Searcharxiv⌕ Search

arXiv subjects

Jihao Liu

Publications and source records attributed to Jihao Liu.

At least 19 recordsLinked to original sources

Non-algebraicity of non-abundant foliations and abundance for adjoint foliated structures

Assuming the abundance conjecture in dimension $d$, we establish a non-algebraicity criterion of foliations: any log canonical foliation of rank $\le d$ with $ν\neqκ$ is not algebraically integrable, answering question of Ambro--Cascini--Shokurov--Spicer. Under the same hypothesis, we prove abundance for klt algebraically integrable adjoint foliated structures of dimension $\le d$ and show the existence of good minimal models or Mori fiber spaces. In particular, when $d=3$, all these results hold unconditionally. Using similar arguments, we solve a problem proposed by Lu and Wu on abundance of surface adjoint foliated structures that are not necessarily algebraically integrable.

math.AG↗

Large Language Models as a Universal Searcher for Open-World Automated Machine Learning

Developing effective machine learning models requires complex manual tuning, while current Automated Machine Learning (AutoML) frameworks remain domain-specific and isolate model architecture, training hyperparameter, and data ratio optimization. Here, we bridge this gap by proposing a universal, large language model (LLM)-based AutoML framework that jointly optimizes these configurations for the first time without the need for task-specific fine-tuning. We propose NCode, a unified numerical encoding scheme that translates diverse AutoML design choices into universal discrete sequences and effectively reframes the optimization process as a unified ranking task. Trained solely on optimization trajectories derived from a limited, single-domain vision task, the proposed LLM-based searcher demonstrates impressive cross-task, cross-modality and joint search-space optimization generalizability, achieving state-of-the-art performance across image classification, image segmentation, speaker verification and text-to-image benchmarks. Crucially, it exhibits superior scalability in the large-scale training and joint optimization of a billion-parameter Diffusion Transformer, requiring 896 GPU-days and attaining a state-of-the-art GenEval score of 0.91. By leveraging reinforcement learning to instill Chain-of-Thought reasoning, we provide a systematic way to interpret the LLM-based searcher's internal decision-making mechanisms, revealing its optimization strategy shifted from the early-stage diversity-driven exploration to the late-stage anchor exploitation. Our proposed framework provides a scalable, generalizable, and interpretable pathway toward general-purpose open-world machine learning optimization.

cs.CL↗

Criteria for Weighted Homogeneity via Logarithmic Vector Fields

Recently in [6] the authors proved a lower bound for the $\GSV$ index of a logarithmic vector field along an isolated hypersurface singularity and proposed a conjecture that the weighted homogeneity of this hypersurface singularity can be detected by the existence of either a transverse or a non-degenerate holomorphic logarithmic vector fields. In this paper we prove this conjecture affirmatively. We also prove that the $\GSV$ index attains the lower bound if and only if the hypersurface singularity is weighted homogeneous.

math.AG↗

The minimal volume of normal stable surface pairs with reduced boundary containing a zero-dimensional non-klt center

We show that the minimum volume among normal stable surface pairs with nonzero reduced boundary containing a zero-dimensional non-klt center is $\frac{1}{42}$, and show that the surface pair attaining this minimum is unique up to isomorphism. As an application, we show that $\frac{1}{42}$ is an infinite accumulation point of the volume set of normal stable surfaces.

math.AG↗

Embedded constant mean curvature tori in hemispheres of Berger spheres

We prove that for all real numbers $κ,τ$ with $κ>8τ^2$ and $τ\neq 0$, the Berger sphere $\mathbb{E}^3(κ,τ)$ contains smooth closed embedded rotational surfaces of genus one with constant mean curvature which are contained in an open hemisphere. This disproves a conjecture of Fernández and Mira. The main result of this paper is obtained by generative AI, particularly GPT-5.6-sol, Fable 5, and the Danus system, and then verified by the authors.

