Workshop on Composite AI (CompAI)

Aim and Scope

Recent rapid growth of AI has shown the potential to revolutionize most of the everyday aspects of human lives. The field of AI is traditionally divided into a number of subfields such as Machine Learning, Knowledge and Reasoning, Planning and Scheduling, SAT solving, Computer Vision (and others) that are usually pursued individually. But the challenges of real-world applications are often too hard for a single AI approach. Hence there is a need for Composite AI, which integrates several different AI approaches that complement each other to solve the problem. Development of Composite AI system is still ad-hoc and premature.

The workshop aims at bringing together researchers and practitioners from different AI subfields to discuss challenges that they currently face and to initiate a discussion about the benefits and challenges of Composite AI.

Submission Instructions

All papers must be submitted in a PDF format and must conform to the (ECAI formatting). The extra page can only be used for acknowledgements and references. The reviewing process will be single-blind.

The submission is done via (OpenReview).

Authors of rejected papers from ECAI-24 wishing to submit their paper to CompAI can submit a request by July 11 through a dedicated form (which will beome available after July 1)

Schedule (October 20)

9:00–10:30 Welcome + Invited Talk

• Welcome (15min)
• Invited Talk - Giuseppe de Giacomo (title: From Infinite to Finite Traces and Back: Linear Temporal Logic in Sequential Decision Making )

10:30–11:00 Coffee break

11:00–12:30 Paper Session 1

• Uwe Köckemann: Goal-based Composition of Hybrid AI Systems
• Markus Iser: Automated Explanation Selection for Scientific Discovery
• Arthur Ledaguenel, Céline Hudelot and Mostepha Khouadjia: Improving Neural-based Classification with Logical Background Knowledge
• Caglar Demir, N'Dah Jean Kouagou, Arnab Sharma and Axel-Cyrille Ngonga Ngomo: Inference over Unseen Entities, Relations and Literals on Knowledge Graphs

12:30–14:00 Lunch break

14:00–15:30 Paper Session 2

• Slavomir Svancar, Lukas Chrpa, Filip Dvorak, Tomas Balyo: Cloud Kitchen: Using Planning-based Composite AI to Optimize Food Delivery Process
• Filippos Gouidis, Konstantinos Papoutsakis, Theodore Patkos, Antonis Argyros, Dimitris Plexousakis: Enabling Visual Intelligence by Leveraging Visual Object States in a Neurosymbolic Framework (presented by: Filippos Gouidis and Konstantinos Papoutsakis)
• Victor Sim, Wang Ngai Ng, Josiah Essiam: MP-End2End: Low Cost 3D analysis of Microplastics with DFF
• Kossi Amouzouvi, Yashrajsinh Chudasama, Jens Lehmann, Disha Purohit, Ariam Rivas, Bowen Song, Sahar Vahdati and Maria Esther-Vidal: Caving Knowledge Graphs in Latent Space with Similarity Group
• Karthika Vijayan and Shruti Dhavalikar: Combining Discriminative and Generative AI for Dedicated Conversational Assistants (virtual)

15:30–16:00 Coffee break

16:00–17:30 Panel Discussion

• Andrei Muratov (AWS GameTech, AWSI Media, Entertainment, Games & Sports), Mikoláš Janota (Czech Technical University in Prague), Julien Millet (Game Only cluster and United Bits Games)
• moderated by Filip Dvořák (Filuta AI)

Invited Talk Details

Speaker: Giuseppe de Giacomo

Title: From Infinite to Finite Traces and Back: Linear Temporal Logic in Sequential Decision Making

Abstract: Linear Temporal Logic (LTL) has a long history in CS and AI due to its ability to express sophisticated temporal properties over infinite traces. Recently, finite-trace variants of LTL, such as LTL on Finite Traces (LTLf) and Pure Past LTL (PPLTL), have gained popularity in AI, particularly in sequential decision-making tasks where an autonomous agent nominally loops through three finite phases: acquiring a goal, reasoning strategically to achieve it, and executing the resulting strategy (or plan). A key advantage of these finite-trace variants is their reducibility to equivalent regular automata, which can be determinized and transformed into two-player games on graphs. This gives them unprecedented computational effectiveness and scalability. Can these advantages be extended to infinite traces? In this talk, we provide a positive answer. By leveraging Manna and Pnueli’s safety-progress hierarchy for LTL, we introduce infinite-trace extensions of LTLf and PPLTL that retain the full expressive power of LTL, while preserving the crucial feature that the game arena for strategy extraction can still be derived from deterministic finite automata.

Accepted Papers

Slavomir Svancar, Lukas Chrpa, Filip Dvorak, Tomas Balyo:
Cloud Kitchen: Using Planning-based Composite AI to Optimize Food Delivery Process Paper

Uwe Köckemann:
Goal-based Composition of Hybrid AI Systems Paper

Markus Iser:
Automated Explanation Selection for Scientific Discovery Paper

Filippos Gouidis, Konstantinos Papoutsakis, Theodore Patkos, Antonis Argyros, Dimitris Plexousakis:
Enabling Visual Intelligence by Leveraging Visual Object States in a Neurosymbolic Framework Paper

Victor Sim, Wang Ngai Ng, Josiah Essiam:
MP-End2End: Low Cost 3D analysis of Microplastics with DFF Paper

Arthur Ledaguenel, Céline Hudelot and Mostepha Khouadjia:
Improving Neural-based Classification with Logical Background Knowledge Paper

Caglar Demir, N'Dah Jean Kouagou, Arnab Sharma and Axel-Cyrille Ngonga Ngomo:
Inference over Unseen Entities, Relations and Literals on Knowledge Graphs Paper

Kossi Amouzouvi, Yashrajsinh Chudasama, Jens Lehmann, Disha Purohit, Ariam Rivas, Bowen Song, Sahar Vahdati and Maria Esther-Vidal:
Caving Knowledge Graphs in Latent Space with Similarity Group Paper

Karthika Vijayan and Shruti Dhavalikar:
Combining Discriminative and Generative AI for Dedicated Conversational Assistants Paper