arXiv subjects

Peifeng Liu

Publications and source records attributed to Peifeng Liu.

8 recordsLinked to original sources

arXiv2026

Evidence-Guided Detection, Localization and Explanation for Text-Centric Image Forensics

The rapid progress of AIGC has made text-centric image manipulation increasingly accessible, creating new forensic challenges that require not only authenticity detection but also spatial grounding and evidence-based explanation. This paper presents our solution to the GenText-Forensics Challenge at ACM Multimedia 2026. We propose an evidence-guided detector-localizer-reasoner system, where an image-level detector provides a global authenticity prior, a dedicated localizer extracts tampered regions as spatial grounding evidence, and an MLLM-based reasoner generates structured forensic reports grounded in this expert forensic evidence. These modules are connected through a cascaded evidence flow: the detector gates the subsequent localization and prompting process, the localizer converts tamper responses into grounding boxes, and the reasoner is trained to synthesize the detector decision and localized evidence into the final report. As a key part of our method, we introduce iterative difficulty-aware mining to improve localization quality and apply report-mask consistency post-processing to align report grounding with predicted masks. On the official hidden test set, our system achieves a final score of 0.638 and ranks second in the challenge, validating the effectiveness of the proposed evidence-guided system. The code is available at https://github.com/peifengLiu42/ACMMM26-evidence-guided-detector-localizer-reasoner-system.

cs.CV↗

arXiv2026

Active Evidence-Seeking and Diagnostic Reasoning in Large Language Models for Clinical Decision Support

Large language models perform well on static medical examinations, yet clinical diagnosis often requires iterative evidence gathering under uncertainty. Building on prior interactive evaluation efforts, we introduce an OSCE-inspired standardized patient simulator and a controlled, reproducible benchmark for active diagnostic inquiry. Across 468 cases and 15 models in our protocol, we observe that multi-turn evidence seeking reduces diagnostic accuracy by 12.75% and lowers supporting-evidence quality by 24.36% relative to full-context evaluation; error analyses associate these drops with premature diagnostic closure and inefficient questioning. Together, these results suggest that static full-context benchmarks may overestimate performance in interactive evidence-seeking settings, motivating complementary interactive assessment for safer clinical decision support.

cs.AI↗

arXiv2026

A multi-stage soft computing framework for complex disease modelling and decision support: A liver cirrhosis case study

Liver cirrhosis is a major global health problem causing millions of deaths annually, and timely detection with aggressive treatment can significantly improve patients' quality of life. Modelling complex diseases from biomedical data is computationally challenging due to high dimensionality, strong feature correlations, noise, and limited labelled samples. Conventional Machine Learning (ML) pipelines often struggle with robustness, interpretability, and generalisation under such conditions. In this study, we propose an ML-driven multi-stage decision framework for complex disease modelling and therapeutic exploration. The framework integrates single-cell transcriptomic profiling, high-dimensional network-based feature stabilisation, multi-model learning, deep representation construction, and post-hoc decision support. Specifically, single-cell sequencing data were analysed to identify key cellular subpopulations, followed by high-dimensional weighted gene co-expression network analysis (hdWGCNA) to stabilise gene modules under sparsity and noise. To enhance non-linear feature interaction modelling, tabular molecular features were restructured into two-dimensional disease maps and analysed using a CNN. Finally, molecular docking was incorporated as a decision-support module to evaluate candidate therapeutic compounds. Using liver cirrhosis as a representative case, the framework identified a disease-associated endothelial subpopulation and extracted seven robust signature genes (HSPB1, GADD45A, CLDN5, ATP1B3, C1QBP, ENPP2, and PARL). The CNN-based representation learning module outperformed conventional pipelines in classification. The framework is disease-agnostic and readily extends to other omics-driven biomedical applications involving uncertainty, heterogeneity, and limited samples.

q-bio.OT↗

arXiv2018

Examination of the observability of a chiral magnetically-driven charge-separation difference in collisions of the $\mathrm{^{96}_{44}Ru +\, ^{96}_{44}Ru}$ and $\mathrm{^{96}_{40}Zr +\, ^{96}_{40}Zr}$ isobars at energies available at RHIC

Anomalous Viscous Fluid Dynamics (AVFD) model calculations for $\mathrm{^{96}_{44}Ru +\, ^{96}_{44}Ru}$ and $\mathrm{^{96}_{40}Zr +\, ^{96}_{40}Zr}$ collisions ($\sqrt{s_{\rm NN}} = 200$ GeV) are used in concert with a charge-sensitive correlator, to test its ability to detect and characterize the charge separation difference expected from the Chiral Magnetic Effect (CME) in these isobaric collisions. The tests indicate a larger charge separation for $\mathrm{^{96}_{44}Ru +\, ^{96}_{44}Ru}$ than for $\mathrm{^{96}_{40}Zr +\, ^{96}_{40}Zr}$ collisions, and a discernible CME-driven difference of $\sim 10$\% in the presence of realistic non-CME backgrounds. They also indicate a strategy for evaluating the relative influence of the background correlations, present for each isobar. These results suggest that charge separation measurements for these isobaric species could serve to further constrain unambiguous identification and characterization of the CME in upcoming measurements at RHIC.

