arXiv subjects

Jun Chen

Publications and source records attributed to Jun Chen.

At least 37 records · Page 2Linked to original sources

arXiv2026

How Sensitive Are LLM Leaderboard Claims to Hidden Model Selection?

LLM leaderboard gains can reflect selection among privately evaluated model variants, yet neither the number of variants nor their dependence is public. We ask how many hidden variants a published margin can support while retaining statistical evidence of a provider's advantage over a fixed comparator. For a fixed candidate family under a Gaussian margin model, we derive a sensitivity curve that reports this maximum count as a function of a lower bound on within-family correlation. The relevant correlation must match the score used for ranking and the sampling model: in a controlled family, pooled item correlation is 0.90, whereas composite-score correlation is 0.46 under item resampling and 0.92 when MMLU subjects are resampled. An item-based audit of 394 adjacent-rank claims on the Open LLM Leaderboard finds that 391 lack statistical support even before accounting for selection. Among claims that pass the uncorrected test, certification can depend on assumptions about the hidden family's correlation. The resulting curves make these assumptions explicit without estimating the unobserved search size.

stat.ML↗

arXiv2026

Vacancy-mediated nitrogen diffusion and aggregation via high-fluence electron beam irradiation in HPHT synthesized diamond crystal

The negatively charged nitrogen vacancy (NV-) center in diamond is a promising point defect for highly sensitive quantum sensing. The formation of high-density NV- centers is essential for improving sensitivity. We performed room-temperature electron beam irradiation (EBI) and annealing on nitrogen-doped high-pressure high-temperature diamond crystals, aiming to convert all substitutional nitrogen into NV structures by increasing EBI fluence. While the Ns0 to NV0 and NV- conversion process dominated at low EBI fluences, in a high-fluence region, NV0 and NV- center and Ns0 and Ns+ concentrations decreased with increasing EBI fluence, indicating the formation of unknown nitrogen-related defects such as the H3 center, which is a nitrogen and vacancy aggregation defect. Although H3 centers were observed at high EBI fluence, their annealing temperature of 1375 +- 25 °C was lower than the typically reported temperatures over 1600 °C. We attribute this low-temperature formation to vacancy-mediated nitrogen diffusion and aggregation.

cond-mat.mtrl-sci↗

arXiv2026

BioEVAL: A global, multi-institutional benchmark of large language and multimodal models for bioengineering

Large Language Models (LLMs) have demonstrated historic breakthroughs in general reasoning with early successes in biomedical science. However, existing LLM benchmarking emphasizes factual recall, offering limited insight into model performance on frontier and multimodal tasks. We assembled BioEVAL (BioEngineering Validation of AI and LLMs), a global, multi-institutional initiative designed to assess experimental reasoning capability across bioengineering (BE) subfields. BioEVAL spans 11 major BE subfields plus a set of uncategorized items, bringing together 22 research groups to create a PhD-level benchmark comprising 608 evaluation items: 1) 380 multiple-choice questions (MCQs, 359 retained after audit), 2) 218 literature synthesis tasks, and 3) 10 multimodal problems with experimental image interpretation. Benchmark items underwent authoring-group expert review and centralized quality control before evaluation. Following evaluation, a blinded cross-group consensus audit of the highest- and lowest-accuracy MCQ items flagged 21 questions for revision or removal; these were withheld, and all reported MCQ results are computed on the 359 retained items. We evaluated diverse cloud-scale foundation/multimodal models (e.g., ChatGPT, Gemini, and Grok) and locally deployable models suitable for inference on consumer-grade GPUs. Models achieved the highest accuracy of up to 90% on MCQs, similarity score of 0.72 on literature synthesis, and accuracy of 80% on a small sample of multimodal reasoning questions, with substantial performance variation across subfields. Leaderboard rankings characterize current capabilities, limitations, and development priorities across the evaluated BE task categories. BioEVAL is maintained as an extensible benchmark with standardized protocols for continuing expert item contribution and model evaluation.

