arXiv subjects

Pengcheng Li

Publications and source records attributed to Pengcheng Li.

At least 37 records · Page 2Linked to original sources

arXiv2026

PointCoT: A Multi-modal Benchmark for Explicit 3D Geometric Reasoning

While Multimodal Large Language Models (MLLMs) demonstrate proficiency in 2D scenes, extending their perceptual intelligence to 3D point cloud understanding remains a significant challenge. Current approaches focus primarily on aligning 3D features with pre-trained models. However, they typically treat geometric reasoning as an implicit mapping process. These methods bypass intermediate logical steps and consequently suffer from geometric hallucinations. They confidently generate plausible responses that fail to ground in precise structural details. To bridge this gap, we present PointCoT, a novel framework that empowers MLLMs with explicit Chain-of-Thought (CoT) reasoning for 3D data. We advocate for a \textit{Look, Think, then Answer} paradigm. In this approach, the model is supervised to generate geometry-grounded rationales before predicting final answers. To facilitate this, we construct Point-Reason-Instruct, a large-scale benchmark comprising $\sim$86k instruction-tuning samples with hierarchical CoT annotations. By leveraging a dual-stream multi-modal architecture, our method synergizes semantic appearance with geometric truth. Extensive experiments demonstrate that PointCoT achieves state-of-the-art performance on complex reasoning tasks.

cs.CV↗

arXiv2026

Effects of the centrality determination method for the equation of state and nucleonic observables from Au+Au collisions at $\sqrt{s_{NN}}$ = 2.4 GeV

Centrality determination remains one of the major sources of systematic uncertainty in intermediate-energy heavy-ion collision analyses, especially for probing the nuclear equation of state (EoS) at supra-saturation densities. To quantitatively assess the uncertainties associated with different centrality determination methods and to investigate their effects on final-state EoS-sensitive observables. Within the ultra-relativistic quantum molecular dynamics (UrQMD) model, Au+Au collisions at $\sqrt{s_{NN}}$=2.4 GeV are performed within a soft and a hard EoS. Event centrality is determined using the multiplicity of all charged particles ($M_\mathrm{ch}$) and two impact parameter-based centrality filters, one based on a geometrical interpretation and the other based on the Glauber Monte Carlo (MC) model, denoted as $b_{f}$ and $b_{r}$, respectively. It is shown that there exist significant differences between the real impact parameter distributions of event samples selected by $M_\mathrm{ch}$, $b_{f}$, and $b_{r}$, particularly between $M_\mathrm{ch}$ and $b_{r}$. When the $b_{f}$ is employed, uncertainties associated with centrality selection have a weaker influence on observables than the effects induced by the EoS. In contrast, when the $b_{r}$ is used, the influence of centrality-related uncertainties becomes more pronounced than that of the EoS. These results demonstrate that a rigorous and consistent mapping between $M_\mathrm{ch}$ and impact parameter is essential to impose quantitative constraints on the high-density nuclear EoS. Furthermore, our study indicates that the geometrical interpretation of centrality remains valid and consistent with dynamical multiplicity selection, whereas the Glauber MC-based centrality determination becomes unreliable at the investigated energy.

nucl-th↗

arXiv2026

ThinkDrive: Chain-of-Thought Guided Progressive Reinforcement Learning Fine-Tuning for Autonomous Driving

With the rapid advancement of large language models (LLMs) technologies, their application in the domain of autonomous driving has become increasingly widespread. However, existing methods suffer from unstructured reasoning, poor generalization, and misalignment with human driving intent. While Chain-of-Thought (CoT) reasoning enhances decision transparency, conventional supervised fine-tuning (SFT) fails to fully exploit its potential, and reinforcement learning (RL) approaches face instability and suboptimal reasoning depth. We propose ThinkDrive, a CoT guided progressive RL fine-tuning framework for autonomous driving that synergizes explicit reasoning with difficulty-aware adaptive policy optimization. Our method employs a two-stage training strategy. First, we perform SFT using CoT explanations. Then, we apply progressive RL with a difficulty-aware adaptive policy optimizer that dynamically adjusts learning intensity based on sample complexity. We evaluate our approach on a public dataset. The results show that ThinkDrive outperforms strong RL baselines by 1.45%, 1.95%, and 1.01% on exam, easy-exam, and accuracy, respectively. Moreover, a 2B-parameter model trained with our method surpasses the much larger GPT-4o by 3.28% on the exam metric.

