arXiv subjects

Yong Li

Publications and source records attributed to Yong Li.

At least 289 records · Page 16Linked to original sources

arXiv2025

The Thinking Spectrum: An Empirical Study of Tunable Reasoning in LLMs through Model Merging

The growing demand for large language models (LLMs) with tunable reasoning capabilities in many real-world applications highlights a critical need for methods that can efficiently produce a spectrum of models balancing reasoning depth and computational cost. Model merging has emerged as a promising, training-free technique to address this challenge by arithmetically combining the weights of a general-purpose model with a specialized reasoning model. While various merging techniques exist, their potential to create a spectrum of models with fine-grained control over reasoning abilities remains largely unexplored. This work presents a large-scale empirical study evaluating a range of model merging techniques across multiple reasoning benchmarks. We systematically vary merging strengths to construct accuracy-efficiency curves, providing the first comprehensive view of the tunable performance landscape. Our findings reveal that model merging offers an effective and controllable method for calibrating the trade-off between reasoning accuracy and token efficiency, even when parent models have highly divergent weight spaces. Crucially, we identify instances of Pareto Improvement, where a merged model achieves both higher accuracy and lower token consumption than one of its parents. Our study provides the first comprehensive analysis of this tunable space, offering practical guidelines for creating LLMs with specific reasoning profiles to meet diverse application demands.

cs.AI↗

arXiv2025

Collaborative Deterministic-Probabilistic Forecasting for Diverse Spatiotemporal Systems

Probabilistic forecasting is crucial for real-world spatiotemporal systems, such as climate, energy, and urban environments, where quantifying uncertainty is essential for informed, risk-aware decision-making. While diffusion models have shown promise in capturing complex data distributions, their application to spatiotemporal forecasting remains limited due to complex spatiotemporal dynamics and high computational demands. we propose CoST, a general forecasting framework that collaborates deterministic and diffusion models for diverse spatiotemporal systems. CoST formulates a mean-residual decomposition strategy: it leverages a powerful deterministic model to capture the conditional mean and a lightweight diffusion model to learn residual uncertainties. This collaborative formulation simplifies learning objectives, improves accuracy and efficiency, and generalizes across diverse spatiotemporal systems. To address spatial heterogeneity, we further design a scale-aware diffusion mechanism to guide the diffusion process. Extensive experiments across ten real-world datasets from climate, energy, communication, and urban systems show that CoST achieves 25\% performance gains over state-of-the-art baselines, while significantly reducing computational cost.

cs.LG↗

arXiv2025

LongScape: Advancing Long-Horizon Embodied World Models with Context-Aware MoE

Video-based world models hold significant potential for generating high-quality embodied manipulation data. However, current video generation methods struggle to achieve stable long-horizon generation: classical diffusion-based approaches often suffer from temporal inconsistency and visual drift over multiple rollouts, while autoregressive methods tend to compromise on visual detail. To solve this, we introduce LongScape, a hybrid framework that adaptively combines intra-chunk diffusion denoising with inter-chunk autoregressive causal generation. Our core innovation is an action-guided, variable-length chunking mechanism that partitions video based on the semantic context of robotic actions. This ensures each chunk represents a complete, coherent action, enabling the model to flexibly generate diverse dynamics. We further introduce a Context-aware Mixture-of-Experts (CMoE) framework that adaptively activates specialized experts for each chunk during generation, guaranteeing high visual quality and seamless chunk transitions. Extensive experimental results demonstrate that our method achieves stable and consistent long-horizon generation over extended rollouts. Our code is available at: https://github.com/tsinghua-fib-lab/Longscape.

cs.CV↗

arXiv2025

The LLN and CLT for the statistical ensembles of discrete integrable Hamiltonian systems

This paper investigates the behavior of statistical ensembles under iteration map induced by discrete integrable Hamiltonian systems in deterministic case and stochastic case, addressing the problem from two perspectives: the Law of Large Numbers and the Central Limit Theorem. In deterministic case, the Law of Large Numbers simplifies the convergence conditions to the extent that the Riemann-Lebesgue lemma is no longer required. In the stochastic setting, we extend the results to general stochastic processes, beginning with the perturbation term represented by standard Brownian motion. Moreover, we establish a Central Limit Theorem for the statistical ensemble. A numerical example is also included.