cs.AI↗

arXiv2026

Privacy-Preserving Prompted Policy Search for Robotic Control

Large language models (LLMs) have recently demonstrated promising capabilities as in-context policy optimizers for Reinforcement Learning (RL), enabling policy search driven by both numerical reward signals and natural language reasoning. However, deploying such methods in practice requires transmitting raw policy parameters and rewards history to cloud-based LLM APIs, exposing proprietary control strategies to third-party service providers. To address this issue, this paper introduces Privacy-Preserving Prompted Policy Search (PP-ProPS), a framework that enables LLM-guided policy optimization while keeping policy and environmental parameters confidential. PP-ProPS encodes policy parameters and reward values using secret client-side transformations before they are included in each API request, ensuring that the LLM provider observes only encoded policy parameters and scaled reward information. Furthermore, unlike Vanilla ProPS, the proposed framework does not require the true optimal episodic return to be known or disclosed to the LLM. Beyond protecting the optimization data, PP-ProPS improves the search process in two ways. First, it provides the LLM with individual reward components instead of only a single total return, offering more informative feedback about each candidate policy. Second, it uses a bounded history that prevents the prompt from growing indefinitely, improving search with high-dimensional policies and supporting the use of open-weight LLMs. The proposed PP-ProPS is evaluated on both continuous and discrete control problems spanning Multi-Joint dynamics with Contact (MuJoCo) locomotion, classic control, highway driving, and robotic arm manipulation. Compared to Vanilla ProPS, the proposed PP-ProPS outperforms ProPS in seven of the ten evaluated tasks, and surpasses conventional RL methods including PPO, SAC, and TRPO, in five of the six tasks.

cs.RO↗

arXiv2026

Parametric Study of the Torus Instability Threshold

The torus instability of an arched current channel has been suggested to initiate and drive major solar and stellar eruptions. Its threshold, given by the critical decay index of the equilibrium external poloidal field (the so-called strapping field) at the position of the current channel, is insufficiently known. Here, we carry out a parametric numerical study of the threshold, employing the force-free Titov-Démoulin (TD) equilibrium of a line-tied partial toroidal current channel and flux rope. This addresses the scatter of the threshold about its canonical value, $n_\mathrm{cr}=3/2$. Values scattering in the range $n_\mathrm{cr}\approx$\,1--2 are typically found in numerical and observational studies of flux rope eruptions on the Sun. For zero external toroidal (guide, or shear) field and approximately semicircular geometry (corresponding to minimal photospheric line-tying), we find the threshold to lie in the theoretically expected range of $\approx$\,1--1.5. An external toroidal field introduces a strong stabilizing effect on the instability, raising the threshold up to $\sim$\,2.5, which can explain observational and numerical results above the canonical value. Line-tying is found to act as a stabilizer as well. We also consider the approximate threshold based on the potential field and find a very good agreement with the exact numerical value, provided the horizontal component perpendicular to the flux rope axis is used to approximate the external poloidal field.

astro-ph.SR↗

arXiv2026

An Overview of Rate-Distortion-Perception Theory

This paper provides a comprehensive overview of the rate-distortion-perception framework, tracing its evolution from mathematical theory to practical deployment. We review the mathematical definition of the the mathematical definition of the Blau--Michaeli function and the associated coding theorems, examine computational methods for its evaluation, and discuss alternative formulations of the framework. Operational principles and design guidelines for modern image and video coding are also presented, highlighting the rate-distortion-perception framework as a rigorous foundation for developing next-generation perceptual compression systems.

cs.IT↗

arXiv2026

Dual-Process Atomic Skill Learning: Decoupling Semantic Reasoning and Real-Time Control

Language-conditioned Imitation Learning (IL) is essential for enabling robots to perform complex tasks following natural language instructions. However, generalizing to multi-step compositional tasks remains a significant challenge. While hierarchical approaches attempt to address this by decomposing tasks into atomic skills, existing methods often suffer from training instability and codebook collapse due to the tight coupling between high-level skill reasoning and low-level action generation in joint training paradigms. Inspired by the Dual-Process Theory of cognition, we propose Dual-Process Atomic Skill Learning (DASL), a novel asynchronous hierarchical imitation learning framework that decouples slow semantic reasoning from fast, real-time motion control. DASL comprises a Slow-Frequency Policy that predicts interpretable, discrete skills via Vector Quantization, and a High-Frequency Policy that leverages a latent diffusion model and a Decision Transformer to generate precise actions conditioned on these latent skills. By asynchronously coordinating these modules and utilizing diffusion to structure the latent space, our framework mitigates the skill codebook interference problem common in joint training paradigms. Evaluations across simulation benchmarks and experiment demonstrate that DASL significantly outperforms state-of-the-art baselines, excelling in skill acquisition and compositional generalization to unseen instructions. GitHub page: https://github.com/Hatakekaka/DASL