cs.AI↗

arXiv2026

AIVD: Adaptive Edge-Cloud Collaboration for Accurate and Efficient Industrial Visual Detection

Multimodal large language models (MLLMs) demonstrate exceptional capabilities in semantic understanding and visual reasoning, yet they still face challenges in precise object localization and resource-constrained edge-cloud deployment. To address this, this paper proposes the AIVD framework, which achieves unified precise localization and high-quality semantic generation through the collaboration between lightweight edge detectors and cloud-based MLLMs. To enhance the cloud MLLM's robustness against edge cropped-box noise and scenario variations, we design an efficient fine-tuning strategy with visual-semantic collaborative augmentation, significantly improving classification accuracy and semantic consistency. Furthermore, to maintain high throughput and low latency across heterogeneous edge devices and dynamic network conditions, we propose a heterogeneous resource-aware dynamic scheduling algorithm. Experimental results demonstrate that AIVD substantially reduces resource consumption while improving MLLM classification performance and semantic generation quality. The proposed scheduling strategy also achieves higher throughput and lower latency across diverse scenarios.

cs.CV↗

arXiv2025

SOFTooth: Semantics-Enhanced Order-Aware Fusion for Tooth Instance Segmentation

Three-dimensional (3D) tooth instance segmentation remains challenging due to crowded arches, ambiguous tooth-gingiva boundaries, missing teeth, and rare yet clinically important third molars. Native 3D methods relying on geometric cues often suffer from boundary leakage, center drift, and inconsistent tooth identities, especially for minority classes and complex anatomies. Meanwhile, 2D foundation models such as the Segment Anything Model (SAM) provide strong boundary-aware semantics, but directly applying them in 3D is impractical in clinical workflows. To address these issues, we propose SOFTooth, a semantics-enhanced, order-aware 2D-3D fusion framework that leverages frozen 2D semantics without explicit 2D mask supervision. First, a point-wise residual gating module injects occlusal-view SAM embeddings into 3D point features to refine tooth-gingiva and inter-tooth boundaries. Second, a center-guided mask refinement regularizes consistency between instance masks and geometric centroids, reducing center drift. Furthermore, an order-aware Hungarian matching strategy integrates anatomical tooth order and center distance into similarity-based assignment, ensuring coherent labeling even under missing or crowded dentitions. On 3DTeethSeg'22, SOFTooth achieves state-of-the-art overall accuracy and mean IoU, with clear gains on cases involving third molars, demonstrating that rich 2D semantics can be effectively transferred to 3D tooth instance segmentation without 2D fine-tuning.

cs.CV↗

arXiv2025

Probing the Three-dimension Emission Source and Neutron Skin via $π$-$π$ Correlations in Heavy-Ion Collisions

The Richardson-Lucy algorithm is applied to reconstruct the three-dimensional source function of identical pions from their two-particle correlation functions. The algorithm's performance is first evaluated through simulations with Gaussian-type initial source functions. Its imaging quality and robustness are further demonstrated with experimental data from Au+Au collisions at 1.23 A GeV, collected by the HADES Collaboration. Additionally, using UrQMD simulations of Pb+Pb collisions at 1.5 A GeV, we show that the deblurred source functions exhibit sensitivity to the initial neutron skin thickness of the colliding nuclei. This highlights the potential of the Richardson-Lucy algorithm as a tool for probing the neutron density distribution in heavy nuclei.

nucl-ex↗

arXiv2025

Label-Free Intraoperative Imaging of Hemodynamics using Deep Learning

Intraoperative visualization of hemodynamics is crucial for accurate diagnosis and informed surgical decision-making. In neurosurgery, indocyanine green fluorescence imaging (ICG-FI) is the gold standard for assessing blood flow and identifying vascular structures. However, it is limited by time-consuming data acquisition, mandatory waiting periods, potential allergic reactions, and operational complexities. Label-free alternatives, such as laser speckle contrast imaging (LSCI) and white light imaging (WLI), offer real-time vascular assessment but cannot resolve arterial-venous differentiation or blood flow direction determination. To address these challenges, we present a label-free cross-modal generation framework to synthesize mean transition time (MTT) maps from LSCI and WLI. MTT maps encode local hemodynamics, enabling artery-vein differentiation and flow direction inference. Experimental validation in rat brains demonstrates that the proposed method presents clear vasculature delineation, accurate artery-vein differentiation, and reliable blood flow direction decoding, while reducing total imaging time by 95.8% compared to conventional ICG protocols. This approach offers a fast, efficient, and contrast-free solution for continuous intraoperative surgical guidance.