math.PR↗

arXiv2025

KeyWorld: Key Frame Reasoning Enables Effective and Efficient World Models

Robotic world models are a promising paradigm for forecasting future environment states, yet their inference speed and the physical plausibility of generated trajectories remain critical bottlenecks, limiting their real-world applications. This stems from the redundancy of the prevailing frame-to-frame generation approach, where the model conducts costly computation on similar frames, as well as neglecting the semantic importance of key transitions. To address this inefficiency, we propose KeyWorld, a framework that improves text-conditioned robotic world models by concentrating transformers computation on a few semantic key frames while employing a lightweight convolutional model to fill the intermediate frames. Specifically, KeyWorld first identifies significant transitions by iteratively simplifying the robot's motion trajectories, obtaining the ground truth key frames. Then, a DiT model is trained to reason and generate these physically meaningful key frames from textual task descriptions. Finally, a lightweight interpolator efficiently reconstructs the full video by inpainting all intermediate frames. Evaluations on the LIBERO benchmark demonstrate that KeyWorld achieves a 5.68$\times$ acceleration compared to the frame-to-frame generation baseline, and focusing on the motion-aware key frames further contributes to the physical validity of the generated videos, especially on complex tasks. Our approach highlights a practical path toward deploying world models in real-time robotic control and other domains requiring both efficient and effective world models. Code is released at https://anonymous.4open.science/r/Keyworld-E43D.

cs.RO↗

arXiv2025

Angular Linearization and Quantitative Convergence of Statistical Ensembles for Weighted Integrable Hamiltonian Systems

Within the framework of weighted integrable Hamiltonian systems, we study the long-time behavior of the associated statistical ensembles. We construct an action-dependent angular conjugacy that rectifies the nonuniform angular flow into a constant-speed linear flow, thereby reducing the ensemble dynamics to oscillatory integrals over the action domain. Using integration by parts along a tailored vector field and $W^{1,1}$ approximations with controlled boundary terms, we derive inverse-in-time decay for each Fourier mode and complete the summation under nonresonance and uniform nondegeneracy assumptions. Consequently, under suitable regularity and admissible initial data, we prove that the ensemble converges to the weighted equilibrium measure at rate $O(1/t)$, with an explicit parameter-dependent bound. The methodology provides a transferable analytic scheme for establishing limit theorems in near-integrable regimes.

math.DS↗

arXiv2025

Structure-prior Informed Diffusion Model for Graph Source Localization with Limited Data

Source localization in graph information propagation is essential for mitigating network disruptions, including misinformation spread, cyber threats, and infrastructure failures. Existing deep generative approaches face significant challenges in real-world applications due to limited propagation data availability. We present SIDSL (\textbf{S}tructure-prior \textbf{I}nformed \textbf{D}iffusion model for \textbf{S}ource \textbf{L}ocalization), a generative diffusion framework that leverages topology-aware priors to enable robust source localization with limited data. SIDSL addresses three key challenges: unknown propagation patterns through structure-based source estimations via graph label propagation, complex topology-propagation relationships via a propagation-enhanced conditional denoiser with GNN-parameterized label propagation module, and class imbalance through structure-prior biased diffusion initialization. By learning pattern-invariant features from synthetic data generated by established propagation models, SIDSL enables effective knowledge transfer to real-world scenarios. Experimental evaluation on four real-world datasets demonstrates superior performance with 7.5-13.3\% F1 score improvements over baselines, including over 19\% improvement in few-shot and 40\% in zero-shot settings, validating the framework's effectiveness for practical source localization. Our code can be found \href{https://github.com/tsinghua-fib-lab/SIDSL}{here}.

cs.SI↗

arXiv2025

Seeing 3D Through 2D Lenses: 3D Few-Shot Class-Incremental Learning via Cross-Modal Geometric Rectification

