arXiv subjects

Minghao Guo

Publications and source records attributed to Minghao Guo.

At least 55 records · Page 3Linked to original sources

arXiv2025

TwinVoice: A Multi-dimensional Benchmark Towards Digital Twins via LLM Persona Simulation

Large Language Models (LLMs) are exhibiting emergent human-like abilities and are increasingly envisioned as the foundation for simulating an individual's communication style, behavioral tendencies, and personality traits. However, current evaluations of LLM-based persona simulation remain limited: most rely on synthetic dialogues, lack systematic frameworks, and lack analysis of the capability requirement. To address these limitations, we introduce TwinVoice, a comprehensive benchmark for assessing persona simulation across diverse real-world contexts. TwinVoice encompasses three dimensions: Social Persona (public social interactions), Interpersonal Persona (private dialogues), and Narrative Persona (role-based expression). It further decomposes the evaluation of LLM performance into six fundamental capabilities, including opinion consistency, memory recall, logical reasoning, lexical fidelity, persona tone, and syntactic style. Experimental results reveal that while advanced models achieve moderate accuracy in persona simulation, they still fall short of capabilities such as syntactic style and memory recall. Consequently, the average performance achieved by LLMs remains considerably below the human baseline.

cs.CL↗

arXiv2025

FlashBias: Fast Computation of Attention with Bias

Attention with bias, which extends standard attention by introducing prior knowledge as an additive bias matrix to the query-key scores, has been widely deployed in vision, language, protein-folding and other advanced scientific models, underscoring its status as a key evolution of this foundational module. However, introducing bias terms creates a severe efficiency bottleneck in attention computation. It disrupts the tightly fused memory-compute pipeline that underlies the speed of accelerators like FlashAttention, thereby stripping away most of their performance gains and leaving biased attention computationally expensive. Surprisingly, despite its common usage, targeted efficiency optimization for attention with bias remains absent, which seriously hinders its application in complex tasks. Diving into the computation of FlashAttention, we prove that its optimal efficiency is determined by the rank of the attention weight matrix. Inspired by this theoretical result, this paper presents FlashBias based on the low-rank compressed sensing theory, which can provide fast-exact computation for many widely used attention biases and a fast-accurate approximation for biases in general formalizations. FlashBias can fully take advantage of the extremely optimized matrix multiplication operation in modern GPUs, achieving 1.5$\times$ speedup for Pairformer in AlphaFold 3, and over 2$\times$ speedup for attention with bias in vision and language models without loss of accuracy. Code is available at this repository: https://github.com/thuml/FlashBias.

cs.LG↗

arXiv2025

Post Hoc Regression Refinement via Pairwise Rankings

Accurate prediction of continuous properties is essential to many scientific and engineering tasks. Although deep-learning regressors excel with abundant labels, their accuracy deteriorates in data-scarce regimes. We introduce RankRefine, a model-agnostic, plug-and-play post hoc method that refines regression with expert knowledge coming from pairwise rankings. Given a query item and a small reference set with known properties, RankRefine combines the base regressor's output with a rank-based estimate via inverse variance weighting, requiring no retraining. In molecular property prediction task, RankRefine achieves up to 10% relative reduction in mean absolute error using only 20 pairwise comparisons obtained through a general-purpose large language model (LLM) with no finetuning. As rankings provided by human experts or general-purpose LLMs are sufficient for improving regression across diverse domains, RankRefine offers practicality and broad applicability, especially in low-data settings.

cs.LG↗

arXiv2025

Leveraging Geometric Visual Illusions as Perceptual Inductive Biases for Vision Models

Contemporary deep learning models have achieved impressive performance in image classification by primarily leveraging statistical regularities within large datasets, but they rarely incorporate structured insights drawn directly from perceptual psychology. To explore the potential of perceptually motivated inductive biases, we propose integrating classic geometric visual illusions well-studied phenomena from human perception into standard image-classification training pipelines. Specifically, we introduce a synthetic, parametric geometric-illusion dataset and evaluate three multi-source learning strategies that combine illusion recognition tasks with ImageNet classification objectives. Our experiments reveal two key conceptual insights: (i) incorporating geometric illusions as auxiliary supervision systematically improves generalization, especially in visually challenging cases involving intricate contours and fine textures; and (ii) perceptually driven inductive biases, even when derived from synthetic stimuli traditionally considered unrelated to natural image recognition, can enhance the structural sensitivity of both CNN and transformer-based architectures. These results demonstrate a novel integration of perceptual science and machine learning and suggest new directions for embedding perceptual priors into vision model design.

