Lightnews — Scholar-powered news

UniStuttgartAI @unistuttgartai.bsky.social · Aug 12

🚀 New accepted at ECAI 2025!

Researchers from AC at the KI Institute will present “Full-History Graphs with Edge-Type Decoupled Networks for Temporal Reasoning” - a new temporal graph neural network design that decouples spatial & temporal edges for better long-range reasoning.

#ECAI2025

MultiADS: Defect-aware Supervision for Multi-type Anomaly Detection and Segmentation in Zero-Shot Learning

UniStuttgartAI @unistuttgartai.bsky.social · Aug 7

Sadikaj, Y., Zhou, H., Halilaj, L., Schmid, S., Staab, S., & Plant, C. (2025). MultiADS: Defect-aware Supervision for Multi-type Anomaly Detection and Segmentation in Zero-Shot Learning. ICCV 2025. arxiv.org/abs/2504.06740

Precise optical inspection in industrial applications is crucial for minimizing scrap rates and reducing the associated costs. Besides merely detecting if a product is anomalous or not, it is crucial ...

AccessGuru: Leveraging LLMs to Detect and Correct Web Accessibility Violations in HTML Code

UniStuttgartAI @unistuttgartai.bsky.social · Aug 7

Fathallah, N., Hernández, D., & Staab, S. (2025). AccessGuru: Leveraging LLMs to detect and correct web accessibility violations in HTML code. ASSETS 2025. arxiv.org/abs/2507.19549

The vast majority of Web pages fail to comply with established Web accessibility guidelines, excluding a range of users with diverse abilities from interacting with their content. Making Web pages acc...

Two Papers Addressing Real-World Problems Through AI to be presented at ICCV and ASSETS | News | Jul 29, 2025 | Institute for Artificial Intelligence | University of Stuttgart

UniStuttgartAI @unistuttgartai.bsky.social · Aug 7

🔗 See more details about the papers in our latest post: www.ki.uni-stuttgart.de/institute/ne...

www.ki.uni-stuttgart.de

UniStuttgartAI @unistuttgartai.bsky.social · Aug 7

🎉 Researchers from AC @ KI Institute will present two AI papers tackling real-world challenges:

🌐 "Making the Web More Inclusive: Enter AccessGuru" (ASSETS 2025).
🏭 "MultiADS: Defect-aware Supervision for Multi-type Anomaly Detection and Segmentation in Zero-Shot Learning" (ICCV 2025).

Reposted by UniStuttgartAI

Daniel Hernández @daniel.mstdn.degu.cl.ap.brid.gy · Aug 2

Yuqicheng Zhu (@yuqichengzhu.bsky.social) presented our paper, Predicate-Conditional Conformalized Answer Sets for Knowledge Graph Embeddings, at #acl2025. This paper studies a key question for the reliability of applications using knowledge graph embeddings […]

[Original post on mstdn.degu.cl]

Yuqicheng is standing in front of his poster. The poster is on a panel with the number 299.

3 3

UniStuttgartAI @unistuttgartai.bsky.social · May 28

In particular, Dr. Stolle will introduce the model of “continuous safety engineering for high-risk AI systems,” which explicitly models uncertainties and confidences and propagates these during design time and operation time.

UniStuttgartAI @unistuttgartai.bsky.social · May 28

Dr. Stolle will outline two perspectives on safety engineering for AI-enabled systems—a safety-centric view and an AI-centric view—and then present his team's approach of combining the best of these two worlds.

Conformalized Answer Set Prediction for Knowledge Graph Embedding

UniStuttgartAI @unistuttgartai.bsky.social · May 28

We are honored to host 𝐃𝐫. 𝐑𝐞𝐢𝐧𝐡𝐚𝐫𝐝 𝐒𝐭𝐨𝐥𝐥𝐞 for his talk, “𝐸𝑛𝑔𝑖𝑛𝑒𝑒𝑟𝑖𝑛𝑔 𝑆𝑎𝑓𝑒 𝑆𝑦𝑠𝑡𝑒𝑚𝑠 𝑤𝑖𝑡ℎ 𝐴𝐼”, on 𝐉𝐮𝐧𝐞 𝟓 𝐚𝐭 𝟏𝟓:𝟒𝟓 in 𝐑𝐨𝐨𝐦 𝐔𝟑𝟐.𝟏𝟎𝟏, 𝐔𝐧𝐢𝐯𝐞𝐫𝐬𝐢𝐭ä𝐭𝐬𝐬𝐭𝐫𝐚ß𝐞 𝟑𝟐. Students, staff, and all interested guests are warmly invited to attend!

#AI #Safety

1 2 3

UniStuttgartAI @unistuttgartai.bsky.social · Apr 24

📍 NAACL 2025

Conformalized Answer Set Prediction for Knowledge Graph Embedding: arxiv.org/abs/2408.08248

Knowledge graph embeddings (KGE) apply machine learning methods on knowledge graphs (KGs) to provide non-classical reasoning capabilities based on similarities and analogies. The learned KG embeddings...

DAGE: DAG Query Answering via Relational Combinator with Logical Constraints

UniStuttgartAI @unistuttgartai.bsky.social · Apr 24

📍 WWW 2025

DAGE: DAG Query Answering via Relational Combinator with Logical Constraints: arxiv.org/abs/2410.22105

MaSH: Maximal Separating Poincaré Hyperplanes for Hierarchical Imbalanced Learning: www.cs.emory.edu/~jyang71/fil...

Predicting answers to queries over knowledge graphs is called a complex reasoning task because answering a query requires subdividing it into subqueries. Existing query embedding methods use this deco...

From Tokens to Lattices: Emergent Lattice Structures in Language Models

UniStuttgartAI @unistuttgartai.bsky.social · Apr 24

📍 ICLR 2025

From Tokens to Lattices: Emergent Lattice Structures in Language Models: arxiv.org/abs/2504.08778

Pretrained masked language models (MLMs) have demonstrated an impressive capability to comprehend and encode conceptual knowledge, revealing a lattice structure among concepts. This raises a critical ...