arXiv subjects

Zihong Chen

Publications and source records attributed to Zihong Chen.

11 recordsLinked to original sources

arXiv2026

VINE: Taming Generative Control Policies for Reinforcement Learning

Flow-matching policies have emerged as an effective policy parameterization for robot learning. They iteratively generate actions from noise, enabling highly expressive modeling of complex and multimodal action distributions. However, prior works observed that scaling these policies with value-gradient reinforcement learning (RL) often leads to training instability. Existing methods attribute this instability to iterative generation and therefore avoid end-to-end value-gradient optimization by sacrificing iterative generation, high expressiveness, or value-gradient optimization. Contrary to prior belief, we show the instability does not stem from iterative generation itself, but from the vanilla sampling strategy originally designed for behavior cloning, which becomes brittle under value-gradient RL. Motivated by this insight, we propose VINE, an RL-oriented sampling method that enables stable end-to-end value-gradient optimization for flow-matching policies. Instead of following a single flow trajectory, VINE reconstructs a new interpolation state at every denoising step, creating a stable differentiable path for value-gradient propagation while remaining compatible with the original flow-matching denoising process. As a result, VINE preserves the expressiveness and iterative generation of flow-matching without sacrificing end-to-end value-gradient optimization. Despite performing end-to-end backpropagation through all ten denoising steps, VINE achieves stable policy improvement and consistently outperforms state-of-the-art RL methods on the OGBench offline RL benchmark and real-world robotic manipulation task. Videos are available on our website: https://agibottech.github.io/vine.

cs.RO↗

arXiv2026

UniFS: Unified Fast-to-Slow Hierarchical Architecture for Vision-Language-Action Models

Mainstream Fast-Slow dual system vision-language-action models decouple a high-frequency action expert from a low-frequency vision-language model for efficiency, yet they face a fundamental frequency dilemma: large update gaps cause semantic drift from stale context, while small gaps erode the intended computational savings. Moreover, because the action expert receives only the VLM's final-layer representation at a single fixed frequency, rich intermediate features are discarded, limiting both information coupling and manipulation precision. Inspired by multi-timescale neural processing in the human brain, we introduce UniFS, a unified fast-to-slow architecture that resolves these challenges through three key designs. First, we stratify the VLM layers into groups with progressively decreasing update frequencies, enabling shallow layers to capture fast-changing dynamics while deeper layers cache stable semantic context. Second, a latent vector inversion mechanism re-routes the interaction order between multi-scale VLM features and the action expert, aligning fast-varying representations with fine-grained action decoding and slow-varying ones with coarse planning. Third, a multi-level supervision strategy enforces a coarse-to-fine learning hierarchy across temporal scales. Together, these designs enable richer cross-frequency information transfer within a single backbone, while the low-frequency pathways additionally preserve temporal context across steps. Experiments on LIBERO show that UniFS achieves state-of-the-art performance (98.3\% average success rate, a 2.5\% gain over VLA-Adapter baseline) while reducing average inference latency from 36.5~ms to 17.8~ms (2.1$\times$ speedup). Real-robot experiments on a Franka platform further validate its practical applicability. Code is opensourced at https://github.com/linsun449/UniFS.

cs.RO↗

arXiv2026

Knowledge-to-Verification: Exploring RLVR for LLMs in Knowledge-Intensive Domains

Reinforcement learning with verifiable rewards (RLVR) has demonstrated promising potential to enhance the reasoning capabilities of large language models (LLMs) in domains such as mathematics and coding. However, its applications on knowledge-intensive domains have not been effectively explored due to the scarcity of high-quality verifiable data. Furthermore, current RLVR focuses solely on the correctness of final answers, leading to the limitations of flawed reasoning and sparse reward signals. In this work, we propose Knowledge-to-Verification (K2V), a framework that extends RLVR to knowledge-intensive domains through automated verifiable data synthesis, while enabling verification of the LLM's reasoning process. Extensive experiments demonstrate that K2V enhances the reasoning of LLM in knowledge-intensive domains without significantly compromising the model's general capabilities. This study also suggests that integrating automated data synthesis with reasoning verification is a promising direction to enhance model capabilities in these broader domains. Code is available at https://github.com/SeedScientist/K2V.