cs.RO↗

arXiv2026

StepAudio 3 Gen Technical Report

We introduce StepAudio 3 Gen, a general-purpose audio generation model that supports zero-shot text-to-speech (TTS), voice design, vocal generation, sound effects, music, vibe speech, and mixtures of multiple audio types within a unified framework. At its core, StepAudio 3 Gen is a discrete autoregressive generator that models audio directly over residual vector quantization (RVQ) tokens, departing from the diffusion Transformer-based continuous generation paradigm prevalent in recent general audio models. Its StepAudio Tokenizer represents general audio at 12.5 Hz in a shared $16 \times 2048$ residual code space, jointly quantizing semantic and waveform-level acoustic features so that each code layer preserves both types of information. For generation, the backbone predicts the first codebook along the time axis using autoregressive modeling, while a lightweight causal Transformer completes the remaining fifteen codebooks along the codebook axis. Our study further identifies three key design principles: (1) interference-aware progressive pretraining for acquiring audio capabilities while preserving the textual abilities of the large language model, (2) RVQ Adaptor for effectively incorporating multi-codebook acoustic representations, and (3) discrete autoregressive modeling over a shared representation across general audio domains. With progressive pretraining, multi-task instruction training, and supervised fine-tuning, StepAudio 3 Gen achieves state-of-the-art performance on both TTS and voice design, while retaining strong generation capabilities across speech, vocals, sound effects, and music. Audio samples are available at https://stepaudiollm.github.io/step-audio-3-gen/.

cs.SD↗

arXiv2026

ChatPlanner: A Large Language Model Framework for Personalized Public Transit Routing

Personalized public transit routing in public transit systems remains challenging due to the difficulty of capturing and integrating diverse user preferences into routing algorithms. This paper presents ChatPlanner, a novel framework that leverages Large Language Models (LLMs) to enable preference-aware public transit routing. Our approach employs fine-tuned LLMs with Retrieval-Augmented Generation (RAG) to extract routing parameters and interpret conversationally expressed preferences from natural language queries as preference scores, subsequently integrating these preferences into the objective function of a public transit routing algorithm. This study designs preference-aware datasets incorporating eight personas and five contexts to establish scoring standards for both fine-tuning and RAG. This work conducted four experiments to validate the solutions' feasibility, extraction of routing information and preferences, solution set quality and completeness, and latency and computational tractability. Results demonstrate that ChatPlanner generates feasible solutions reliably. Fine-tuning enforces the required output structure and learns general preference patterns, while RAG provides query-specific context to resolve imprecise or conversational expressions and calibrate continuous scores. The combination of both achieves the highest accuracy in routing information extraction and rubric-consistent user preference interpretation. Results based on selected case studies show that by capturing user conversationally expressed preferences, ChatPlanner identifies preference-relevant solutions across different dimensions that existing route planners overlook, generating more route alternatives. The latency evaluation confirms that the framework is computationally tractable. This research establishes a new paradigm for integrating natural language understanding into transportation optimization.

cs.AI↗

arXiv2026

Bi-HYCO: Bi-Objective Cooperative Learning for PDE Parameter Identification under Fragmented Observations

Physical and synthetic models may describe complementary aspects of the same PDE-governed system while receiving different, possibly fragmented, observations. We propose Bi-Objective HYCO (Bi-HYCO), a cooperative framework that retains both representations and their local observational objectives while coupling their predicted states at unlabeled interaction points. These points contain no measurements and do not augment the data; they provide a communication mechanism in the common state space. The two criteria form a vector-valued objective, and weighted scalarizations provide computational realizations. For the deterministic shared-observation algorithm with fixed interaction points, we prove sufficient decrease and finite length of the whole alternating sequence, which converges to a mixed critical point under the stated Kurdyka-Lojasiewicz-type assumptions. Elliptic transmission and two-dimensional Navier-Stokes experiments assess parameter and state reconstruction, noise and scalarization effects, and PINN/XPINN references. Ablations show that removing state interaction while retaining aggregation deteriorates parameter recovery in the tested configurations, particularly for Navier-Stokes.

