arXiv subjects

Jian Li

Publications and source records attributed to Jian Li.

At least 37 records · Page 2Linked to original sources

arXiv2026

AdaRoPE: Not All Attention Heads Should Rotate and Scale Equally

Rotary Position Embedding (RoPE) is widely adopted in Transformers to encode positional information, yet standard implementations enforce a uniform frequency schedule and scaling across all attention heads. Using simplified retrieval tasks and length generalization scenarios, we show -- both empirically and theoretically -- that heads with different functional roles require distinct frequency ranges and attention scaling factors to operate effectively. Ignoring this structure leads to suboptimal utilization of embedding dimensions and degraded performance, particularly under long-context settings. To address these limitations, we propose AdaRoPE, which equips each attention head with learnable rotation frequencies and attention scaling factors. Pretrained LLMs with AdaRoPE consistently outperform existing RoPE variants, including partial RoPE and NoPE baselines. For context extension, we further show that uniform frequency and attention scaling, used in methods such as YaRN, are suboptimal. By applying head-specific scaling, AdaRoPE enables better context extension while better preserving short-context performance in both the extrapolation setting and the long-context continued pretraining setting. These results highlight the importance of optimizing rotary position embedding at the level of individual attention heads.

cs.AI↗

arXiv2026

SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers

Looped Transformers increase effective depth by iterating a shared block of layers, but most evaluations compare at fixed model size, conflating architectural advantage with extra FLOPs. We study looping on Mixture-of-Experts Transformers while closely matching per-token FLOPs, total non-embedding parameters, and KV cache. Through a series of ablations, we arrive at a recipe we call SMELT (Sparse MoE Transformer, middle layers Loop Twice), which loops the middle half of layers twice while matching the unlooped Baseline on all three budgets. We scale SMELT across four sizes up to 54B non-embedding parameters and fit a separate Chinchilla-style scaling law for each architecture. SMELT's loss drops faster with compute, saving 6.8--18.0\% of training FLOPs on the compute-optimal frontier. The advantage transfers to downstream benchmarks beyond what validation loss predicts, is largest on Code, and grows with sample length and the number of in-context examples. Mechanistic analysis shows that the second visit reduces the attention sink and redirects mass toward content-relevant tokens, an inductive bias that may underlie the observed performance gains. These results show that looping can improve Transformers even under budget matching, offering a practical recipe that turns depth reuse into measurable gains.

cs.LG↗

arXiv2026

Quantum convolutional neural network for predicting nuclear charge radii

Quantum machine learning has the potential to become a new tool for understanding complex nuclear structures. In this work, we apply a hybrid quantum convolutional neural network (QCNN) to nuclear charge-radius prediction for the first time, aiming to explore the feasibility of quantum machine learning in nuclear-physics data analysis. Based on a classical convolutional neural network (CNN) framework, a small variational quantum convolutional filter is introduced as a quantum feature map to extract local correlations on the nuclear chart. The QCNN shows promising predictive accuracy and training stability, and provides a reliable description of the charge-radius evolution along several representative isotopic chains. These results support further investigation of quantum convolutional architectures for nuclear-structure data analysis.

nucl-th↗

arXiv2026

Gaussian Belief Propagation Network for Depth Completion

Depth completion aims to predict a dense depth map from a color image with sparse depth measurements. Although deep learning methods have achieved state-of-the-art (SOTA), effectively handling the sparse and irregular nature of input depth data in deep networks remains a significant challenge, often limiting performance, especially under high sparsity. To overcome this limitation, we introduce the Gaussian Belief Propagation Network (GBPN), a novel hybrid framework synergistically integrating deep learning with probabilistic graphical models for end-to-end depth completion. Specifically, a scene-specific Markov Random Field (MRF) is dynamically constructed by the Graphical Model Construction Network (GMCN), and then inferred via Gaussian Belief Propagation (GBP) to yield the dense depth distribution. Crucially, the GMCN learns to construct not only the data-dependent potentials of MRF but also its structure by predicting adaptive non-local edges, enabling the capture of complex, long-range spatial dependencies. Furthermore, we enhance GBP with a serial \& parallel message passing scheme, designed for effective information propagation, particularly from sparse measurements. Extensive experiments demonstrate that GBPN achieves SOTA performance on the NYUv2 and KITTI benchmarks. Evaluations across varying sparsity levels, sparsity patterns, and datasets highlight GBPN's superior performance, notable robustness, and generalizable capability.