cs.CL↗

arXiv2026

NTIRE 2026 Challenge on Short-form UGC Video Restoration in the Wild with Generative Models: Datasets, Methods and Results

This paper presents an overview of the NTIRE 2026 Challenge on Short-form UGC Video Restoration in the Wild with Generative Models. This challenge utilizes a new short-form UGC (S-UGC) video restoration benchmark, termed KwaiVIR, which is contributed by USTC and Kuaishou Technology. It contains both synthetically distorted videos and real-world short-form UGC videos in the wild. For this edition, the released data include 200 synthetic training videos, 48 wild training videos, 11 validation videos, and 20 testing videos. The primary goal of this challenge is to establish a strong and practical benchmark for restoring short-form UGC videos under complex real-world degradations, especially in the emerging paradigm of generative-model-based S-UGC video restoration. This challenge has two tracks: (i) the primary track is a subjective track, where the evaluation is based on a user study; (ii) the second track is an objective track. These two tracks enable a comprehensive assessment of restoration quality. In total, 95 teams have registered for this competition. And 12 teams submitted valid final solutions and fact sheets for the testing phase. The submitted methods achieved strong performance on the KwaiVIR benchmark, demonstrating encouraging progress in short-form UGC video restoration in the wild.

cs.CV↗

arXiv2026

The Getzler-Gauss-Manin connection and Kontsevich-Soibelman operations on the periodic cyclic homology

We study equivariant operations on the periodic cyclic homology of dg algebras that arise from the chain level action of the two-colored Kontsevich-Soibelman operad. The first main result is that these operations are covariantly constant with respect to the Getzler-Gauss-Manin connection on the periodic cyclic homology of a family of dg algebras. Then, using classical computations of Cohen \cite{Coh}, we explicitly compute a set of generators for these operations under composition, and show that these generators are closely related to the $p$-fold equivariant cap products previously studied by the author \cite{Che2} in relation to equivariant Gromov-Witten theory with mod $p$ coefficients. The main technical novelty is a re-formulation of the Kontsevich-Soibelman operad in terms of a two-colored version of the cacti operad, and a proof that it is \emph{equivariantly} quasi-equivalent to the two-colored operad of little disks on a disk/cylinder.

math.QA↗

arXiv2025

GraphGen: Enhancing Supervised Fine-Tuning for LLMs with Knowledge-Driven Synthetic Data Generation

Fine-tuning for large language models (LLMs) typically requires substantial amounts of high-quality supervised data, which is both costly and labor-intensive to acquire. While synthetic data generation has emerged as a promising solution, existing approaches frequently suffer from factual inaccuracies, insufficient long-tail coverage, simplistic knowledge structures, and homogenized outputs. To address these challenges, we introduce GraphGen, a knowledge graph-guided framework designed for three key question-answering (QA) scenarios: atomic QA, aggregated QA, and multi-hop QA. It begins by constructing a fine-grained knowledge graph from the source text. It then identifies knowledge gaps in LLMs using the expected calibration error metric, prioritizing the generation of QA pairs that target high-value, long-tail knowledge. Furthermore, GraphGen incorporates multi-hop neighborhood sampling to capture complex relational information and employs style-controlled generation to diversify the resulting QA data. Experimental results on knowledge-intensive tasks under closed-book settings demonstrate that GraphGen outperforms conventional synthetic data methods, offering a more reliable and comprehensive solution to the data scarcity challenge in supervised fine-tuning. The code and data are publicly available at https://github.com/open-sciencelab/GraphGen.

cs.CL↗

arXiv2025

SeedBench: A Multi-task Benchmark for Evaluating Large Language Models in Seed Science