nucl-ex↗

arXiv2018

System-size dependence of the viscous attenuation of anisotropic flow in p+Pb and Pb+Pb collisions at LHC energies

The elliptic and triangular flow coefficients ($\mathrm{v_n, \, n=2,3}$) measured in Pb+Pb ($\sqrt{s_{_{\rm NN}}} = 2.76$ TeV) and p+Pb ($\sqrt{s_{_{\rm NN}}} = 5.02$ TeV) collisions, are studied as a function of initial-state eccentricity ($\varepsilon_n$), and dimensionless size characterized by the cube root of the mid-rapidity charged hadron multiplicity density $\mathrm{\left< N_{ch} \right>^{1/3}}$. The results indicate that the influence of eccentricity ($\mathrm{v_n} \propto \varepsilon_n$) observed for large $\mathrm{\left< N_{ch} \right>}$, is superseded by the effects of viscous attenuation for small $\mathrm{\left< N_{ch} \right>}$, irrespective of the colliding species. Strikingly similar acoustic scaling patterns of exponential viscous modulation, with a damping rate proportional to $\mathrm{n^2}$ and inversely proportional to the dimensionless size, are observed for the eccentricity-scaled coefficients for the two sets of colliding species. The resulting scaling parameters suggest that, contrary to current predilections, the patterns of viscous attenuation, as well as the specific shear viscosity $\left<\fracη{s}(\text{T})\right>$ for the matter created in p+Pb and Pb+Pb collisions, are comparable.

nucl-ex↗

arXiv2018

Acoustic scaling of linear and mode-coupled anisotropic flow; implications for precision extraction of the specific shear viscosity

The $\mathrm{n^{th}}$-order linear flow coefficients $\mathrm{v^L_n \, (n=2,3,4,5)}$, and the corresponding nonlinear mode-coupled ($\mathrm{mc}$) coefficients $\mathrm{v^{mc}_{4,(2,2)}}$, $\mathrm{v^{mc}_{5,(2,3)}}$, $\mathrm{v^{mc}_{6,(3,3)}}$ and $\mathrm{v^{mc}_{6,(2,2,2)}}$, are studied for Pb+Pb collisions at $\sqrt{s_{_{\rm NN}}} = 2.76$ TeV. Both sets of coefficients indicate a common acoustic scaling pattern of exponential viscous modulation, with a rate proportional to the square of the harmonic numbers and the mean transverse momenta (respectively), and inversely proportional to the cube root of the charged particle multiplicity ($\mathrm{(N_{ch})^{1/3}}$), that characterizes the dimensionless size of the systems produced in the collisions. These patterns and their associated scaling parameters, provide new stringent constraints for eccentricity independent estimates of the specific shear viscosity ($η/s$) and the viscous correction to the thermal distribution function for the matter produced in the collisions. They also give crucial constraints for extraction of the initial-state eccentricity spectrum.

nucl-ex↗

arXiv2016

Finite-Size Scaling of Non-Gaussian Fluctuations Near the QCD Critical Point

An effective Finite-Size Scaling (FSS) of moment products from recent STAR measurements of the variance $σ$, skewness $S$ and kurtosis $κ$ of net-proton multiplicity distributions, are reported for a broad range of collision centralities in Au+Au ($\sqrt{s_{NN}}= 7.7 - 200$ GeV) collisions. The products $Sσ$ and $κσ^2 $, which are directly related to the hgher-order baryon number susceptibility ratios $χ^{(3)}_B/χ^{(2)}_B$ and $χ^{(4)}_B/χ^{(2)}_B$, show scaling patterns consistent with earlier indications for a second order phase transition at a critical end point (CEP) in the plane of temperature vs. baryon chemical potential ($T,μ_B$) of the QCD phase diagram. The resulting scaling functions validate the earlier estimates of $T^{\text{cep}} \sim 165$ MeV and $μ_B^{\text{cep}} \sim 95$ MeV for the location of the CEP, and the critical exponents used to assign its 3D Ising model universality class.

nucl-ex↗

arXiv2016

Scaling properties of the mean multiplicity and pseudorapidity density in $e^{-}+e^{+}$, $e^{\pm}$+p, p($\bar{\mathrm{p}}$)+p, p+A and A+A(B) collisions

The pseudorapidity density (dN/deta) for p+p, p+A and A+A(B) collisions, and the mean multiplicity for ee, ep, and p+p collisions, are studied for an inclusive range of beam energies (Root_s). Characteristic scaling patterns are observed for both dN/deta and , consistent with a thermal particle production mechanism for the bulk of the soft particles produced in all of these systems. They also validate an essential role for quark participants in these collisions. The scaled values for dN/deta and are observed to factorize into contributions which depend on log(Root_s) and the number of nucleon or quark participant pairs (Npp). Quantification of these contributions give expressions which serve to systematize dN/deta and measurements spanning nearly four orders of magnitude in Root_s, and to predict their values as a function of Root_s and Npp.

nucl-ex↗