cs.CV↗

arXiv2026

Weak disjointness of hypercyclic operators

We study the weak disjointness of hypercyclic operators to refine the classifications of hypercyclic operators. We establish an analogue of the Weiss-Akin-Glasner Theorem from topological dynamics within the framework of linear dynamics, which gives a characterization of the weak disjointness of each class of mixing operators with respect to a given Furstenberg family. The key ingredient is an analogue of the Weiss-Akin-Glasner Lemma from topological dynamics, which gives a characterization of subsets of non-negative integers which can be realized by the return time sets of mixing operators with respect to a given Furstenberg family. We also provide several examples to distinguish some classes of hypercyclic operators and end with the characterization of the mixing property with respect to Furstenberg families of direct sums of backward shifts on Fréchet sequence spaces.

math.DS↗

arXiv2026

Non-Hermitian Topology and Boundary Jordan Chains with Generalized Chiral Symmetry

We study a generalization of chiral symmetry applicable to non-Hermitian systems and its topological consequences on one-dimensional chains. We find a rich family of topological phases characterized not by a single winding number, but a vector of them. More importantly, we uncover a novel type of bulk-boundary correspondence, where the vector of winding numbers in the bulk corresponds to the set of Jordan chains of various length at the boundary. This in turn leads to highly unconventional chiral-charge distributions on both edges. Our work extends the topological classification of the non-Hermitian AIII class along a new axis.

cond-mat.mes-hall↗

arXiv2026

A semi-analytical Ring-Arc model for Plutino perturbations

Neptune's 2:3 mean motion resonance induces an asymmetric spatial distribution of Plutinos. We previously proposed a 9000-point-mass arc model that successfully captures the resulting influence on planetary ephemerides. However, integrating thousands of point masses involves significant computational cost. In this paper, to more efficiently simulate the total perturbation of Plutinos, we develop a semi-analytical Ring-Arc model described by gravitational potentials. This model consists of an outer ring and two inner arcs, representing the resonant spatial distribution of Plutinos. The model parameters are derived from observed Plutinos and theoretical sampling, yielding consistent structures. Numerical simulations show that the perturbations on the Sun-Neptune and Sun-Saturn distances from the semi-analytical Ring-Arc model agree with those from the 9000-point-mass arc model within 4\% over 50 years, while outperforming a simple ring model in comparison with the bias-corrected Plutino population. Our semi-analytical approach preserves the dynamical characteristics while significantly enhancing computational efficiency, offering a state-of-the-art tool for ephemeris modelling.

astro-ph.EP↗

arXiv2026

ArtSociety: Multi-Agent Multimodal Collaboration for Art Emotion Understanding

The AffectiveArt Multidimensional Art Emotion Understanding task asks to jointly predict an artwork's fine-grained emotion (12 classes, 1549:1 head-to-tail ratio), binary valence/arousal, and five attribute-grounded descriptions -- sub-tasks that exhibit strong empirical trade-offs, so the single-model solutions we tried do not jointly optimize all of them well. We present ArtSociety, a multi-agent framework that assembles heterogeneous multimodal experts -- a DINOv2-Giant vision agent (A1), a scene-grounded CoT fine-tuned MLLM (A2), and three closed-source reasoning agents (A3-A5) -- and coordinates them with two training-free controllers: (i) a rare-class-aware voting arbiter that lowers the agreement threshold for tail emotions, exploiting decorrelated error patterns across agent families; and (ii) a description-first reasoning agent whose DESCRIBE-then-CLASSIFY chain of thought forces visual evidence before label commitment, yielding near-perfect grounded descriptions. A task-routing policy directs the hard emotion task to the full five-agent ensemble while assigning the near-saturated valence/arousal and generative description tasks to the single strongest reasoning agent. On the official test set (1,000 artworks), ArtSociety achieves an Overall Score of 0.8870 (Classification 0.7789, Description 0.9952). An eleven-variant ablation study reveals that, once method and scale saturate at around 0.76, the decisive gains come from agent collaboration and data-side supervision -- a 30B MoE model trained on older data does not outperform an 8B model trained on better data. Code is available at https://github.com/swordlidev/ArtSociety

cs.CV↗

arXiv2026

Automatic Item Generation for Personality Situational Judgment Tests with Large Language Models