The rapid growth of 3D digital content necessitates expandable recognition systems for open-world scenarios. However, existing 3D class-incremental learning methods struggle under extreme data scarcity due to geometric misalignment and texture bias. While recent approaches integrate 3D data with 2D foundation models (e.g., CLIP), they suffer from semantic blurring caused by texture-biased projections and indiscriminate fusion of geometric-textural cues, leading to unstable decision prototypes and catastrophic forgetting. To address these issues, we propose Cross-Modal Geometric Rectification (CMGR), a framework that enhances 3D geometric fidelity by leveraging CLIP's hierarchical spatial semantics. Specifically, we introduce a Structure-Aware Geometric Rectification module that hierarchically aligns 3D part structures with CLIP's intermediate spatial priors through attention-driven geometric fusion. Additionally, a Texture Amplification Module synthesizes minimal yet discriminative textures to suppress noise and reinforce cross-modal consistency. To further stabilize incremental prototypes, we employ a Base-Novel Discriminator that isolates geometric variations. Extensive experiments demonstrate that our method significantly improves 3D few-shot class-incremental learning, achieving superior geometric coherence and robustness to texture bias across cross-domain and within-domain settings.

cs.CV↗

arXiv2025

Learning Attribute-Aware Hash Codes for Fine-Grained Image Retrieval via Query Optimization

Fine-grained hashing has become a powerful solution for rapid and efficient image retrieval, particularly in scenarios requiring high discrimination between visually similar categories. To enable each hash bit to correspond to specific visual attributes, we propoe a novel method that harnesses learnable queries for attribute-aware hash codes learning. This method deploys a tailored set of queries to capture and represent nuanced attribute-level information within the hashing process, thereby enhancing both the interpretability and relevance of each hash bit. Building on this query-based optimization framework, we incorporate an auxiliary branch to help alleviate the challenges of complex landscape optimization often encountered with low-bit hash codes. This auxiliary branch models high-order attribute interactions, reinforcing the robustness and specificity of the generated hash codes. Experimental results on benchmark datasets demonstrate that our method generates attribute-aware hash codes and consistently outperforms state-of-the-art techniques in retrieval accuracy and robustness, especially for low-bit hash codes, underscoring its potential in fine-grained image hashing tasks.

cs.CV↗

arXiv2025

Secure Multicast Communications with Pinching-Antenna Systems (PASS)

This article investigates secure multicast communications in pinching-antenna systems (PASS), where pinching beamforming is enabled by adaptively adjusting pinching antenna (PAs) positions along waveguides to improve multicast security. Specifically, a PASS-based secure multicast framework is proposed, in which joint optimization of transmit and pinching beamforming is conducted to maximize the secrecy multicast rate. i) For the single-group multicast scenario, an alternating optimization (AO) framework is employed, where the pinching beamformer is updated via an element-wise sequential optimization method. The transmit beamformer is designed via a semidefinite relaxation (SDR) formulation for an upper-bound solution, while a Dinkelbach-alternating direction method of multipliers (ADMM) offers a low-complexity alternative. ii) For the multi-group multicast scenario, transmit and pinching beamformers are alternately optimized under a majorization-minimization (MM) framework. The transmit beamformer is obtained via SDR or an efficient second-order cone programming (SOCP) method, while the pinching beamformer is updated through MM-based element-wise sequential update strategy. Numerical results are provided to demonstrate that: (i) PASS consistently outperform conventional fixed-location antenna architectures in terms of secrecy performance across various configurations; and (ii) the performance advantage of PASS over fixed-location architectures becomes more significant with increased service region, larger antenna arrays, and higher user and eavesdropper densities.

eess.SP↗

arXiv2025

Rationality Check! Benchmarking the Rationality of Large Language Models

Large language models (LLMs), a recent advance in deep learning and machine intelligence, have manifested astonishing capacities, now considered among the most promising for artificial general intelligence. With human-like capabilities, LLMs have been used to simulate humans and serve as AI assistants across many applications. As a result, great concern has arisen about whether and under what circumstances LLMs think and behave like real human agents. Rationality is among the most important concepts in assessing human behavior, both in thinking (i.e., theoretical rationality) and in taking action (i.e., practical rationality). In this work, we propose the first benchmark for evaluating the omnibus rationality of LLMs, covering a wide range of domains and LLMs. The benchmark includes an easy-to-use toolkit, extensive experimental results, and analysis that illuminates where LLMs converge and diverge from idealized human rationality. We believe the benchmark can serve as a foundational tool for both developers and users of LLMs.

