We are pleased to announce that two papers from researchers of AC have been accepted to three upcoming conferences in early 2026: AAAI, and EACL.
Hongkuan Zhou travelled to Singapore to present [1] at AAAI.
This work addresses Open-domain Visual Entity Recognition, a task that links images to real-world concepts. Unlike standard classification with fixed labels, it operates under open-set conditions where most entities are unseen during training, leading to high visual ambiguity and limited supervision. To address this, Hongkuan and his collaborators propose KnowCoL (Knowledge-guided Contrastive Learning), a dual-encoder framework that projects images and entity text descriptions into a shared semantic embedding space and enriches that space with structured knowledge (e.g., type hierarchies and relations from Wikidata) to improve zero-shot recognition and disambiguation.
You will find Yuqicheng Zhu in Rabat, as he got his last work accepted in EACL, in collaboration with Hongkuan.
The authors focus on Knowledge Graph-based Retrieval-Augmented Generation (KG-RAG). Many existing benchmarks use questions that simple retrieval can answer, which makes it hard to tell if models are actually reasoning or just finding facts. Yuqicheng and Hongkuan, among others, present a new benchmarking BRINK (Benchmark for Reasoning under Incomplete Knowledge). Their standardized evaluation reveals that current KG-RAG methods struggle with reasoning under incomplete knowledge, often defaulting to internal memorization rather than effectively leveraging graph structures and exhibiting varying degrees of generalization depending on their design.
[1] H. Zhou, L. Halilaj, S. Monka, S. Schmid, Y. Zhu, J. Wu, N. Nazer, S. Staab. Seeing and Knowing in the Wild: Open-domain Visual Entity Recognition with Large-scale Knowledge Graphs via Contrastive Learning. In: 40th AAAI Conference on Artificial Intelligence, AAAI 2026, Singapore, January 20 to January 27, 2026.
[2] D. Zhou, Y. Zhu, X. Wang, H. Zhou, Y. He, J. Chen, S. Staab, E. Kharlamov. What Breaks Knowledge Graph based RAG? Empirical Insights into Reasoning under Incomplete Knowledge. In: The 19th Conference of the European Chapter of the Association for Computational Linguistics (EACL-2026). Rabat, Morocco, March 24-29, 2026.
