Recent closed-source multimodal systems have made great advances, but their hidden language for understanding the world remains opaque because of their black-box architectures. In this paper, we use the systems' preference bias to study their hidden language: During the process of compressing the input images (typically containing multiple concepts) into texts and then reconstructing them into images, the systems' inherent preference bias introduces specific shifts in the outputs, disrupting the original input concept co-occurrence. We employ the multi-round "telephone game" to strategically leverage this bias. By observing the co-occurrence frequencies of concepts in telephone games, we quantitatively investigate the concept connection strength in the understanding of multimodal systems, i.e., "hidden language." We also contribute Telescope, a dataset of 10,000+ concept pairs, as the database of our telephone game framework. Our telephone game is test-time scalable: By iteratively running telephone games, we can construct a global map of concept connections in multimodal systems' understanding. Here we can identify preference bias inherited from training, assess generalization capability advancement, and discover more stable pathways for fragile concept connections. Furthermore, we use Reasoning-LLMs to uncover unexpected concept relationships that transcend textual and visual similarities, inferring how multimodal systems understand and simulate the world. This study offers a new perspective on the hidden language of multimodal systems and lays the foundation for future research on the interpretability and controllability of multimodal systems.

In recent years, the rapid increase in academic publications across various fields has posed severe challenges for academic paper analysis: scientists struggle to timely and comprehensively track the latest research findings and methodologies. Key concept extraction has proven to be an effective analytical paradigm, and its automation has been achieved with the widespread application of language models in industrial and scientific domains. However, existing paper databases are mostly limited to similarity matching and basic classification of key concepts, failing to deeply explore the relational networks between concepts. This paper is based on the OpenAlex opensource knowledge graph. By analyzing nearly 8,000 open-source paper data from Novosibirsk State University, we discovered a strong correlation between the distribution patterns of paper key concept paths and both innovation points and rare paths. We propose a prompt engineering-based key concept path analysis method. This method leverages small language models to achieve precise key concept extraction and innovation point identification, and constructs an agent based on a knowledge graph constraint mechanism to enhance analysis accuracy. Through fine-tuning of the Qwen and DeepSeek models, we achieved significant improvements in accuracy, with the models publicly available on the Hugging Face platform.

One of the most intriguing features of language is its constant change, with ongoing shifts in how meaning is expressed. Despite decades of research, the factors that determine how and why meanings evolve remain only partly understood. Colexification -- the phenomenon of expressing multiple distinct concepts using the same word form -- serves as a valuable window onto the dynamics of meaning change across languages. Here, we apply phylogenetic comparative models to dictionary data from three language families, Austronesian, Indo-European, and Uralic, in order to shed light on the evolutionary dynamics underlying the colexification of concept pairs. We assess the effects of three predictors: associativity, borrowability, and usage frequency. Our results show that more closely related concept pairs are colexified across a larger portion of the family tree and exhibit slower rates of change. In contrast, concept pairs that are more frequent and more prone to borrowing tend to change more rapidly and are less often colexified. We also find considerable differences between the language families under study, suggesting that areal and cultural factors may play a role.

Educational Knowledge Graphs (EduKGs) organize various learning entities and their relationships to support structured and adaptive learning. Prerequisite relationships (PRs) are critical in EduKGs for defining the logical order in which concepts should be learned. However, the current EduKG in the MOOC platform CourseMapper lacks explicit PR links, and manually annotating them is time-consuming and inconsistent. To address this, we propose an unsupervised method for automatically inferring concept PRs without relying on labeled data. We define ten criteria based on document-based, Wikipedia hyperlink-based, graph-based, and text-based features, and combine them using a voting algorithm to robustly capture PRs in educational content. Experiments on benchmark datasets show that our approach achieves higher precision than existing methods while maintaining scalability and adaptability, thus providing reliable support for sequence-aware learning in CourseMapper.

Clinical notes contain rich clinical narratives but their unstructured format poses challenges for large-scale analysis. Standardized terminologies such as SNOMED CT improve interoperability, yet understanding how concepts relate through co-occurrence and semantic similarity remains underexplored. In this study, we leverage the MIMIC-IV database to investigate the relationship between SNOMED CT concept co-occurrence patterns and embedding-based semantic similarity. Using Normalized Pointwise Mutual Information (NPMI) and pretrained embeddings (e.g., ClinicalBERT, BioBERT), we examine whether frequently co-occurring concepts are also semantically close, whether embeddings can suggest missing concepts, and how these relationships evolve temporally and across specialties. Our analyses reveal that while co-occurrence and semantic similarity are weakly correlated, embeddings capture clinically meaningful associations not always reflected in documentation frequency. Embedding-based suggestions frequently matched concepts later documented, supporting their utility for augmenting clinical annotations. Clustering of concept embeddings yielded coherent clinical themes (symptoms, labs, diagnoses, cardiovascular conditions) that map to patient phenotypes and care patterns. Finally, co-occurrence patterns linked to outcomes such as mortality and readmission demonstrate the practical utility of this approach. Collectively, our findings highlight the complementary value of co-occurrence statistics and semantic embeddings in improving documentation completeness, uncovering latent clinical relationships, and informing decision support and phenotyping applications.

