Sitemap
A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.
Pages
Posts
Happy Chinese New Year!
Published:
Happy Chinese New Year! Wish that all of us will make great progress in our studies and researches! 🤗
Built a Xi-Gundam
Published:
14 hours of work, it is so far the coolest Gundam (HGUC-\(\Xi\) Gumdan) I’ve ever built…Although with some little defects.
New paper on AI4Physics
Published:
My new AI4S paper has come out.🥳
Talk slides on my onging work
Published:
I gave a talk on my research project yesterday on group meeting, which is a little bit interesting…
Seems that my model skills have improved
Published:
I have tried assembling and painting a new model—Wheeljack in Bumblebee (2019), a.k.a. Transformers 6 (2019). This time the sketch lines, battle-damage and matte paint effects look well improved.🤗🤗🤗
新课题成果完成/还是在玩胶
Published:
重新把玩上起源电影的擎天柱了🥳把纯黑色塑料骨架做了一下金属干扫, 质感就舒服了非常多, 随便拍几张
最近拼的一些胶
Published:
昨晚把电镀彩透 RG 沙扎比拼完了, 熬了好几天夜😭, 白天干活, 晚上继续一坐四五个小时玩胶.
科研进展~
Published:
在 24 号的 post 里, 我发现了一个很奇怪的平衡极限环, 今天我应该差不多弄清楚它的机制了! 很嗨皮…
用摄像头捕捉手势实现酷炫的粒子效果!
Published:
今天干了个很有意思的东西, 通过摄像头追踪手势的握拳和张开, 来渲染图案的粒子效果, 建议在 PC 网页上浏览.
玩玩卡冈图雅黑洞可视化
Published:
很多人都知道星际穿越里的黑洞形象, Gemini 3.0 已经可以帮助我们很方便地可视化他了🤗
科研问题: 奇怪的极限环?
Published:
感觉 Su & Lai (2022b) 的文章现象学研究是有问题的…
神秘提示词
Published:
主播今天看见了一个帮助人拆分问题, 理解问题, 解决问题的优秀提示词:
动态&课题介绍
Published:
Gemini 3.0 今天发布了, 但是效果貌似没有预期的那么厉害…? 推理感觉都不如 o3, 尤其是 Canvas 功能, 感觉好不稳定…
动态
Published:
周四有李所的 “浦江创新论坛” 学术会议, 控告用户 @h4ppy_le 放了我的鸽子, 懒得喷🤗
publications
PRD: Power corrections in the determination of heavy meson LCDAs: A renormalon-based estimation
Published in Physical Review D, 2025
Abstract: At leading power accuracy the QCD light-cone distribution amplitudes (LCDAs) for a heavy meson can be matched onto the LCDAs in the framework of heavy-quark effective theory (HQET) through a factorization formula. We examine the power corrections to this factorization in the renormalon model, which can associate the power corrections originating from high-twist contributions to the divergent series in a matching kernel. Our analysis indicates that the dominant power corrections originate from the virtual part of the vertex bubble chain diagrams, which generate poles at w=n+12,∀n∈ℕ and w=1 in the Borel plane. Employing phenomenological models for both HQET and QCD LCDA, we present a numerical estimate. The results indicate that the power corrections in the peak region are approximately 22% for the D meson and 7% for the B–meson. These findings showcase the magnitude and the potential importance of power corrections in achieving high-precision phenomenological predictions for heavy mesons.
ICML 2026 accepted / Nat. Comm. submitted: PhysMaster: Building an Autonomous AI Physicist for Theoretical and Computational Physics Research
Published in ICML 2026 accepted / Nat. Comm. submitted, 2025
Abstract: Advances in LLMs have produced agents with knowledge and operational capabilities comparable to human scientists, suggesting potential to assist, accelerate, and automate research. However, existing studies mainly evaluate such systems on well-defined benchmarks or general tasks like literature retrieval, limiting their end-to-end problem-solving ability in open scientific scenarios. This is particularly true in physics, which is abstract, mathematically intensive, and requires integrating analytical reasoning with code-based computation. To address this, we propose PhysMaster, an LLM-based agent functioning as an autonomous theoretical and computational physicist. PhysMaster couples absract reasoning with numerical computation and leverages LANDAU, the Layered Academic Data Universe, which preserves retrieved literature, curated prior knowledge, and validated methodological traces, enhancing decision reliability and stability. It also employs an adaptive exploration strategy balancing efficiency and open-ended exploration, enabling robust performance in ultra-long-horizon tasks. We evaluate PhysMaster on problems from high-energy theory, condensed matter theory to astrophysics, including: (i) acceleration, compressing labor-intensive research from months to hours; (ii) automation, autonomously executing hypothesis-driven loops ; and (iii) autonomous discovery, independently exploring open problems.