cs.LG↗

arXiv2026

EvoGS: Modeling Deformation Evolution for Dynamic Gaussian Splatting

Recent extensions of 3D Gaussian Splatting (3DGS) enable real-time novel view synthesis in dynamic scenes by learning time-conditioned Gaussian deformations. However, existing MLP-based methods typically estimate deformations independently at each timestamp, making them less robust to large or abrupt motions. To address this issue, we propose \textbf{EvoGS}, a 3DGS-based dynamic reconstruction framework that models Gaussian deformation as a temporal evolution process. EvoGS maintains persistent deformation states for each Gaussian, extrapolates future states from historical deformation states, and corrects the predictions with MLP-derived observations. The correction is adaptively weighted using a temporal residual memory and evolution statistics such as deformation velocity and trajectory deviation. To further improve reconstruction quality, EvoGS introduces deformation-aware densification. Clone and split operations are performed along corrected deformation directions, while an uncertainty-aware strategy suppresses densification for Gaussians with unstable deformation histories. Experiments show that EvoGS improves dynamic novel view synthesis quality and achieves competitive performance across benchmarks.

cs.CV↗

arXiv2026

GTA-RAG: Graph-Trajectory-Augmented Reinforcement Learning for Multi-Turn Retrieval-Augmented Reasoning

Retrieval-augmented generation (RAG) enables LLMs to access external knowledge for answering knowledge-intensive questions. For complex multi-hop questions, multi-turn retrieval-augmented reasoning extends RAG into an iterative process that repeatedly searches for and integrates evidence across documents. However, existing reinforcement-learning (RL) approaches for agentic RAG are typically optimized with final-answer rewards, which provide sparse supervision and overlook whether the model actually retrieves the required evidence chain. We present \textsc{GTA-RAG}, a graph-trajectory-augmented RL framework for multi-turn retrieval-augmented reasoning. From an entity--document graph, we sample connected document paths, synthesize multi-hop QA trajectories, and validate them with the deployed retriever to obtain executable trajectory-level supervision. We then optimize the retrieval policy with Group Relative Policy Optimization (GRPO) and a trajectory-guided reward that encourages both accurate answers and acquisition of target evidence documents, followed by answer-reward training on natural QA instances. Experiments on three multi-hop and two simple QA benchmarks show that \method{} consistently outperforms RL-based RAG baselines with both Qwen2.5-3B and Qwen2.5-7B backbones, while substantially improving evidence-chain coverage. Our code is available at https://github.com/cjcj46262/GTA-RAG.

cs.CL↗

arXiv2026

Imaging and characterization of spontaneous vortices in a proximity-induced superconductor

Observation of spontaneous symmetry breaking is crucial for understanding continuous second-order phase transitions from disordered to ordered states, which often leads to the formation of topological defects. In superconductors, such topological defects manifest as quantized vortices. However, the formation and observation of spontaneous vortices in a uniform superconductor are challenging because extremely rapid cooling (>108 K/s) is generally required for that purpose. Here we conducted scanning superconducting quantum interference device microscope (SSM) measurements on an MgB2-based proximity-induced superconductor, an intrinsically inhomogeneous system. In this system, individual superconducting domains will reach internal equilibrium independently during cooling and choose their own phase before the global phase coherence is established via the long-range proximity coupling. The SSM measurements demonstrate that vortices are nucleated spontaneously even at a relatively slow cooling rate (~0.2 K/s). We also find that the vortices with different polarities, sizes, and shapes appear stochastically under near-zero-field conditions. The geometry of the spontaneous vortices is more extended than that of the field-induced Abrikosov vortices. Magnetic field profile analysis based on the London model elucidates that penetration depths of the extended vortices are anomalously large, exceeding several micrometers. This unusual morphology of the spontaneous vortices most likely imprints the information that is frozen at the moment of vortex formation. Our findings not only provide insights into the local phase differences present in the early stage of the phase transition in this proximity-induced superconducting system, but they also shed insights into the structure, formation, and stabilization of topological defects in highly disordered and inhomogeneous superconducting systems.

cond-mat.supr-con↗

arXiv2026

MDCPP: Multi-Robot Dynamic Coverage Path Planning for Workload Adaptation

Multi-robot coverage path planning commonly balances geometric area or path length under a constantspeed assumption. This assumption is inadequate when sensing or interaction tasks cause spatially varying traversal speeds, because equal areas can induce markedly different completion times. We propose Multi-Robot Dynamic Coverage Path Planning (MDCPP), which learns a Gaussian-mixture workload field from partial observations, predicts cell-wise service times, and repeatedly repartitions the uncovered cells through a distributed capacity-constrained assignment. We establish finite termination and pairwise local optimality of each synchronized assignment round, bound the service-time makespan degradation due to estimation error, and state sufficient conditions for complete coverage. A 600-run benchmark against sweeping, LS-MCPP, reactive reassignment, and an oracle shows that prediction is most valuable under strong heterogeneity and improves aggregate paired makespan over the nonpredictive alternatives. A three-UGV experiment further validates route execution and spatial speed adaptation under localization, drivetrain, and wireless-control effects.