Personality assessment through situational judgment tests (SJTs) offers unique advantages over traditional Likert-type self-report scales, yet their development remains labor-intensive, time-consuming, and heavily dependent on subject matter experts. Recent advances in large language models (LLMs) have shown promise for automatic item generation (AIG). Building on these developments, the present study focuses on developing and evaluating a structured and generalizable framework for automatically generating personality SJTs, using GPT-4 and ChatGPT-5 as empirical examples. Three studies were conducted. Study 1 systematically compared the effects of prompt design and temperature settings on the content validity of LLM-generated items to develop an effective and stable LLM-based AIG approach for personality SJT. Results showed that optimized prompts and a temperature of 1.0 achieved the best balance of creativity and accuracy on GPT-4. Study 2 examined the cross-model generalizability and reproducibility of this automated SJT generation approach through multiple rounds. The results showed that the approach consistently produced reproducible and high-quality items on ChatGPT-5. Study 3 evaluated the psychometric properties of LLM-generated SJTs covering five facets of the Big Five personality traits. Results demonstrated satisfactory reliability and validity across most facets, though limitations were observed in the convergent validity of the compliance facet and certain aspects of criterion-related validity. These findings provide robust evidence that the proposed LLM-based AIG approach can produce culturally appropriate and psychometrically sound SJTs with efficiency comparable to or exceeding traditional methods.

cs.CL↗

arXiv2026

Summarization is Not Dead Yet

The progress of large language models (LLMs) has fueled claims that model-generated summaries rival or even surpass human-written references, raising questions about whether summarization remains an open research problem. We re-examine this narrative through a multi-track evaluation covering diverse datasets and state-of-the-art LLMs, combining controlled human assessment, bias-mitigated LLM-as-Judge protocols, factuality verification against external knowledge, and corpus-level linguistic analysis. Our findings reveal a more nuanced landscape in which human references continue to demonstrate advantages in informativeness and faithfulness, whereas LLM outputs are preferred mainly for surface-level coherence and fluency. Factuality verification indicates that human references remain more reliable, particularly for claims involving reasoning or synthesis, and linguistic analysis uncovers a pattern of stylistic homogeneity across different models. These observations suggest that current LLMs have raised the floor of summarization quality, but the ceiling of their performance remains below human capabilities.

cs.CL↗

arXiv2026

Fast radio bursts, magnetars and earthquakes: their "family feud"?

Fast radio bursts (FRBs) are millisecond-duration cosmic transients whose origin remains elusive. Competing models invoke either earthquake-like processes or flare-like mechanisms. To discriminate between these scenarios, we develop a novel diagnostic, the Pincus-Lyapunov diagram (PLD), to characterize the energetic transients in the stochasticity-chaos phase space. We compile burst sequences from five representative FRBs (FRB 20121102A, FRB 20190520B, FRB 20201124A, FRB 20220912A, and FRB 20240114A), together with those from magnetar flares (SGR J1550$-$5418, SGR J0501+4516, SGR 1806$-$20, SGR 1900+14, and SGR J1935+2154), pulsar glitches, solar flares, and earthquakes, and map them onto the PLD for comparative analysis. The resulting diagram shows that FRBs occupy a distinct region of the phase space. Specifically, a permutation test reveals a statistically significant difference in the distributions of magnetar flares and pulsar glitches compared to those of repeating FRBs ($p$-value $\simeq 0.05$). To examine whether temporal variations in source activity can shift a repeater's position in this phase space, we analyze the time evolution of the most prolific repeater, FRB~20240114A. For this repeating FRB, both Pincus Index and Lyapunov Exponent demonstrate statistically stable behaviour over the eight-month observation session, with Augmented Dickey--Fuller tests yielding $p \simeq 1.78\times10^{-3}$ and $9.91\times10^{-3}$, respectively. By assembling the most comprehensive dataset to date, our work indicates that the trigger mechanisms of repeating FRBs are likely to be distinct from those driving magnetar flares, pulsar glitches, solar flares, and earthquakes.