Seed science is essential for modern agriculture, directly influencing crop yields and global food security. However, challenges such as interdisciplinary complexity and high costs with limited returns hinder progress, leading to a shortage of experts and insufficient technological support. While large language models (LLMs) have shown promise across various fields, their application in seed science remains limited due to the scarcity of digital resources, complex gene-trait relationships, and the lack of standardized benchmarks. To address this gap, we introduce SeedBench -- the first multi-task benchmark specifically designed for seed science. Developed in collaboration with domain experts, SeedBench focuses on seed breeding and simulates key aspects of modern breeding processes. We conduct a comprehensive evaluation of 26 leading LLMs, encompassing proprietary, open-source, and domain-specific fine-tuned models. Our findings not only highlight the substantial gaps between the power of LLMs and the real-world seed science problems, but also make a foundational step for research on LLMs for seed design.

cs.CL↗

arXiv2024

On the exponential type conjecture

We prove that the small quantum t-connection on a closed monotone symplectic manifold is of exponential type and has quasi-unipotent regularized monodromies at t=0. This answers a conjecture of Katzarkov-Kontsevich-Pantev and Galkin-Golyshev-Iritani for those classes of symplectic manifolds. The proof follows a reduction to positive characteristics argument, and the main tools of the proof are Katz's local monodromy theorem in differential equations and quantum Steenrod operations in equivariant Gromov-Witten theory with mod p coefficients.

math.SG↗

arXiv2024

Quantum Steenrod operations and Fukaya categories

This paper is concerned with quantum cohomology and Fukaya categories of a closed monotone symplectic manifold X, where we use coefficients in a field k of characteristic p > 0. The main result of this paper is that the quantum Steenrod operations QΣadmit an interpretation in terms of certain operations on the (equivariant) Hochschild invariants of the Fukaya category of X, via suitable (equivariant) versions of the open-closed maps. As an application, we demonstrate how the categorical perspective provides new tools for computing QΣbeyond the reach of known technology. We also explore potential connections of our work to arithmetic homological mirror symmetry.

math.SG↗

arXiv2024

On operadic open-closed maps in characteristic $p$

Consider a closed monotone symplectic manifold $(M,ω)$. \cite{Gan2} constructed a cyclic open-closed map, which goes from the cyclic homology of the Fukaya category of $M$ to the $S^1$-equivariant quantum cohomology of $M$. In this paper, we show that with mod $p$ coefficients, Ganatra's cyclic open-closed map is compatible with a certain $\mathbb{Z}/p$-equivariant open-closed map under the natural $\mathbb{Z}/p$-Gysin type comparison map for Hochschild homology. Along with the proof, this paper gives a new homotopy theoretic framework for studying open-closed maps in symplectic topology. These will be used in an upcoming work \cite{Che} to study mod $p$ equivariant enumerative invariants such as the Quantum Steenrod operations. The main insights of this paper are: 1) a $\mathbb{Z}/p$-Gysin comparison result for ($\mathcal{A}_{\infty}$-) cyclic objects, 2) a new construction of the open-closed map using operadic Floer theory of \cite{AGV}, which gives rise to a new interpretation of its `$S^1$-equivariant' property, and 3) comparison of the new construction with its classical counterpart.

math.SG↗

arXiv2018

Stability Conditions and Exceptional Objects in Triangulated Categories

The goal of this paper is to study the subspace of stability condition $Σ_{\mathcal{E}}\subset \mathrm{Stab}(X)$ associated to an exceptional collection $\mathcal{E}$ on a projective variety $X$. Following Emanuele Macrì's approach, we show a certain correspondence between the homotopy class of continuous loops in $Σ_{\mathcal{E}}$ and words of the braid group. In particular, we prove that in the case $X=\mathbb{P}^3$ and $\mathcal{E}=\{\mathcal{O},\mathcal{O}(1),\mathcal{O}(2),\mathcal{O}(3)\}$, the space $Σ_{\mathcal{E}}$ is a connected and simply connected 4-dimensional manifold.

math.AG↗