsion integrals. We implement both direct diagonalization and the iterative Davidson algorithm for solving the ppRPA equations, as well as active-space approximations, allowing users to balance accuracy and computational efficiency. We demonstrate the performance of LibppRPA through benchmark calculations on singlet-triplet gaps, double excitations, charge-transfer excitations, and valence/Rydberg excitations, showcasing its reliability across diverse molecular systems. The library provides a robust platform for studying electronic excitations and offers new opportunities for future developments in electronic structure theory.

physics.chem-ph↗

ChemFM as a Scaling Law Guided Foundation Model Pre-trained on Informative Chemicals

Traditional AI methods often rely on task-specific model designs and training, which constrain both the scalability of model size and generalization across different tasks. Here, we introduce ChemFM, a large foundation model specifically developed for chemicals. By conducting a series of scaling experiments, we identify UniChem as the informative molecular database for pre-training the foundation model. ChemFM comprises 3 billion parameters and is pre-trained on 178 million molecules using self-supervised causal language modeling to extract generalizable molecular representations. This model can be adapted to diverse downstream chemical applications using either full-parameter or parameter-efficient fine-tuning methods. ChemFM consistently outperforms state-of-the-art task-specific AI models across all tested tasks. Notably, it achieves up to 67.48% performance improvement across 34 property prediction benchmarks, up to 33.80% reduction in mean average deviation between conditioned and actual properties of generated molecules in conditional molecular generation tasks, and up to 3.7% top-1 accuracy improvement across 4 reaction prediction datasets. Moreover, ChemFM demonstrates its superior performance in predicting antibiotic activity and cytotoxicity, highlighting its potential to advance the discovery of novel antibiotics. Furthermore, we demonstrate that, as a foundation model, ChemFM exhibits strong data efficiency, requiring significantly fewer labeled training samples to achieve state-of-the-art performance. We anticipate that ChemFM will significantly advance chemistry research by providing a foundation model capable of effectively generalizing across a broad range of tasks with minimal additional training.

cs.CE↗