physics.med-ph↗

arXiv2025

Unlocking the initial neutron density distribution from the two-pion HBT correlation function in heavy-ion collisions

Revealing the neutron density distribution in the nucleus is one of the crucial tasks of nuclear physics. Within the framework of the ultrarelativistic quantum molecular dynamic model followed by a correlation afterburner program, we investigate the effects of the initial neutron density distribution on the charged-pion yield ratio $π^{-}/π^{+}$, the two-pion momentum correlation function, and the emission source dimension. It is found that the $π^{-}/π^{+}$ ratio is sensitive to the initial neutron density distribution and the impact parameter, especially for collisions at large impact parameter. However, the charge splitting in the correlation functions between positively $π^{+}π^{+}$ and negatively $π^{-}π^{-}$, as well as the source radii and volumes extracted exhibit a stronger dependence on the initial neutron density distribution, but a weaker dependence on the impact parameter. The present study highlights that $π^{+}π^{+}$ and $π^{-}π^{-}$ correlation functions in heavy-ion collisions could be used to probe the initial neutron density distribution of nuclei.

nucl-th↗

arXiv2025

EMO-RL: Emotion-Rule-Based Reinforcement Learning Enhanced Audio-Language Model for Generalized Speech Emotion Recognition

Although Large Audio-Language Models (LALMs) have exhibited outstanding performance in auditory understanding, their performance in affective computing scenarios, particularly in emotion recognition, reasoning, and subtle sentiment differentiation, remains suboptimal. Recent advances in Reinforcement Learning (RL) have shown promise in improving LALMs' reasoning abilities. However, two critical challenges hinder the direct application of RL techniques to Speech Emotion Recognition (SER) tasks: (1) convergence instability caused by ambiguous emotional boundaries and (2) limited reasoning ability when using relatively small models (e.g., 7B-parameter architectures). To overcome these limitations, we introduce EMO-RL, a novel framework incorporating reinforcement learning with two key innovations: Emotion Similarity-Weighted Reward (ESWR) and Explicit Structured Reasoning (ESR). Built upon pretrained LALMs, our method employs group-relative policy optimization with emotion constraints. Comprehensive experiments demonstrate that our EMO-RL training strategies can significantly enhance the emotional reasoning capabilities of LALMs, attaining state-of-the-art results on both the MELD and IEMOCAP datasets, and cross-dataset experiments prove the strong superiority of generalization.

cs.SD↗

arXiv2025

Bayesian analysis of properties of nuclear matter with the FOPI experimental data

Based on the ultra-relativistic quantum molecular dynamics (UrQMD) transport model, combined with experimental data of directed flow, elliptic flow, and nuclear stopping power measured by FOPI in $\rm ^{197}Au+^{197}Au$ collisions at beam energies ($E_{lab}$) of 0.25 and 0.4 GeV/nucleon, the incompressibility of the nuclear equation of state $K_0$, the nucleon effective mass $m^*$, and the in-medium correction factor ($F$, with respect to free-space values) on the nucleon-nucleon elastic cross sections are studied by Bayesian analysis. It is found that both $m^*$ and $F$ can be tightly constrained with the uncertainty $\le$ 15\%, however, $K_0$ cannot be constrained tightly. We deduce $m^*/m_0 = 0.78^{+0.09}_{-0.10}$ and $F = 0.75^{+0.08}_{-0.07}$ with experimental data at $E_{lab}$ = 0.25 GeV/nucleon, and the obtained values increased to $m^*/m_0 = 0.88^{+0.03}_{-0.03}$ and $F = 0.88^{+0.06}_{-0.07}$ at $E_{lab}$ = 0.4 GeV/nucleon. The obtained results are further verified with rapidity-dependent flow data.

nucl-th↗

arXiv2025

Suspension Splitting and Cohomotopy Sets of Simply Connected $7$-manifolds

Let $M$ be a closed simply connected $7$-manifold. In this paper we establish homotopy decompositions of the reduced suspension space $ΣM$ into a wedge sum of simpler spaces when localized at a set of primes. These decompositions are applied to study the cohomotopy sets $π^k(M)$ and the $p$-local cohomotopy sets $π^4(M;\mathbb{Z}_{(p)})$.

