Autoregressive Mosaics: Probing 2D Spatial Reasoning in Text-Only Language Models
研究通过Autoregressive Mosaics基准测试探究文本模型的2D空间推理能力，区分了模型的空间布局能力和代码生成能力。
0 citationsRead paper
Selected work
Representative Papers
Aging of Prompt Engineering Techniques Across LLM Versions
研究评估了五种提示工程技术在六个指令调优模型上的效果变化，发现不同模型家族对提示技术的响应差异，建议根据具体模型调整提示策略。
0 citationsRead paper
Lost in Aggregation: How Benchmarks Overlook Irreplaceable Model Strengths
论文探讨了通过关注数据集特定峰值性能来评估模型不可替代性的问题，提出了一种新的评价框架以补充传统聚合指标的不足。
0 citationsRead paper
CytoBERT: A Foundation Model for Cytometry Data
This study addresses the challenge of cross-study model transfer in flow cytometry caused by data heterogeneity and the lack of standardized markers. To overcome this, we present the first open-source single-cell flow foundation model designed for variable marker panels. Leveraging self-supervised pretraining and marker standardization on a corpus of 50 million cells, the model learns transferable features across heterogeneous panels, effectively breaking data migration bottlenecks. Experimental results demonstrate robust cross-dataset generalization in sample-level classification tasks. By enabling reliable feature extraction despite panel variability, this work establishes a methodological foundation for scalable, universal flow cytometry analysis and facilitates broader application of foundation models in immunophenotyping research.
0 citationsRead paper
Meandering Photo Lines: Fluid Multiscale Exploration of Photographic Archives
This study addresses the common limitation of existing photo browsing interfaces, which typically decouple spatial and temporal dimensions, thereby obscuring the spatiotemporal narratives embedded in personal photographic archives. To overcome this, the authors propose a river-inspired “photo stream” visualization that integrates spatial, temporal, and thematic information into a continuous, explorable meandering path. The system combines two complementary views: a grid-based clustered overview for macro-level context and a curved-detail view for micro-level exploration. Employing collapsible interactions and animated transitions, it enables seamless multiscale navigation. Evaluations on two large-scale datasets—one comprising 100,000 photographs from the Jewish diaspora and another with 20,000 personal mobile photos—demonstrate that the approach successfully unifies analytical scrutiny with curiosity-driven, immersive exploration.
0 citationsRead paper
Recent publications
Latest Papers
Autoregressive Mosaics: Probing 2D Spatial Reasoning in Text-Only Language Models
研究通过Autoregressive Mosaics基准测试探究文本模型的2D空间推理能力，区分了模型的空间布局能力和代码生成能力。
0 citationsRead paper
Aging of Prompt Engineering Techniques Across LLM Versions
研究评估了五种提示工程技术在六个指令调优模型上的效果变化，发现不同模型家族对提示技术的响应差异，建议根据具体模型调整提示策略。
0 citationsRead paper
Lost in Aggregation: How Benchmarks Overlook Irreplaceable Model Strengths
论文探讨了通过关注数据集特定峰值性能来评估模型不可替代性的问题，提出了一种新的评价框架以补充传统聚合指标的不足。
0 citationsRead paper
CytoBERT: A Foundation Model for Cytometry Data
This study addresses the challenge of cross-study model transfer in flow cytometry caused by data heterogeneity and the lack of standardized markers. To overcome this, we present the first open-source single-cell flow foundation model designed for variable marker panels. Leveraging self-supervised pretraining and marker standardization on a corpus of 50 million cells, the model learns transferable features across heterogeneous panels, effectively breaking data migration bottlenecks. Experimental results demonstrate robust cross-dataset generalization in sample-level classification tasks. By enabling reliable feature extraction despite panel variability, this work establishes a methodological foundation for scalable, universal flow cytometry analysis and facilitates broader application of foundation models in immunophenotyping research.
0 citationsRead paper
Meandering Photo Lines: Fluid Multiscale Exploration of Photographic Archives
This study addresses the common limitation of existing photo browsing interfaces, which typically decouple spatial and temporal dimensions, thereby obscuring the spatiotemporal narratives embedded in personal photographic archives. To overcome this, the authors propose a river-inspired “photo stream” visualization that integrates spatial, temporal, and thematic information into a continuous, explorable meandering path. The system combines two complementary views: a grid-based clustered overview for macro-level context and a curved-detail view for micro-level exploration. Employing collapsible interactions and animated transitions, it enables seamless multiscale navigation. Evaluations on two large-scale datasets—one comprising 100,000 photographs from the Jewish diaspora and another with 20,000 personal mobile photos—demonstrate that the approach successfully unifies analytical scrutiny with curiosity-driven, immersive exploration.
0 citationsRead paper