astro-ph.HE↗

arXiv2026

Density properties of orbits for a hypercyclic operator on a Banach space

We study density properties of orbits for a hypercyclic operator $T$ on a separable Banach space $X$, and show that exactly one of the following four cases holds: (1) every vector in $X$ is asymptotic to zero with density one; (2) generic vectors in $X$ are distributionally irregular of type $1$; (3) generic vectors in $X$ are distributionally irregular of type $2\frac{1}{2}$ and no hypercyclic vector is distributionally irregular of type $1$; (4) every hypercyclic vector in $X$ is divergent to infinity with density one. We also present some examples concerned with weighted backward shifts on $\ell^p$ to show that all the above four cases can occur. Furthermore, we show that similar results hold for $C_0$-semigroups.

math.FA↗

arXiv2026

PiT: Progressive Diffusion Transformer

Diffusion Transformers (DiTs) achieve remarkable performance within image generation via the transformer architecture. Conventionally, DiTs are constructed by stacking serial isotropic global modeling transformers, which face significant quadratic computational cost. However, through empirical analysis, we find that DiTs do not rely as heavily on global information as previously believed. In fact, most layers exhibit significant redundancy in global computation. Additionally, conventional attention mechanisms suffer from low-frequency inertia, limiting their efficiency. To address these issues, we propose Pseudo Shifted Window Attention (PSWA), which fundamentally mitigates global attention redundancy. PSWA achieves moderate global-local information through window attention. It further utilizes a high-frequency bridging branch to simulate shifted window operations, which both enrich the high-frequency information and strengthen inter-window connections. Furthermore, we propose the Progressive Coverage Channel Allocation (PCCA) strategy that captures high-order attention without additional computational cost. Based on these innovations, we propose a series of Pseudo Progressive Diffusion Transformer (PiT). Our extensive experiments show their superior performance; for example, our proposed PiT-L achieves 54% FID improvement over DiT-XL/2 while using less computation.

cs.CV↗

arXiv2026

Impact of Nuclear Level Density on $r$-Process Rare-Earth Peak Nucleosynthesis

The rare-earth peak ($A\sim164$) is a prominent feature of the $r$-process, and previous theoretical studies suggest that it is possibly linked to local nuclear structural effects. However, the nuclear level density (NLD), a physical quantity directly reflecting these properties, has been largely overlooked compared to other structural properties such as nuclear masses. To address this, we perform $r$-process simulations across three astrophysical scenarios using neutron-capture rates derived from six distinct NLD models. Our results reveal that microscopic models yield systematic deviations in NLD relative to phenomenological ones, leading to critical impacts on nucleosynthesis. Specifically, systematic NLD differences in even-$A$ nuclei redirect the nuclear flow, accelerating the early formation of the rare-earth peak and temporarily enhancing its magnitude. This underlying structural shift also fundamentally alters the $r$-process sensitivity to the neutron-capture rate, effectively eliminating its dependence on the odd-even nature of protons. Overall, these findings demonstrate that the internal nuclear structure encoded within NLDs can collectively induce a global redirection of the nucleosynthesis pathway, highlighting the critical need for self-consistent microscopic inputs in future simulations.

nucl-th↗

arXiv2026

Cost-Utility Alignment in LLM Agent Trajectories:Profiling,Attribution,Diagnosis,Adaptation,and Evaluation

LLM agents execute tasks through multi-step trajectories that accumulate cost in tokens, latency, monetary fees, and environmental risk while producing utility only at the aggregate task level. Prior surveys address inference optimization, agent capabilities, or evaluation in isolation, leaving practitioners without principled tools to determine whether a trajectory's resource expenditure is justified by its task contribution. We address this gap by developing a trajectory-centric cost-utility alignment framework that treats resource consumption and task contribution as dual ledgers over the same execution, organized around five analytical stages: cost profiling, utility attribution, misalignment diagnosis, targeted adaptation, and evaluation. Utility attribution is central to this structure: rather than relying on aggregate outcomes, it organizes contribution methods by evidential strength, from process proxies and information dependency to counterfactual replay, supplying the causal evidence that grounds diagnosis and guides adaptation. Using this framework, we analyze recent agent systems, attribution methods, and evaluation protocols covering efficiency, reliability, and economic value, as well as five forms of misalignment spanning cognitive and context use, external interaction, recovery-loop control, resource-capability allocation, and multi-agent coordination, together with their targeted adaptations. The result is a closed analytical loop connecting the cost side of agent execution to its utility side, providing a structured basis for resource-aware agent design and deployment.