cs.AI↗

arXiv2025

MobiGPT: A Foundation Model for Mobile Wireless Networks

With the rapid development of mobile communication technologies, future mobile networks will offer vast services and resources for commuting, production, daily life, and entertainment. Accurate and efficient forecasting of mobile data (e.g., cell traffic, user behavior, channel quality) helps operators monitor network state changes, orchestrate wireless resources, and schedule infrastructure and users, thereby improving supply efficiency and service quality. However, current forecasting paradigms rely on customized designs with tailored models for exclusive data types. Such approaches increase complexity and deployment costs under large-scale, heterogeneous networks involving base stations, users, and channels. In this paper, we design a foundation model for mobile data forecasting, MobiGPT, with a unified structure capable of forecasting three data types: base station traffic, user app usage, and channel quality. We propose a soft-prompt learning method to help the model understand features of different data types, and introduce a temporal masking mechanism to guide the model through three forecasting tasks: short-term prediction, long-term prediction, and distribution generation, supporting diverse optimization scenarios. Evaluations on real-world datasets with over 100,000 samples show that MobiGPT achieves accurate multi-type forecasting. Compared to existing models, it improves forecasting accuracy by 27.37%, 20.08%, and 7.27%, reflecting strong generalization. Moreover, MobiGPT exhibits superior zero/few-shot performance in unseen scenarios, with over 21.51% improvement, validating its strong transferability as a foundation model.

cs.LG↗

arXiv2025

FLAME: A Serving System Optimized for Large-Scale Generative Recommendation with Efficiency

Generative recommendation (GR) models possess greater scaling power compared to traditional deep learning recommendation models (DLRMs), yet they also impose a tremendous increase in computational burden. Measured in FLOPs, a typical GR model's workload sits in $10^9 \sim 10^{11}$ range, roughly four orders of magnitude higher than traditional DLRMs. Delivering accurate results in a few tens of milliseconds while processing billions of such requests per day puts extreme demands on the performance of the online serving system. Therefore, for industry practitioners, the alluring gains of GR models are tempered by the formidable challenge of online deployment at scale in production services. In this work, we introduce a comprehensive solution of online serving system tailored For Large-scale GenerAtive RecoMmendation with Efficiency (FLAME). Specifically, we leveraging CPU-GPU heterogeneous hardware to decouple feature pre-processing and model computation. We encapsulated several memory optimization features as the Proximal Data Accelerator (PDA) module to make full use of limited bandwidth and storage resources, which achieves a 1.9x throughput gain and a 1.7x latency reduction. We implement the Fused Kernel Engine (FKE) module based on the functionality and interface of NVIDIA TensorRT to boost model computation, delivering a speedup ratio of 4.6x-6.1x, throughput gain ratio of 4.7x-6.3x one step further. In addition, we design the Dynamic Stream Orchestrator (DSO) module to coordinate concurrent requests, enhancing the system throughput performance with 1.3x improvement in throughput and 2.3x speed-up under non-uniform distribution of upstream candidates. Comprehensive evaluations demonstrate that our FLAME effectively supports large-scale online deployment of GR models and achieves remarkable improvements in system performance.

cs.DC↗

arXiv2025

Towards sharp regularity: Full-dimensional tori in $ C^\infty $ vector fields over $ \mathbb{T}^\infty $

We consider the $ C^1 $ linearization of a perturbed vector field $ ω+P $ over the infinite-dimensional torus $ \mathbb{T}^\infty $, and determine the sharp regularity requirement of the perturbation $ P $ conjugating the unperturbed one $ ω$ onto $ ω-\tildeω+P $ via a small modifying term $ \tildeω $. We discuss the Diophantine type introduced by Bourgain, investigate the universal nonresonance, and provide the weakest known regularity of perturbations for which KAM applies. Our results allow for lower regularity than analyticity such as Gevrey regularity or even $ C^\infty $ regularity. We propose a new KAM scheme with a balancing sequence to overcome the non-polynomial nonresonance, differing from the usual Newtonian approach. Thereby, in addition to deriving the sharp Gevrey exponent along Diophantine nonresonance, we answer the fundamental question of what the minimum regularity required for KAM is in the infinite-dimensional setting: $ C^\infty $ regularity. Our linearization technique also applies to the quasi-periodic case over $ \mathbb{T}^n $.

math.DS↗

arXiv2025

Symbolic regression via MDLformer-guided search: from minimizing prediction error to minimizing description length

