Deep neural networks are known to develop similar representations for semantically related data, even when they belong to different domains, such as an image and its description, or the same text in different languages. We present a method for quantitatively investigating this phenomenon by measuring the relative information content of the representations of semantically related data and probing how it is encoded into multiple tokens of large language models (LLMs) and vision transformers. Looking first at how LLMs process pairs of translated sentences, we identify inner ``semantic'' layers containing the most language-transferable information. We find moreover that, on these layers, a larger LLM (DeepSeek-V3) extracts significantly more general information than a smaller one (Llama3.1-8B). Semantic information of English text is spread across many tokens and it is characterized by long-distance correlations between tokens and by a causal left-to-right (i.e., past-future) asymmetry. We also identify layers encoding semantic information within visual transformers. We show that caption representations in the semantic layers of LLMs predict visual representations of the corresponding images. We observe significant and model-dependent information asymmetries between image and text representations.

Abstract--We present an online semantic object mapping system for a quadruped robot operating in real indoor environments, turning sensor detections into named objects in a global map. During a run, the mapper integrates range geometry with camera detections, merges co-located detections within a frame, and associates repeated detections into persistent object instances across frames. Objects remain in the map when they are out of view, and repeated sightings update the same instance rather than creating duplicates. The output is a compact object layer that can be queried (class, pose, and confidence), is integrated with the occupancy map and readable by a planner . In on-robot tests, the layer remained stable across viewpoint changes.

We introduce SpreadPy as a Python library for simulating spreading activation in cognitive single-layer and multiplex networks. Our tool is designed to perform numerical simulations testing structure-function relationships in cognitive processes. By comparing simulation results with grounded theories in knowledge modelling, SpreadPy enables systematic investigations of how activation dynamics reflect cognitive, psychological and clinical phenomena. We demonstrate the library's utility through three case studies: (1) Spreading activation on associative knowledge networks distinguishes students with high versus low math anxiety, revealing anxiety-related structural differences in conceptual organization; (2) Simulations of a creativity task show that activation trajectories vary with task difficulty, exposing how cognitive load modulates lexical access; (3) In individuals with aphasia, simulated activation patterns on lexical networks correlate with empirical error types (semantic vs. phonological) during picture-naming tasks, linking network structure to clinical impairments. SpreadPy's flexible framework allows researchers to model these processes using empirically derived or theoretical networks, providing mechanistic insights into individual differences and cognitive impairments. The library is openly available, supporting reproducible research in psychology, neuroscience, and education research.

Understanding the meaning of data is crucial for performing data analysis, yet for the users to gain insight into the content and structure of their database, a tedious data exploration process is often required [2, 16]. A common industry practice taken on by specialists such as Knowledge Engineers is to explicitly construct an intermediate layer between the database and the user -- a semantic layer -- abstracting away certain details of the database schema in favor of clearer data semantics [3, 10]. In the era of Large Language Models (LLMs), industry practitioners and researchers attempt to circumvent this costly process using LLM-powered Natural Language Interfaces [4, 6, 12, 18, 19, 22]. The promise of such Text-to-SQL solutions is to allow users without technical expertise to seamlessly interact with databases. For example, a new company employee could effectively issue queries in natural language without programming expertise or even explicit knowledge of the database structure, e.g., knowing the names of entities or properties, the exact location of data sources, etc.

LiDAR bundle adjustment (BA) is an effective approach to reduce the drifts in pose estimation from the front-end. Existing works on LiDAR BA usually rely on predefined geometric features for landmark representation. This reliance restricts generalizability, as the system will inevitably deteriorate in environments where these specific features are absent. To address this issue, we propose SGBA, a LiDAR BA scheme that models the environment as a semantic Gaussian mixture model (GMM) without predefined feature types. This approach encodes both geometric and semantic information, offering a comprehensive and general representation adaptable to various environments. Additionally, to limit computational complexity while ensuring generalizability, we propose an adaptive semantic selection framework that selects the most informative semantic clusters for optimization by evaluating the condition number of the cost function. Lastly, we introduce a probabilistic feature association scheme that considers the entire probability density of assignments, which can manage uncertainties in measurement and initial pose estimation. We have conducted various experiments and the results demonstrate that SGBA can achieve accurate and robust pose refinement even in challenging scenarios with low-quality initial pose estimation and limited geometric features. We plan to open-source the work for the benefit of the community https://github.com/Ji1Xinyu/SGBA.

Back in 2018, I had the privilege of keynoting at one of Semantic Web Company's events in Vienna, as well as attending the full event. It was a great opportunity to immerse myself in the Central European perspective on the utility of Linked Open Data standards and how those standards were being applied. I got to meet innovators at enterprises making good use of Linked Open Vocabularies with the help of SWC's PoolParty semantics platform, Ontotext's GraphDB and Semiodesk's semantic graph development acceleration software, for example. There is so much that is impactful and powerful going on at these kinds of semantic technology events. So many people in the audience grasp the importance of a semantic layer to findable, accessible, interoperable, and reusable (FAIR) data, regardless of its origin and its original form–whether structured data, or document and multimedia content.

A semantic layer is a business-friendly representation of data, allowing for explanation of complex business logic in simpler terms. In Business Intelligence (BI), it has been called the metadata layer, semantic model, business view, or BI model. When the semantic layer was first introduced to BI tools 30 years ago, it defined table joins, metric aggregation, user-friendly names and more, allowing BI end-users to simply drag-and-drop fields like Product Name and Sales onto a report. Yes, "no-code" BI has been around for at least 30 years. This allowed early data teams to start thinking more strategically about where to put business logic, but also opened up a lot of complex issues.

Different BI tools offer different approaches to building dashboards. On one end of the spectrum, you have tools that prescribe having one query per chart and on the other end you have tools that espouse implementing a complex semantic layer. I believe there's a middle path that lies between both extremes, and I call it the dataset-centric approach. In the dataset-centric approach, the tool is connected to individual datasets that are expected to contain all of the metrics and dimensions for a given subject area. In this post, I'll describe the strengths and tradeoffs for each of the approaches and make the case for the dataset-centric approach as the ideal one for fast-moving data teams.

Every time we deal with an image transformer network what we end up doing is the exact same thing: finetuning a pretrained backbone of the encoder part. This is the traditional approach, not just for the semantic segmentation task. However, not taking into account the semantic information of the image to solve this task may not be the optimal method especially if we are talking about semantic segmentation. The authors of this paper have addressed the above problem by proposing a new simple and effective framework that can incorporate the semantic information of the image into a pretrained hierarchical transformer-based backbone with the help of a semantic attention operation. The authors provide empirical evidence by integrating SeMask into Swin-Transformer.

Semantic layers can be the key to a powerful, widely used business intelligence solution and an analytical waste of money--but what does a good semantic layer for self-service BI look like today? Tune into this webinar to learn how you can architect your systems and train your employees to encourage data sharing, improve data quality, and enhance your ability to produce average, powerful analytics across your organization.