cs.CV↗

arXiv2025

Newton to Einstein: Axiom-Based Discovery via Game Design

This position paper argues that machine learning for scientific discovery should shift from inductive pattern recognition to axiom-based reasoning. We propose a game design framework in which scientific inquiry is recast as a rule-evolving system: agents operate within environments governed by axioms and modify them to explain outlier observations. Unlike conventional ML approaches that operate within fixed assumptions, our method enables the discovery of new theoretical structures through systematic rule adaptation. We demonstrate the feasibility of this approach through preliminary experiments in logic-based games, showing that agents can evolve axioms that solve previously unsolvable problems. This framework offers a foundation for building machine learning systems capable of creative, interpretable, and theory-driven discovery.

cs.CE↗

arXiv2025

The Type Ia Supernova and Asymptotic Giant Branch Stellar Ejecta-regulated Interstellar Medium of Massive Galaxies

Observations and theory suggest that Type Ia supernovae (SNIa) heating and mass loss from asymptotic giant branch (AGB) stars play a crucial role in the interstellar medium (ISM) of massive galaxies. We perform 3D hydrodynamic simulations of the central few kiloparsecs of massive galaxies, including radiative cooling and mass and energy injection from AGB winds and SNIa (resolving each SNIa remnant, a few $\times10~\mathrm{pc}$ in size), excluding black hole feedback. We study systems with different initial core thermodynamic profiles, focusing on NGC 1399. Our simulations reproduce its observed density and entropy profiles well. Over $100~\mathrm{Myr}$, two steady-state profiles emerge, depending on the inner circumgalactic medium (CGM) pressure and the ratio of Ia heating to cooling: (i) if SNIa heating is less than cooling, a cooling flow develops; (ii) if SNIa heating is comparable to or exceeds cooling, SNIa heating drives a slow subsonic outflow of AGB ejecta, with black hole accretion at small radii. This outflow, pressure-confined by the CGM, adapts the ISM to the CGM properties: a low entropy CGM results in a dense, low entropy ISM with higher black hole accretion, while a high entropy CGM leads to a less dense, high entropy ISM with lower accretion. This suggests that the AGB-SNIa regulated ISM connects CGM and galaxy scales, potentially influencing black hole feedback in massive halos. Approximate methods of modeling Ia heating, such as clustered SNIa and smoothly distributed heating, produce unrealistic ISM profiles over $100~\mathrm{Myr}$, highlighting the importance of resolving SNIa in simulations.

astro-ph.GA↗

arXiv2025

Two-Stage Pretraining for Molecular Property Prediction in the Wild

Molecular deep learning models have achieved remarkable success in property prediction, but they often require large amounts of labeled data. The challenge is that, in real-world applications, labels are extremely scarce, as obtaining them through laboratory experimentation is both expensive and time-consuming. In this work, we introduce MoleVers, a versatile pretrained molecular model designed for various types of molecular property prediction in the wild, i.e., where experimentally-validated labels are scarce. MoleVers employs a two-stage pretraining strategy. In the first stage, it learns molecular representations from unlabeled data through masked atom prediction and extreme denoising, a novel task enabled by our newly introduced branching encoder architecture and dynamic noise scale sampling. In the second stage, the model refines these representations through predictions of auxiliary properties derived from computational methods, such as the density functional theory or large language models. Evaluation on 22 small, experimentally-validated datasets demonstrates that MoleVers achieves state-of-the-art performance, highlighting the effectiveness of its two-stage framework in producing generalizable molecular representations for diverse downstream properties.

