When Successful Knowledge Graph Edits Displace Correct Answers: Rank-Level Locality beyond Parameter Support
研究解决知识图谱嵌入编辑导致正确答案位移的问题，通过引入不同范围的排名位移审计及条件推导方法来减少位移效应。
0 citationsRead paper
Selected work
Representative Papers
Attention Capture Is Not Detection: A Two-Stage Account of How Humans Miss Localized AI Image Edits
This study addresses the oversight of dissociated human attention and recognition mechanisms in existing AI image editing detection. We propose a two-stage cognitive model demonstrating that edited regions drive attentional capture while semantic plausibility determines judgment accuracy. As the first work to introduce pre-attentive and recognition distinctions into this domain, we construct a generative eye-movement prediction framework. Validated through eye-tracking and mixed-effects analyses, the significant dissociation between stages is confirmed. The model achieves attention prediction correlations of 0.77–0.82, and its missed-detection behavioral prediction performance (r=0.52) significantly outperforms linear baselines (r=0.48). These findings establish a novel paradigm for understanding detection blind spots in human-AI interaction, highlighting the critical role of cognitive separation in evaluating synthetic imagery.
0 citationsRead paper
HNDiff: Haze-Noise Diffusion for Image Dehazing
This work addresses the limitation of existing diffusion-based dehazing methods, which disregard the physical formation mechanism of haze and merely reconstruct images from Gaussian noise, thereby constraining restoration performance. To overcome this, the authors propose HNDiff, a novel framework that embeds the atmospheric scattering model as an inductive bias into the diffusion process: during the forward pass, haze and noise are jointly injected, while the reverse pass simultaneously performs dehazing and denoising to achieve physically consistent image recovery. The method introduces a haze-aware noise scheduler that adaptively modulates noise intensity according to haze density and further presents Latent HNDiff to enhance off-the-shelf dehazing networks. Extensive experiments demonstrate that HNDiff significantly boosts the performance of mainstream backbone architectures across multiple benchmarks, achieving state-of-the-art results.
0 citationsRead paper
MAPLE: Efficient and Diverse Multi-Alpha Generation for Portfolio Construction
This work addresses the limitation of conventional deep learning approaches for stock ranking, which typically produce a single alpha signal and lack explicit control over correlations among multiple alphas, resulting in insufficient portfolio diversity. The authors propose MAPLE, a novel framework that, within a single training run, jointly incorporates a unified capacity-scaled prediction head, an extreme-rank weighted listwise loss, and an explicit diversity regularizer to enable controllable generation of multiple alpha signals with desired correlation structures—all within a single model. Notably, MAPLE achieves this without increasing architectural complexity and is compatible with various backbone networks. Evaluated across four major equity markets in the U.S., China, and Japan, MAPLE significantly outperforms nine baselines, achieving up to 55× fewer parameters and 2.5× faster training while improving Sharpe ratios by 10–23% and Calmar ratios by 17–43%.
0 citationsRead paper
ConcoLixir: Reactive LLM Discovery Oracles for Python Concolic Testing
This work addresses the limitations of Python concolic testing, which often suffers from symbolic degradation due to library calls, intractable semantic operations, and stalled path exploration. The paper proposes the first integration of a large language model (LLM) as a lightweight, reactive oracle within the concolic execution loop. Without replacing the underlying symbolic solver, the LLM leverages execution feedback and path constraints to dynamically generate initial seeds, suggest concrete inputs upon solver failure, and guide targeted exploration toward uncovered code when coverage plateaus. This approach substantially enhances exploration efficacy, particularly across semantic barriers and library boundaries. Experimental results demonstrate average line coverage improvements of 8.6, 15.1, and 17.0 percentage points on synthetic benchmarks, real-world programs, and library-centric targets, respectively, with a total API cost of only \$1.63.
0 citationsRead paper
Recent publications
Latest Papers
When Successful Knowledge Graph Edits Displace Correct Answers: Rank-Level Locality beyond Parameter Support
研究解决知识图谱嵌入编辑导致正确答案位移的问题，通过引入不同范围的排名位移审计及条件推导方法来减少位移效应。
0 citationsRead paper
Attention Capture Is Not Detection: A Two-Stage Account of How Humans Miss Localized AI Image Edits
This study addresses the oversight of dissociated human attention and recognition mechanisms in existing AI image editing detection. We propose a two-stage cognitive model demonstrating that edited regions drive attentional capture while semantic plausibility determines judgment accuracy. As the first work to introduce pre-attentive and recognition distinctions into this domain, we construct a generative eye-movement prediction framework. Validated through eye-tracking and mixed-effects analyses, the significant dissociation between stages is confirmed. The model achieves attention prediction correlations of 0.77–0.82, and its missed-detection behavioral prediction performance (r=0.52) significantly outperforms linear baselines (r=0.48). These findings establish a novel paradigm for understanding detection blind spots in human-AI interaction, highlighting the critical role of cognitive separation in evaluating synthetic imagery.
0 citationsRead paper
HNDiff: Haze-Noise Diffusion for Image Dehazing
This work addresses the limitation of existing diffusion-based dehazing methods, which disregard the physical formation mechanism of haze and merely reconstruct images from Gaussian noise, thereby constraining restoration performance. To overcome this, the authors propose HNDiff, a novel framework that embeds the atmospheric scattering model as an inductive bias into the diffusion process: during the forward pass, haze and noise are jointly injected, while the reverse pass simultaneously performs dehazing and denoising to achieve physically consistent image recovery. The method introduces a haze-aware noise scheduler that adaptively modulates noise intensity according to haze density and further presents Latent HNDiff to enhance off-the-shelf dehazing networks. Extensive experiments demonstrate that HNDiff significantly boosts the performance of mainstream backbone architectures across multiple benchmarks, achieving state-of-the-art results.
0 citationsRead paper
MAPLE: Efficient and Diverse Multi-Alpha Generation for Portfolio Construction
This work addresses the limitation of conventional deep learning approaches for stock ranking, which typically produce a single alpha signal and lack explicit control over correlations among multiple alphas, resulting in insufficient portfolio diversity. The authors propose MAPLE, a novel framework that, within a single training run, jointly incorporates a unified capacity-scaled prediction head, an extreme-rank weighted listwise loss, and an explicit diversity regularizer to enable controllable generation of multiple alpha signals with desired correlation structures—all within a single model. Notably, MAPLE achieves this without increasing architectural complexity and is compatible with various backbone networks. Evaluated across four major equity markets in the U.S., China, and Japan, MAPLE significantly outperforms nine baselines, achieving up to 55× fewer parameters and 2.5× faster training while improving Sharpe ratios by 10–23% and Calmar ratios by 17–43%.
0 citationsRead paper
ConcoLixir: Reactive LLM Discovery Oracles for Python Concolic Testing
This work addresses the limitations of Python concolic testing, which often suffers from symbolic degradation due to library calls, intractable semantic operations, and stalled path exploration. The paper proposes the first integration of a large language model (LLM) as a lightweight, reactive oracle within the concolic execution loop. Without replacing the underlying symbolic solver, the LLM leverages execution feedback and path constraints to dynamically generate initial seeds, suggest concrete inputs upon solver failure, and guide targeted exploration toward uncovered code when coverage plateaus. This approach substantially enhances exploration efficacy, particularly across semantic barriers and library boundaries. Experimental results demonstrate average line coverage improvements of 8.6, 15.1, and 17.0 percentage points on synthetic benchmarks, real-world programs, and library-centric targets, respectively, with a total API cost of only \$1.63.
0 citationsRead paper