Symbolic regression, a task discovering the formula best fitting the given data, is typically based on the heuristical search. These methods usually update candidate formulas to obtain new ones with lower prediction errors iteratively. However, since formulas with similar function shapes may have completely different symbolic forms, the prediction error does not decrease monotonously as the search approaches the target formula, causing the low recovery rate of existing methods. To solve this problem, we propose a novel search objective based on the minimum description length, which reflects the distance from the target and decreases monotonically as the search approaches the correct form of the target formula. To estimate the minimum description length of any input data, we design a neural network, MDLformer, which enables robust and scalable estimation through large-scale training. With the MDLformer's output as the search objective, we implement a symbolic regression method, SR4MDL, that can effectively recover the correct mathematical form of the formula. Extensive experiments illustrate its excellent performance in recovering formulas from data. Our method successfully recovers around 50 formulas across two benchmark datasets comprising 133 problems, outperforming state-of-the-art methods by 43.92%. Experiments on 122 unseen black-box problems further demonstrate its generalization performance. We release our code at https://github.com/tsinghua-fib-lab/SR4MDL .

cs.LG↗

arXiv2025

Flexible Coded Distributed Convolution Computing for Enhanced Straggler Resilience and Numerical Stability in Distributed CNNs

Deploying Convolutional Neural Networks (CNNs) on resource-constrained devices necessitates efficient management of computational resources, often via distributed environments susceptible to latency from straggler nodes. This paper introduces the Flexible Coded Distributed Convolution Computing (FCDCC) framework to enhance straggler resilience and numerical stability in distributed CNNs. We extend Coded Distributed Computing (CDC) with Circulant and Rotation Matrix Embedding (CRME) which was originally proposed for matrix multiplication to high-dimensional tensor convolution. For the proposed scheme, referred to as the Numerically Stable Coded Tensor Convolution (NSCTC) scheme, we also propose two new coded partitioning schemes: Adaptive-Padding Coded Partitioning (APCP) for the input tensor and Kernel-Channel Coded Partitioning (KCCP) for the filter tensor. These strategies enable linear decomposition of tensor convolutions and encoding them into CDC subtasks, combining model parallelism with coded redundancy for robust and efficient execution. Theoretical analysis identifies an optimal trade-off between communication and storage costs. Empirical results validate the framework's effectiveness in computational efficiency, straggler resilience, and scalability across various CNN architectures.

cs.DC↗

arXiv2025

PG-Agent: An Agent Powered by Page Graph

Graphical User Interface (GUI) agents possess significant commercial and social value, and GUI agents powered by advanced multimodal large language models (MLLMs) have demonstrated remarkable potential. Currently, existing GUI agents usually utilize sequential episodes of multi-step operations across pages as the prior GUI knowledge, which fails to capture the complex transition relationship between pages, making it challenging for the agents to deeply perceive the GUI environment and generalize to new scenarios. Therefore, we design an automated pipeline to transform the sequential episodes into page graphs, which explicitly model the graph structure of the pages that are naturally connected by actions. To fully utilize the page graphs, we further introduce Retrieval-Augmented Generation (RAG) technology to effectively retrieve reliable perception guidelines of GUI from them, and a tailored multi-agent framework PG-Agent with task decomposition strategy is proposed to be injected with the guidelines so that it can generalize to unseen scenarios. Extensive experiments on various benchmarks demonstrate the effectiveness of PG-Agent, even with limited episodes for page graph construction.

cs.AI↗

arXiv2025

Resurgence for large $c$ expansion in Coulomb gas formalism

We develop a resurgence analysis for large central charge (large $C$) expansions in two-dimensional CFTs using the Coulomb gas formalism. Through the exact Borel-Laplace representations of the conformal blocks $I_1(C,z)$ and $I_2(C,z)$ associated with the four-point correlation function $ \langle ϕ_{2,1}(0)ϕ_{2,1}(z,\bar{z})ϕ_{2,1}(1)ϕ_{2,1}(\infty)\rangle$, we demonstrate that $I_1(C,z)$ participates in the Stokes phenomenon of $I_2(C,z)$ (and vice versa), and establish that monodromy in $z$ arises from alien calculus in the Borel plane variable $ζ$ (Borel dual to $C$). From a given conformal block, resurgence theory thus enables us to discover other internal operators (conformal blocks). This approach establishes a non-perturbative connection between conformal blocks, shedding light on the resurgence phenomena in more general quantum field theories.

hep-th↗