cs.LG↗

arXiv2025

Cyclic Zoom: Multiscale GRMHD Modeling of Black Hole Accretion and Feedback

We present a ``cyclic zoom'' method to capture the dynamics of accretion flows onto black holes across a vast range of spatial and temporal scales in general relativistic magnetohydrodynamic (GRMHD) simulations. In this method, we cyclically zoom out (derefine) and zoom in (refine) the simulation domain while using a central mask region containing a careful treatment of the coarsened fluid variables to preserve the small-scale physics, in particular the magnetic field dynamics. The method can accelerate GRMHD simulations by $\gtrsim 10^5$ times for problems with large-scale separation. We demonstrate the validity of the technique using a series of tests, including spherically symmetric Bondi accretion, the Blandford-Znajek monopole, magnetized turbulent Bondi accretion, accretion of a magnetized rotating torus, and the long-term evolution of an accreting torus about both Schwarzschild and Kerr black holes. As applications, we simulate Bondi and rotating torus accretion onto black holes from galactic scales, covering an extremely large dynamic range. In Bondi accretion, the accretion rate is suppressed relative to the Bondi rate by $\sim(10r_\mathrm{g}/r_\mathrm{B})^{1/2}$ with a feedback power of $\sim 0.01 \dot{M} c^2$ for vanishing spin, and $\sim 0.1 \dot{M} c^2$ for spin $a\approx0.9$. In the long-term evolution of a rotating torus, the accretion rate decreases with time as $\dot{M}\propto t^{-2}$ on timescales much longer than the viscous timescale, demonstrating that our method can capture not only quasi-steady problems but also secular evolution. Our new method likewise holds significant promise for applications to many other problems that need to cover vast spatial and temporal scales.

astro-ph.HE↗

arXiv2025

Graphics4Science: Computer Graphics for Scientific Impacts

Computer graphics, often associated with films, games, and visual effects, has long been a powerful tool for addressing scientific challenges--from its origins in 3D visualization for medical imaging to its role in modern computational modeling and simulation. This course explores the deep and evolving relationship between computer graphics and science, highlighting past achievements, ongoing contributions, and open questions that remain. We show how core methods, such as geometric reasoning and physical modeling, provide inductive biases that help address challenges in both fields, especially in data-scarce settings. To that end, we aim to reframe graphics as a modeling language for science by bridging vocabulary gaps between the two communities. Designed for both newcomers and experts, Graphics4Science invites the graphics community to engage with science, tackle high-impact problems where graphics expertise can make a difference, and contribute to the future of scientific discovery. Additional details are available on the course website: https://graphics4science.github.io

cs.GR↗

arXiv2025

RobotSmith: Generative Robotic Tool Design for Acquisition of Complex Manipulation Skills

Endowing robots with tool design abilities is critical for enabling them to solve complex manipulation tasks that would otherwise be intractable. While recent generative frameworks can automatically synthesize task settings, such as 3D scenes and reward functions, they have not yet addressed the challenge of tool-use scenarios. Simply retrieving human-designed tools might not be ideal since many tools (e.g., a rolling pin) are difficult for robotic manipulators to handle. Furthermore, existing tool design approaches either rely on predefined templates with limited parameter tuning or apply generic 3D generation methods that are not optimized for tool creation. To address these limitations, we propose RobotSmith, an automated pipeline that leverages the implicit physical knowledge embedded in vision-language models (VLMs) alongside the more accurate physics provided by physics simulations to design and use tools for robotic manipulation. Our system (1) iteratively proposes tool designs using collaborative VLM agents, (2) generates low-level robot trajectories for tool use, and (3) jointly optimizes tool geometry and usage for task performance. We evaluate our approach across a wide range of manipulation tasks involving rigid, deformable, and fluid objects. Experiments show that our method consistently outperforms strong baselines in terms of both task success rate and overall performance. Notably, our approach achieves a 50.0\% average success rate, significantly surpassing other baselines such as 3D generation (21.4%) and tool retrieval (11.1%). Finally, we deploy our system in real-world settings, demonstrating that the generated tools and their usage plans transfer effectively to physical execution, validating the practicality and generalization capabilities of our approach.

cs.RO↗

arXiv2025

Genesis: A Compiler Framework for Hamiltonian Simulation on Hybrid CV-DV Quantum Computers