math.DG↗

Mori dream Jacobian elliptic surfaces of Kodaira dimension one

Let $π\colon X\to\mathbb P^1$ be a Jacobian elliptic surface over $\mathbb C$, and set $χ=χ(\mathcal O_X)\ge3$, so that $κ(X)=1$. Assume that the Mordell--Weil group of $π$ is finite and that $π$ has at least one reducible fiber, the reducible fibers being of types $I_{n_1},\ldots,I_{n_s}$. We prove that the zero section and the components of the reducible fibers generate the closed Mori cone if and only if \[ \sum_{i=1}^s\frac{\lfloor n_i^2/4\rfloor}{n_i}\leχ. \] If $\sum_i n_i\le2χ+3$, then $X$ is a Mori dream surface. The proof combines an explicit description of the facets of the cone generated by the curves visible in the fibration with Artin's criterion applied to the null loci of the dual nef rays. We also show that, in Kodaira dimension one, finiteness of both the Mordell--Weil group and the automorphism group does not mply polyhedrality of the Mori cone. In the polyhedral range, we construct a Jacobian elliptic surface with $(χ,n)=(3,11)$, Picard number $12$, and a big and nef divisor which is not semiample; in particular, this surface is not a Mori dream surface. Finally, for every integer $ρ\ge2$, we construct a Jacobian elliptic surface of Kodaira dimension one and Picard number $ρ$ that is a Mori dream surface.

math.AG↗

Disproof of the Yau--Tian--Donaldson conjecture

We construct a polarized smooth projective fivefold and prove that it is K-polystable but does not admit a constant scalar curvature Kähler metric. This disproves the Yau--Tian--Donaldson conjecture for constant scalar curvature metrics. The main result of this paper was obtained using generative AI, particularly GPT-5.6-sol, Fable 5, and the Danus system. A detailed report on the use of generative AI in this paper is enclosed in the appendix, joint with Bin Dong and Guoxiong Gao.

math.DG↗

The equality case of Ehrhart's volume conjecture

We prove that every full-dimensional compact convex body in $\mathbb R^n$ whose barycenter is its unique interior lattice point and whose volume is $(n+1)^n/n!$ is a unimodular image of the simplex $(n+1)Δ_n-(1,\dots,1)$. This resolves the equality case of Ehrhart's volume conjecture, as a counterpart of the inequality part recently proved by OpenAI. The main result of this paper is obtained by generative AI, particularly GPT-5.6-sol, Fable 5, and the Danus system.

math.CO↗

A klt generalized pair with infinitely generated canonical ring

We construct a projective klt generalized pair over the complex numbers whose generalized log canonical ring is not finitely generated. The main result of this paper is obtained by generative AI, particularly GPT-5.6-sol-ultra, Fable 5, and the Danus system.

math.AG↗

Twelve common flex lines in a general pencil of cubics

We prove that a general pencil of plane cubics over $\mathbb C$ has exactly $12$ common flex lines. This answers a question of Ciliberto, Miranda, and Roé. The main result of this paper was obtained using generative AI, particularly ChatGPT 5.5 Pro and the Danus system.

math.AG↗

Danus: Orchestrating Mathematical Reasoning Agents with Fact-Graph Memory

Recent LLM-based mathematical reasoning agents have begun to tackle research-level problems and, in several cases, have contributed to the resolution of open problems. However, scaling and orchestrating such agents effectively remains challenging, due to the difficulty of coordinating parallel proof search while keeping intermediate claims organized and reliable. In this paper, we propose Danus, an orchestration system for research-level mathematical reasoning centered on a shared fact graph as a global memory-management mechanism. Danus consists of a main agent that performs planning and coordination, multiple worker agents that carry out proof search in parallel, and a stateless verifier that checks proposed mathematical claims before they are admitted into the fact graph. Each verified fact is stored together with its proof and logical dependencies, allowing the system to build long arguments incrementally while keeping the shared proof state organized. The main agent periodically summarizes the evolving proof state, redirects workers across promising directions, and supports interaction with human mathematicians through progress reports. We evaluate Danus through six research-level case studies in algebraic geometry, singularity theory, and combinatorics, illustrating how the fact-graph memory mechanism enables Danus to construct long, detailed mathematical proofs. Our results suggest that fact-graph-based orchestration provides an effective route toward scaling mathematical reasoning agents for long-horizon research problems. Danus is open source at https://github.com/frenzymath/Danus.

cs.AI↗

On two questions of Qi on saturated filtrations

We answer two questions of Qi [J. Reine Angew. Math. 830 (2026)] on the metric space of saturated filtrations, one affirmatively and one negatively. The main result of this paper was obtained by Chatgpt 5.5 pro, and the Danus system based on the Rethlas system.

math.AC↗