CLIP and large multimodal models (LMMs) have better accuracy on examples involving concepts that are highly represented in the training data. However, the role of concept combinations in the training data on compositional generalization is largely unclear -- for instance, how does accuracy vary when a common object appears in an uncommon pairing with another object? In this paper, we investigate how word co-occurrence statistics in the pretraining dataset (a proxy for co-occurrence of visual concepts) impacts CLIP/LMM performance. To disentangle the effects of word co-occurrence frequencies from single-word frequencies, we measure co-occurrence with pointwise mutual information (PMI), which normalizes the joint probability of two words co-occurring by the probability of co-occurring independently. Using synthetically generated images with a variety of concept pairs, we show a strong correlation between PMI in the CLIP pretraining data and zero-shot accuracy in CLIP models trained on LAION-400M (r=0.97 and 14% accuracy gap between images in the top and bottom 5% of PMI values), demonstrating that even accuracy on common concepts is affected by the combination of concepts in the image. Leveraging this finding, we reproduce this effect in natural images by editing them to contain pairs with varying PMI, resulting in a correlation of r=0.75. Finally, we demonstrate that this behavior in CLIP transfers to LMMs built on top of CLIP (r=0.70 for TextVQA, r=0.62 for VQAv2). Our findings highlight the need for algorithms and architectures that improve compositional generalization in multimodal models without scaling the training data combinatorially. Our code is available at https://github.com/helenqu/multimodal-pretraining-pmi.

While large language models like BERT demonstrate strong empirical performance on semantic tasks, whether this reflects true conceptual competence or surface-level statistical association remains unclear. I investigate whether BERT encodes abstract relational schemata by examining internal representations of concept pairs across taxonomic, mereological, and functional relations. I compare BERT's relational classification performance with representational structure in [CLS] token embeddings. Results reveal that pretrained BERT enables high classification accuracy, indicating latent relational signals. However, concept pairs organize by relation type in high-dimensional embedding space only after fine-tuning on supervised relation classification tasks. This indicates relational schemata are not emergent from pretraining alone but can be induced via task scaffolding. These findings demonstrate that behavioral performance does not necessarily imply structured conceptual understanding, though models can acquire inductive biases for grounded relational abstraction through appropriate training.

The exceptional performance of Large Language Models (LLMs) often comes with the unintended propagation of social biases embedded in their training data. While existing benchmarks evaluate overt bias through direct term associations between bias concept terms and demographic terms, LLMs have become increasingly adept at avoiding biased responses, creating an illusion of neutrality. However, biases persist in subtler, contextually hidden forms that traditional benchmarks fail to capture. We introduce the Hidden Bias Benchmark (HBB), a novel dataset designed to assess hidden bias that bias concepts are hidden within naturalistic, subtly framed contexts in real-world scenarios. We analyze six state-of-the-art LLMs, revealing that while models reduce bias in response to overt bias, they continue to reinforce biases in nuanced settings. Data, code, and results are available at https://github.com/JP-25/Hidden-Bias-Benchmark.

The use of computational ontologies is well-established in the field of Medical Informatics. The topic of Social Determinants of Health (SDoH) has also received extensive attention. Work at the intersection of ontologies and SDoH has been published. However, a standardized framework for Social Determinants of Education (SDoEd) is lacking. In this paper, we are closing the gap by introducing an SDoEd ontology for creating a precise conceptualization of the interplay between life circumstances of students and their possible educational achievements. The ontology was developed utilizing suggestions from ChatGPT-3.5-010422 and validated using peer-reviewed research articles. The first version of developed ontology was evaluated by human experts in the field of education and validated using standard ontology evaluation software. This version of the SDoEd ontology contains 231 domain concepts, 10 object properties, and 24 data properties

In an effort to create a more efficient AI system, we introduce a new multi-modality learning framework that leverages a modality-agnostic concept space possessing abstract knowledge and a set of modality-specific projection models tailored to process distinct modality inputs and map them onto the concept space. Decoupled from specific modalities and their associated projection models, the concept space focuses on learning abstract knowledge that is universally applicable across modalities. Subsequently, the knowledge embedded into the concept space streamlines the learning processes of modality-specific projection models. We evaluate our framework on two popular tasks: Image-Text Matching and Visual Question Answering. Our framework achieves performance on par with benchmark models while demonstrating more efficient learning curves.