This paper introduces Genesis, the first compiler designed to support Hamiltonian Simulation on hybrid continuous-variable (CV) and discrete-variable (DV) quantum computing systems. Genesis is a two-level compilation system. At the first level, it decomposes an input Hamiltonian into basis gates using the native instruction set of the target hybrid CV-DV quantum computer. At the second level, it tackles the mapping and routing of qumodes/qubits to implement long-range interactions for the gates decomposed from the first level. Rather than a typical implementation that relies on SWAP primitives similar to qubit-based (or DV-only) systems, we propose an integrated design of connectivity-aware gate synthesis and beamsplitter SWAP insertion tailored for hybrid CV-DV systems. We also introduce an OpenQASM-like domain-specific language (DSL) named CVDV-QASM to represent Hamiltonian in terms of Pauli-exponentials and basic gate sequences from the hybrid CV-DV gate set. Genesis has successfully compiled several important Hamiltonians, including the Bose-Hubbard model, $\mathbb{Z}_2-$Higgs model, Hubbard-Holstein model, Heisenberg model and Electron-vibration coupling Hamiltonians, which are critical in domains like quantum field theory, condensed matter physics, and quantum chemistry. Our implementation is available at Genesis-CVDV-Compiler(https://github.com/ruadapt/Genesis-CVDV-Compiler).

quant-ph↗

arXiv2025

TetSphere Splatting: Representing High-Quality Geometry with Lagrangian Volumetric Meshes

We introduce TetSphere Splatting, a Lagrangian geometry representation designed for high-quality 3D shape modeling. TetSphere splatting leverages an underused yet powerful geometric primitive -- volumetric tetrahedral meshes. It represents 3D shapes by deforming a collection of tetrahedral spheres, with geometric regularizations and constraints that effectively resolve common mesh issues such as irregular triangles, non-manifoldness, and floating artifacts. Experimental results on multi-view and single-view reconstruction highlight TetSphere splatting's superior mesh quality while maintaining competitive reconstruction accuracy compared to state-of-the-art methods. Additionally, TetSphere splatting demonstrates versatility by seamlessly integrating into generative modeling tasks, such as image-to-3D and text-to-3D generation.

cs.CV↗

arXiv2025

Galactic-scale Feeding Reveals Warped Hypermagnetized Multiphase Circumbinary Accretion Around Supermassive Black Hole Binaries

Supermassive black hole (SMBH) binaries in gaseous and stellar environments are prime targets for next-generation space-based gravitational wave detectors. Yet, realistic accretion conditions under which these binary systems evolve are not fully understood. In this work, we demonstrate the hypermagnetized multi-phase nature of the surrounding accretion flow formed by large-scale feeding from a galaxy background. Our simulations indicate that the hypermagnetized circumbinary disk is eccentric and warped, hosting a hot gas core for a parsec-scale separated binary. We also observe collimated bipolar magnetic tower-like outflows launched from each SMBH.

astro-ph.HE↗

arXiv2025

Procedural Synthesis of Synthesizable Molecules

Designing synthetically accessible molecules and recommending analogs to unsynthesizable molecules are important problems for accelerating molecular discovery. We reconceptualize both problems using ideas from program synthesis. Drawing inspiration from syntax-guided synthesis approaches, we decouple the syntactic skeleton from the semantics of a synthetic tree to create a bilevel framework for reasoning about the combinatorial space of synthesis pathways. Given a molecule we aim to generate analogs for, we iteratively refine its skeletal characteristics via Markov Chain Monte Carlo simulations over the space of syntactic skeletons. Given a black-box oracle to optimize, we formulate a joint design space over syntactic templates and molecular descriptors and introduce evolutionary algorithms that optimize both syntactic and semantic dimensions synergistically. Our key insight is that once the syntactic skeleton is set, we can amortize over the search complexity of deriving the program's semantics by training policies to fully utilize the fixed horizon Markov Decision Process imposed by the syntactic template. We demonstrate performance advantages of our bilevel framework for synthesizable analog generation and synthesizable molecule design. Notably, our approach offers the user explicit control over the resources required to perform synthesis and biases the design space towards simpler solutions, making it particularly promising for autonomous synthesis platforms. Code is at https://github.com/shiningsunnyday/SynthesisNet.

q-bio.BM↗

arXiv2025

HST Observations within the Sphere of Influence of the Powerful Supermassive Black Hole in PKS0745-191