cs.RO↗

arXiv2026

Deformations and homotopy theory of Rota-Baxter Lie algebras

For Rota-Baxter Lie algebras, a homotopy cooperad is exhibited, whose cobar construction is shown to be the minimal model of the operad of Rota-Baxter Lie algebras by using algebraic Morse theory. The deformation complex of Rota-Baxter Lie algebras as well as the $L_\infty$-algebra structure on this complex are deduced from the minimal model and the notion of homotopy Rota-Baxter Lie algebras is given as a consequence.

math.KT↗

arXiv2026

Heavy quark collisional energy loss in a nonextensive quark-gluon plasma

In this study, we derive the longitudinal and transverse gluon self-energies and the corresponding dielectric functions for a nonextensive QGP, based on nonextensive statistical mechanics and a kinetic theory framework. The nonextensive parameter $q$ enters these quantities primarily through the modification of the Debye mass. Utilizing the derived dielectric functions, we then calculate the collisional energy loss for a heavy quark using two established formalisms: the plasma physics-based Thoma-Gyulassy formula and the thermal field theory-originated Kirzhnits-Thoma formula. Our results show that for both formalisms, the collisional energy loss increases with the nonextensive parameter $q$ with this enhancement being more significant at higher incident quark momenta and suppressed for a heavier quark mass. The energy loss predicted from the Kirzhnits-Thoma formula is substantially larger than that from the Thoma-Gyulassy formula, and the nonextensive effect on the energy loss is more pronounced in the former. Furthermore, the mass suppression of the nonextensive effect on the energy loss is weaker in the Kirzhnits-Thoma approach. These calculations demonstrate that nonextensive statistics can significantly alter the energy loss in the QGP.

hep-ph↗

arXiv2026

Kimi K2.5: Visual Agentic Intelligence

We introduce Kimi K2.5, an open-source multimodal agentic model designed to advance general agentic intelligence. K2.5 emphasizes the joint optimization of text and vision so that two modalities enhance each other. This includes a series of techniques such as joint text-vision pre-training, zero-vision SFT, and joint text-vision reinforcement learning. Building on this multimodal foundation, K2.5 introduces Agent Swarm, a self-directed parallel agent orchestration framework that dynamically decomposes complex tasks into heterogeneous sub-problems and executes them concurrently. Extensive evaluations show that Kimi K2.5 achieves state-of-the-art results across various domains including coding, vision, reasoning, and agentic tasks. Agent Swarm also reduces latency by up to $4.5\times$ over single-agent baselines. We release the post-trained Kimi K2.5 model checkpoint to facilitate future research and real-world applications of agentic intelligence.

cs.CL↗

arXiv2026

Kimi K3: Open Frontier Intelligence

We introduce Kimi K3, a 2.8T parameter Mixture-of-Experts model with 104 billion activated parameters, native vision capabilities, and a 1-million-token context window. Kimi K3 is built on Kimi Delta Attention and Attention Residuals, which improve information flow across sequence length and model depth. Together with Stable LatentMoE, which effectively activates 16 of 896 routed experts per token, and refined training and data recipes, these advances yield an approximately 2.5x improvement in overall scaling efficiency over Kimi K2. Post-training highlights reinforcement learning across general, agentic, and coding domains and multiple reasoning-effort levels, enabling compositional generalization and robust long-horizon execution. At 2.8T scale, Kimi K3 is supported by infrastructure advances in multiple areas: algorithm-system co-design for KDA, perfectly balanced expert-parallel training with efficient memory management, million-token agentic RL with persistent rollout and sandbox states, and deployment innovations. Extensive evaluations show that Kimi K3 achieves frontier-level performance across long-horizon coding, agentic, knowledge, reasoning, and vision tasks. While its overall performance still trails the most powerful proprietary models, namely Claude Fable 5 and GPT-5.6 Sol, Kimi K3 consistently outperforms other open and proprietary models evaluated in our suite. We release the full Kimi K3 model weights to facilitate future research and accelerate the broader deployment and adoption of frontier intelligence.

cs.CL↗