math.AT↗

arXiv2025

Nucleon-$Δ$ elastic cross section in isospin-asymmetric nuclear medium with inclusion of scalar-isovector $δ$ meson field

The production, dynamic evolution, and decay of $Δ$ particles play a crucial role in understanding the properties of high baryon density nuclear matter in intermediate-energy heavy-ion collisions. In this work, the energy-, density-, and isospin-dependent nucleon-$Δ$ elastic cross section ($σ^{*}_{N Δ}$) is studied within the relativistic Boltzmann-Uehling-Uhlenbeck framework, in which the $δ$ meson field is further considered. The results show that the $δ$ and $ρ$ meson related exchange terms have a nonnegligible contribution to the $σ^{*}_{N Δ}$ compared to only considering the $ρ$ meson exchange terms, although there is a significant cancellation on the cross section among these meson exchange terms. In addition, due to the different effects of the medium correction on the effective mass of neutrons, protons, and differently charged $Δ$s, the individual $σ^{*}_{N Δ}$ exhibits an ordered isospin-asymmetry ($α$) dependence, and $σ^{*}_{nΔ}$ and $σ^{*}_{pΔ}$ have opposite $α$ dependencies. And the $α$ dependence of the ratio $R(α)=σ^{*}(α)/σ^{*}(α=0)$ for $nΔ$ reaction channels follow $nΔ^{++}>nΔ^{+}>nΔ^{0}>nΔ^{-}$, while for $pΔ$ it is $pΔ^{-}>pΔ^{0}>pΔ^{+}>pΔ^{++}$. Moreover, the results also indicate that the isospin effect on the $σ^{*}_{N Δ}$, which is dominantly caused by the isovector $ρ$ and $δ$ meson fields, is still pronounced at densities up to 3 times normal nuclear density. Finally, a parametrization of the energy-, density-, and isospin-dependent $NΔ$ elastic cross section is proposed based on the microscopic calculated results, and the in-medium $σ^{*}_{N Δ}$ in the energy range of $\sqrt{s}$=2.3$\sim$3.0 GeV can be well described.

nucl-th↗

arXiv2025

Atomic-Scale Heterogeneity of Hydrogen in Metal Hydrides Revealed by Electron Ptychography

Hydrogen plays critical roles in materials science, particularly for advancing technologies in hydrogen storage and phase manipulation, while also posing challenges like hydrogen embrittlement. Understanding its behavior, vital for improving material properties, requires precise determination of atomic-scale distribution-a persistent challenge due to hydrogen's weak electron scattering and high mobility, as well as the limitations of conventional transmission electron microscopy. We demonstrate that multislice electron ptychography (MEP) overcomes these constraints through three key advances: exceptional sensitivity for hydrogen occupancy, three-dimensional quantification, and picometer-level precision in atomic positioning. Experimentally, MEP resolves heterogeneous hydrogen distributions and quantifies hydrogen-induced lattice displacements with picometer precision in multi-principal-element alloy hydrides. This work demonstrates MEP as a transformative method for directly probing hydrogen atoms in solids, unlocking fundamental understanding of hydrogen's impact on material properties.

cond-mat.mtrl-sci↗

arXiv2025

Cohomotopy Sets of $(n-1)$-connected $(2n+2)$-manifolds for small $n$

Let $M$ be a closed orientable $(n-1)$-connected $(2n+2)$-manifold, $n\geq 2$. In this paper we combine the Postnikov tower of spheres and the homotopy decomposition of the reduced suspension space $ΣM$ to investigate the cohomotopy sets $π^\ast(M)$ for $n=2,3,4$, under the assumption that $M$ has $2$-torsion-free homology. All cohomotopy sets $π^i(M)$ of such manifolds $M$ are characterized except $π^4(M)$ for $n=3,4$.

math.AT↗

arXiv2025

Categorical action for finite classical groups and its applications: characteristic 0