We present Space Telescope Imaging Spectrograph observations from the Hubble Space Telescope of the supermassive black hole (SMBH) at the center of PKS0745-191, a brightest cluster galaxy (BCG) undergoing powerful radio-mode AGN feedback ($P_{\rm cav}\sim5\times10^{45}$ erg s$^{-1}$). These high-resolution data offer the first spatially resolved map of gas dynamics within a SMBHs sphere of influence under such powerful feedback. Our results reveal the presence of highly chaotic, non-rotational ionized gas flows on sub-kpc scales, in contrast to the more coherent flows observed on larger scales. While radio-mode feedback effectively thermalizes hot gas in galaxy clusters on kiloparsec scales, within the core, the hot gas flow may decouple, leading to a reduction in angular momentum and supplying ionized gas through cooling, which could enhance accretion onto the SMBH. This process could, in turn, lead to a self-regulating feedback loop. Compared to other BCGs with weaker radio-mode feedback, where rotation is more stable, intense feedback may lead to more chaotic flows, indicating a stronger coupling between jet activity and gas dynamics. Additionally, we observe a sharp increase in velocity dispersion near the nucleus, consistent with a very massive $M_{\rm BH}\sim1.5\times10^{10} M_\odot$ SMBH. The density profile of the ionized gas is also notably flat, paralleling the profiles observed in X-ray gas around galaxies where the Bondi radius is resolved. These results provide valuable insights into the complex mechanisms driving galaxy evolution, highlighting the intricate relationship between SMBH fueling and AGN feedback within the host galaxy.

astro-ph.GA↗

arXiv2024

Physically Compatible 3D Object Modeling from a Single Image

We present a computational framework that transforms single images into 3D physical objects. The visual geometry of a physical object in an image is determined by three orthogonal attributes: mechanical properties, external forces, and rest-shape geometry. Existing single-view 3D reconstruction methods often overlook this underlying composition, presuming rigidity or neglecting external forces. Consequently, the reconstructed objects fail to withstand real-world physical forces, resulting in instability or undesirable deformation -- diverging from their intended designs as depicted in the image. Our optimization framework addresses this by embedding physical compatibility into the reconstruction process. We explicitly decompose the three physical attributes and link them through static equilibrium, which serves as a hard constraint, ensuring that the optimized physical shapes exhibit desired physical behaviors. Evaluations on a dataset collected from Objaverse demonstrate that our framework consistently enhances the physical realism of 3D models over existing methods. The utility of our framework extends to practical applications in dynamic simulations and 3D printing, where adherence to physical compatibility is paramount.

cs.CV↗

arXiv2024

A Case for Quantum Circuit Cutting for NISQ Applications: Impact of topology, determinism, and sparsity

We make the case that variational algorithm ansatzes for near-term quantum computing are well-suited for the quantum circuit cutting strategy. Previous demonstrations of circuit cutting focused on the exponential execution and postprocessing costs due to the cuts needed to partition a circuit topology, leading to overly pessimistic evaluations of the approach. This work observes that the ansatz Clifford structure and variational parameter pruning significantly reduce these costs. By keeping track of the limited set of correct subcircuit initializations and measurements, we reduce the number of experiments needed by up to 16x, matching and beating the error mitigation offered by classical shadows tomography. By performing reconstruction as a sparse tensor contraction, we scale the feasible ansatzes to over 200 qubits with six ansatz layers, beyond the capability of prior work.

quant-ph↗

arXiv2024

Evolution of Supernova Remnants in a Cloudy Multiphase Interstellar Medium

We investigate the evolution of supernova remnants (SNRs) in a two-phase cloudy medium by performing a series of high-resolution (up to $Δx\approx0.01\,\mathrm{pc}$), 3D hydrodynamical simulations including radiative cooling and thermal conduction. We aim to reach a resolution that directly captures the shock-cloud interactions for the majority of the clouds initialized by the saturation of thermal instability. In comparison to the SNR in a uniform medium with the volume filling warm medium, the SNR expands similarly (following $\propto t^{2/5}$) but sweeps up more mass as the cold clouds contribute before shocks in the warm medium become radiative. However, the SNR in a cloudy medium continuously loses energy after shocks toward the cold clouds cool, resulting in less hot gas mass, thermal energy, and terminal momentum. Thermal conduction has little effect on the dynamics of the SNR but smooths the morphology and modifies the internal structure by increasing the density of hot gas by a factor of $\sim 3-5$. The simulation results are not fully consistent with many previous 1D models describing the SNR in a cloudy medium including a mass loading term. By direct measurement in the simulations, we find that, apart from the mass source, the energy sink is also important with a spatially flat cooling rate $\dot{e}\propto t^{-11/5}$. As an illustration, we show an example 1D model including both mass source and energy sink terms (in addition to the radiative cooling in the volume filling component) that better describes the structure of the simulated SNR.

astro-ph.GA↗