cs.SE↗

arXiv2026

Continuous-Time Quantum Walks based Graph Neural Network

Graph Neural Networks (GNNs) are widely used on graph-structured data, but most suffer from two key weaknesses. First, message passing behaves as a low-pass filter under the homophily assumption, leading to poor performance on heterophilic graphs. Second, stacking layers drives node features toward constants, causing over-smoothing. Existing methods usually address these issues separately, while the few joint solutions rely largely on empirical heuristics, and many over-smoothing remedies sacrifice model expressiveness. We propose \textbf{CTQW-GNN}, a GNN based on Continuous-Time Quantum Walks (CTQW), to address both issues with theoretical justification. Its design exploits two properties of the CTQW propagator $e^{-\mathrm{i}Ht}$. First, it is unitary and has eigenvalues on the unit circle, so no frequency component is damped, counteracting the low-pass bias. Second, unitarity preserves feature norms and prevents the Dirichlet energy from decaying exponentially with depth, thereby mitigating over-smoothing. CTQW-GNN combines three complementary aggregation modules. \textit{CTQW-based Aggregation} evolves node features through the unitary propagator, preserving mid- and high-frequency signals for heterophilic graphs while preventing Dirichlet-energy collapse. \textit{CTQW-Attention Aggregation} constructs a multi-hop neighbor graph from CTQW amplitudes and applies attention over it, enabling access to distant homophilic nodes missed by single-hop aggregation. \textit{LF Aggregation} uses a standard low-pass GAT branch to retain strong performance on homophilic graphs, where pure CTQW aggregation can be suboptimal. We further provide a spectral-gap analysis explaining energy preservation and a Lieb--Robinson-type bound that gives a principled rule for selecting the walk time $t$.

cs.AI↗

arXiv2026

Power growth of mean-L-stable operators on Banach spaces

We study the growth of powers of mean-L-stable operators on Banach spaces. We show that, for linear operators, mean-L-stability is equivalent to uniform boundedness in density; this yields $\|T^n\|=O(n)$ on every Banach space. On Hilbert spaces we prove that mean-L-stability is equivalent to absolute Cesàro boundedness and obtain $\|T^n\|=O(n^{1/2-c_T})$ for some $c_T>0$. For positive mean-L-stable operators on abstract $L^p$-spaces, $1\le p<\infty$, we similarly obtain $\|T^n\|=O(n^{1/p-c_T})$, where in both cases the positive constant $c_T$ cannot be chosen uniformly over all such operators. For positive mean-L-stable operators on $p$-convex Banach lattices, we prove the bound $O(n^{1/p})$ and construct positive topologically mixing operators $T_p$ for which $\|T_p^n\|\asymp n^{1/p}$, where $1\leq p<\infty$. These operators satisfy a uniform weak $(p,p)$ orbit estimate, while the averages of $\|T_p^nx\|^s$ are bounded for $s p$. The operator $T_1$ is uniformly Kreiss bounded and has linear power growth, answering a question of Montes-Rodríguez, Sánchez-Álvarez and Zemánek (2005). Moreover, $T_1$ is mean-L-stable and mean Li--Yorke chaotic, while it is not distributionally chaotic. This answers a question of Bernardes, Bonilla and Peris (2020).

math.FA↗

arXiv2026

AgentRewind: Recoverable Execution for Long-Horizon LLM Agents

Many real-world tasks require LLM agents to interact with their environments over long execution horizons. Errors that occur early in execution may propagate through both the agent context and environment state, and their effects may be difficult to reverse through subsequent actions. Existing methods mainly seek to reduce such errors through plan refinement and safety checks but provide little support after errors occur. To enable recovery during long-horizon execution, we present AgentRewind, a runtime recovery framework that records aligned checkpoints of the agent context and controlled environment, allowing agents to return to an earlier state and resume execution with information from previous attempts. We also construct MettleBench, a benchmark for evaluating task completion and partial progress on long-horizon engineering assignments containing a series of related requirements. Experiments across tasks, multiple models, execution strategies, and agent harnesses show that AgentRewind improves task success rate and average checklist progress over the compared baselines.

cs.AI↗