In this paper, we construct a categorical double quantum Heisenberg action on the representation category of finite classical groups $\mathrm{O}_{2n+1}(q)$, $\mathrm{Sp}_{2n}(q)$ and $\mathrm{O}^{\pm}_{2n}(q)$ with $q$ odd. Over a field of characteristic zero or characteristic $\ell$ with $\ell\nmid q(q-1)$, we deduce a categorical action of a Kac-Moody algebra $\mathfrak{s}\mathfrak{l}'_{I_+}\oplus\mathfrak{s}\mathfrak{l}'_{I_-}$ on the representation category of finite classical groups. We show that the colored weight functions $\mathbb{O}^+(u)(\bullet)$, $\mathbb{O}^-(v)(\bullet)$ and uniform projection can distinguish all irreducible characters of finite classical groups. In particular, the colored weight functions are complete invariants of quadratic unipotent characters. We also show that using the theta correspondence and extra symmetries of categorical double quantum Heisenberg action, the Kac-Moody action on the Grothendieck group of the whole category can be determined explicitly.

math.RT↗

arXiv2025

Exploring Scene Affinity for Semi-Supervised LiDAR Semantic Segmentation

This paper explores scene affinity (AIScene), namely intra-scene consistency and inter-scene correlation, for semi-supervised LiDAR semantic segmentation in driving scenes. Adopting teacher-student training, AIScene employs a teacher network to generate pseudo-labeled scenes from unlabeled data, which then supervise the student network's learning. Unlike most methods that include all points in pseudo-labeled scenes for forward propagation but only pseudo-labeled points for backpropagation, AIScene removes points without pseudo-labels, ensuring consistency in both forward and backward propagation within the scene. This simple point erasure strategy effectively prevents unsupervised, semantically ambiguous points (excluded in backpropagation) from affecting the learning of pseudo-labeled points. Moreover, AIScene incorporates patch-based data augmentation, mixing multiple scenes at both scene and instance levels. Compared to existing augmentation techniques that typically perform scene-level mixing between two scenes, our method enhances the semantic diversity of labeled (or pseudo-labeled) scenes, thereby improving the semi-supervised performance of segmentation models. Experiments show that AIScene outperforms previous methods on two popular benchmarks across four settings, achieving notable improvements of 1.9% and 2.1% in the most challenging 1% labeled data.

cs.CV↗

arXiv2024

Predicting Artificial Neural Network Representations to Learn Recognition Model for Music Identification from Brain Recordings

Recent studies have demonstrated that the representations of artificial neural networks (ANNs) can exhibit notable similarities to cortical representations when subjected to identical auditory sensory inputs. In these studies, the ability to predict cortical representations is probed by regressing from ANN representations to cortical representations. Building upon this concept, our approach reverses the direction of prediction: we utilize ANN representations as a supervisory signal to train recognition models using noisy brain recordings obtained through non-invasive measurements. Specifically, we focus on constructing a recognition model for music identification, where electroencephalography (EEG) brain recordings collected during music listening serve as input. By training an EEG recognition model to predict ANN representations-representations associated with music identification-we observed a substantial improvement in classification accuracy. This study introduces a novel approach to developing recognition models for brain recordings in response to external auditory stimuli. It holds promise for advancing brain-computer interfaces (BCI), neural decoding techniques, and our understanding of music cognition. Furthermore, it provides new insights into the relationship between auditory brain activity and ANN representations.

q-bio.NC↗

arXiv2024

A Point-Neighborhood Learning Framework for Nasal Endoscope Image Segmentation

Lesion segmentation on nasal endoscopic images is challenging due to its complex lesion features. Fully-supervised deep learning methods achieve promising performance with pixel-level annotations but impose a significant annotation burden on experts. Although weakly supervised or semi-supervised methods can reduce the labelling burden, their performance is still limited. Some weakly semi-supervised methods employ a novel annotation strategy that labels weak single-point annotations for the entire training set while providing pixel-level annotations for a small subset of the data. However, the relevant weakly semi-supervised methods only mine the limited information of the point itself, while ignoring its label property and surrounding reliable information. This paper proposes a simple yet efficient weakly semi-supervised method called the Point-Neighborhood Learning (PNL) framework. PNL incorporates the surrounding area of the point, referred to as the point-neighborhood, into the learning process. In PNL, we propose a point-neighborhood supervision loss and a pseudo-label scoring mechanism to explicitly guide the model's training. Meanwhile, we proposed a more reliable data augmentation scheme. The proposed method significantly improves performance without increasing the parameters of the segmentation neural network. Extensive experiments on the NPC-LES dataset demonstrate that PNL outperforms existing methods by a significant margin. Additional validation on colonoscopic polyp segmentation datasets confirms the generalizability of the proposed PNL.

cs.CV↗