Efficient Constant Optimization for Symbolic Regression with GPU-Accelerated Tree-Based Genetic Programming
本文提出了一种基于GPU的批处理Levenberg-Marquardt求解器，用于树型遗传编程中的常数优化问题，显著提高了符号回归中数值系数优化的速度和质量。
0 citationsRead paper
Selected work
Representative Papers
M$^3$R-Bench: A Unified Benchmark for Evidence-Grounded Multimodal Metaphor Understanding
This work addresses the limitation of existing metaphor understanding benchmarks, which predominantly rely on isolated subtasks and lack evaluation of cross-modal target–source mappings grounded in joint visual and textual evidence. To bridge this gap, we introduce M³R-Bench, a unified multimodal benchmark grounded in Conceptual Metaphor Theory, comprising 1,000 human-verified image–text samples annotated across four layers: metaphor existence, mapping relations, sentiment polarity, and stepwise explanations. We further propose a novel three-stage evaluation framework—evidence identification, mapping construction, and sentiment inference—that reveals current models’ overreliance on textual cues and neglect of visual evidence. Building upon this, we develop M³R-Reasoner, which integrates curriculum-based reasoning supervision with task-aware reinforcement learning to guide multimodal large language models toward evidence–mapping consistent reasoning. Despite using only an 8B-parameter backbone, our model surpasses larger closed-source counterparts across all four metrics, outscoring GPT-5.5 by 28.45 and 30.11 points in visual evidence and sentiment plausibility, respectively, and exceeding Claude-Sonnet-4.6 by an average of 8.00 points.
0 citationsRead paper
Predictive Enhancement Calibration for Latent Breast MRI Virtual Contrast Enhancement
This work addresses radiomic distortion and spatial misalignment in synthetic contrast-enhanced breast MRI, which arise from generator intensity upper-bound constraints and independent intensity scaling between source and target images. To resolve these issues, the authors propose a Predictive Enhancement Calibration (PEC) method that establishes a case-adaptive shared coordinate system and predicts the missing enhancement upper bound directly from pre-contrast images during inference. PEC leverages a pretrained FLUX latent flow model for efficient conditional generation, incorporating parameter-efficient reference conditioning, target round-trip reconstruction, and a unified coordinate strategy within a single training framework. Evaluated on the MAMA100 cohort under a source-only setting, PEC significantly improves all eight assessment metrics, with the most pronounced gains observed in MSE and LPIPS.
0 citationsRead paper
Recent publications
Latest Papers
Efficient Constant Optimization for Symbolic Regression with GPU-Accelerated Tree-Based Genetic Programming
本文提出了一种基于GPU的批处理Levenberg-Marquardt求解器，用于树型遗传编程中的常数优化问题，显著提高了符号回归中数值系数优化的速度和质量。
0 citationsRead paper
M$^3$R-Bench: A Unified Benchmark for Evidence-Grounded Multimodal Metaphor Understanding
This work addresses the limitation of existing metaphor understanding benchmarks, which predominantly rely on isolated subtasks and lack evaluation of cross-modal target–source mappings grounded in joint visual and textual evidence. To bridge this gap, we introduce M³R-Bench, a unified multimodal benchmark grounded in Conceptual Metaphor Theory, comprising 1,000 human-verified image–text samples annotated across four layers: metaphor existence, mapping relations, sentiment polarity, and stepwise explanations. We further propose a novel three-stage evaluation framework—evidence identification, mapping construction, and sentiment inference—that reveals current models’ overreliance on textual cues and neglect of visual evidence. Building upon this, we develop M³R-Reasoner, which integrates curriculum-based reasoning supervision with task-aware reinforcement learning to guide multimodal large language models toward evidence–mapping consistent reasoning. Despite using only an 8B-parameter backbone, our model surpasses larger closed-source counterparts across all four metrics, outscoring GPT-5.5 by 28.45 and 30.11 points in visual evidence and sentiment plausibility, respectively, and exceeding Claude-Sonnet-4.6 by an average of 8.00 points.
0 citationsRead paper
Predictive Enhancement Calibration for Latent Breast MRI Virtual Contrast Enhancement
This work addresses radiomic distortion and spatial misalignment in synthetic contrast-enhanced breast MRI, which arise from generator intensity upper-bound constraints and independent intensity scaling between source and target images. To resolve these issues, the authors propose a Predictive Enhancement Calibration (PEC) method that establishes a case-adaptive shared coordinate system and predicts the missing enhancement upper bound directly from pre-contrast images during inference. PEC leverages a pretrained FLUX latent flow model for efficient conditional generation, incorporating parameter-efficient reference conditioning, target round-trip reconstruction, and a unified coordinate strategy within a single training framework. Evaluated on the MAMA100 cohort under a source-only setting, PEC significantly improves all eight assessment metrics, with the most pronounced gains observed in MSE and LPIPS.
0 citationsRead paper