ICML 2026 accepted / Nat. Comm. submitted: PRL-Bench: A Comprehensive Benchmark Evaluating LLMs’ Capabilities in Frontier Physics Research
Published in ICML 2026 accepted / Nat. Comm. submitted, 2026
Abstract: The paradigm of agentic science requires AI systems to conduct robust reasoning and engage in long-horizon, autonomous exploration. However, current scientific benchmarks remain confined to domain knowledge comprehension and complex reasoning, failing to evaluate the exploratory nature and procedural complexity of real-world research. In this work, we present research-oriented evaluations in theoretical and computational physics, a natural testbed with comprehensive domain knowledge, complex reasoning, and verifiable end-to-end workflows without reliance on experiments. Here we introduce PRL-Bench (Physics Research by LLMs), a benchmark designed to systematically map the capability boundaries of LLMs in executing end-to-end physics research. Constructed from 100 curated papers from the latest issues of Physical Review Letters since August 2025 and validated by domain experts, PRL-Bench covers five major theory- and computation-intensive subfields of modern physics: astrophysics, condensed matter physics, high-energy physics, quantum information, and statistical physics. Each task in the benchmark is designed to replicate the core properties of authentic scientific research, including exploration-oriented formulation, long-horizon workflows, and objective verifiability, thereby reconstructing the essential reasoning processes and research workflows of real physics research. Evaluation across frontier models shows that performance remains limited, with the best overall score below 50, revealing a pronounced gap between current LLM capabilities and the demands of real scientific research. PRL-Bench serves a reliable testbed for accessing next generation AI scientists advancing AI systems toward autonomous scientific discovery.
ApJ: Impact of Resonant Compton Scattering on Magnetar X-Ray Polarization with QED Vacuum Resonance
Published in The Astrophysical Journal, 2026
Abstract: Recent obeservations have revealed significant soft X-ray polarizations from several quiescent magnetars, including the intriguing 90∘ polarization angle (PA) swing as a function of photon energy for some sources. We present a general semi-analytical framework for calculating energy-dependent soft X-ray polarization signatures from magnetars, consistently incorporating both QED vacuum resonance in the atmosphere and resonant Compton scattering (RCS) in the magnetosphere. Starting from the polarized radiative transfer equation for RCS and treating vacuum-resonance-induced mode conversion as an input, we employ a first-order approximation in RCS optical depth to evaluate the effect of different magnetospheric plasma density (which depends on magnetic twist), drift velocity and temperature, and viewing geometry on the observed radiation. Our analysis reveals that magnetic twist and plasma drift velocity are the critical parameters controlling the impact of RCS on both the absolute polarization degree and its variation across the soft X-ray spectrum. We find that sufficiently strong RCS can wash out the PA swing caused by vacuum resonance. Furthermore, in addition to the QED vacuum resonance effect, significant relativistic signatures arising from plasma drift velocity (β0≳0.5) may introduce an extra 90∘ PA swing in the spectrum. Our calculation framework, based on single-scattering approximation, bypasses the need for complex, multi-dimensional Monte Carlo simulations, providing an analytical pathway for modeling full-surface emission and rotational-phase-resolved radiation from magnetic neutron stars, in support of current and future X-ray polarization missions.
ApJ to be submitted: Dynamics of Repeating Partial Tidal Disruption Event. I. Long-term Orbital Evolution with Dynamical Tides and Loss-cone Scattering
Published in ApJ to be submitted, 2026
Abstract: A star passing a supermassive black hole (SMBH) with a pericenter distance smaller than its tidal radius can be fully disrupted, producing a nuclear transient. However, on an elliptical orbit with a pericenter distance of $\sim 1-2$ tidal radii, the star may survive and undergo repeating partial tidal disruption events (rpTDEs) at intervals set by its orbital period. Here we develop an iterative map for the dynamical formation of rpTDEs through coupled loss-cone scattering and dynamical tides in single stars from nuclear star clusters, and simulate ensembles spanning different initial orbital parameters. We find that without stellar structural responses, orbits with initial periods longer than a few hundred years reach rpTDEs rapidly, whereas those with periods shorter than a few tens of years will become extreme mass ratio inspirals (EMRIs). Across the full parameter range, the median semimajor-axis shrinkage factor for rpTDEs is $\sim 2$. Including tides does not substantially increase the rpTDE formation rate relative to the pure loss-cone estimate. When stellar expansion due to accumulated dynamical tidal heating is included, systems initially expected to become EMRIs will reach rpTDEs earlier through high-eccentricity Roche-lobe overflow (HERLO). According to our results, the stellar initial conditions adopted in future simulations of short-period rpTDEs should account for the structural response to accumulated heating by dynamical tides over the preceding long-term evolution.
ApJ to be submitted: Beyond Cassini States: Analytical Framework for Non-Trivial Obliquity Steady States in Multi-Planetary Super-Earth Systems
Published in ApJ to be submitted, 2026
Abstract: Secular forcing in compact multiplanet systems can maintain large planetary obliquities even when stellar tides act to align a planet’s spin with its orbit. We develop an analytical framework for the nontrivial obliquity steady states previously found numerically for close-in super-Earths subject to two orbital-precession modes. In a frame co-precessing with the inner orbit, the dominant mode defines an integrable Colombo’s top, while the weaker mode provides periodic forcing and tides provide weak dissipation. Numerical integrations identify these states as stable periodic attractors, equivalent to dissipative limit cycles in the extended phase space. We associate the attractors with nonlinear $m:n$ commensurabilities between the intrinsic libration or circulation frequency and the difference between the orbital precession frequencies. Their phase-space locations follow from the resonance condition, while their existence further requires the cycle-averaged Hamiltonian changes produced by forcing and tides to balance. This balance locks the phase difference between the spin response and orbital forcing. We develop and test this formulation for the $1:1$ libration and circulation branches. For higher-order branches, the perturbative harmonic structure and scaling arguments provide a qualitative interpretation and suggest that progressively weaker dissipation is required. The nonlinear resonant attractors therefore provide a route to long-lived high-obliquity behavior beyond the classical Cassini states in compact planetary systems.
talks
Group meeting talk about onging work
Published:
This talk summarizes my main result in recent research project, journal article for MNRAS/APJ is in preparation…
BS thesis released
Published:
For those who are interested with my BS thesis, click the link below…