Tanawan Premsri

Tanawan Premsri

He/Him

Exploratory AI Researcher

Sitemap

A list of all the posts and pages found on the site. For you robots out there, there is an XML version available for digesting as well.

Pages

Posts

Future Blog Post

less than 1 minute read

Published:

This post will show up by default. To disable scheduling of future posts, edit config.yml and set future: false.

Blog Post number 4

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 3

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 2

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

Blog Post number 1

less than 1 minute read

Published:

This is a sample blog post. Lorem ipsum I can’t remember the rest of lorem ipsum and don’t have an internet connection right now. Testing testing testing this blog post. Blog posts are cool.

portfolio

publications

Gluecons: A generic benchmark for learning under constraints.

Published in Proceedings of the AAAI Conference on Artificial Intelligence (AAAI 2023), 2023

This work introduces a benchmark of nine NLP and vision tasks to systematically evaluate deep learning models that integrate external knowledge as constraints. The benchmark enables richer comparison through extended evaluation criteria and highlights research challenges in constraint-based learning.

Recommended citation: H.R. Faghihi, A. Nafar, C. Zheng, R. Mirzaee, Y. Zhang, A. Uszok, A. Wan, T. Premsri, P. Kordjamshidi, et al. "Gluecons: A generic benchmark for learning under constraints. " Proceedings of the AAAI Conference on Artificial Intelligence (AAAI). 2023.
Download Paper

A Survey on Compositional Learning of AI Models: Theoretical and Experimetnal Practices

Published in Transactions on Machine Learning Research (TMLR), 2024

We provide a comprehensive survey of compositional learning in AI, linking cognitive theories to computational models and evaluating how language and vision systems—including modern LLMs—handle compositional reasoning. Our analysis clarifies current capabilities, limitations, and future research challenges.

Recommended citation: S. Sinha, T. Premsri, P. Kordjamshidi. "A Survey on Compositional Learning of AI Models: Theoretical and Experimental Practices." Transactions on Machine Learning Research (TMLR). 2024.
Download Paper

Neuro-symbolic Training for Reasoning over Spatial Language

Published in Findings of the Association for Computational Linguistics: NAACL 2025 (NAACL 2025 Finding), 2025

We address LLM limitations in complex spatial reasoning by introducing a neuro-symbolic training method that guides models using spatial logical rules. The technique improves generalization and yields strong gains on spatial QA benchmarks, particularly for multi-step reasoning.

Recommended citation: T. Premsri, P. Kordjamshidi. "Neuro-symbolic Training for Reasoning over Spatial Language." In Findings of the Association for Computational Linguistics: NAACL 2025. 2025.
Download Paper

Toward a Clearer Characterization of Neuro-Symbolic Frameworks: A Brief Comparative Analysis.

Published in NeSy 2025, Proceedings of Machine Learning Research (PMLR) 2025., 2025

We examine the technical foundations of Neurosymbolic (NeSy) modeling, identifying how existing frameworks integrate symbolic representations with neural architectures and where they fall short—particularly in usability and generality. By comparing three generic NeSy systems, we outline key challenges and directions needed to advance problem-solving capabilities in future NeSy frameworks.

Recommended citation: S. Sinha, T. Premsri, P. Kordjamshidi. "Toward a Clearer Characterization of Neuro-Symbolic Frameworks: A Brief Comparative Analysis. " Proceedings of Machine Learning Research (PMLR),. 2025.
Download Paper

FoR-SALE: Frame of Reference-guided Spatial Adjustment in LLM-based Diffusion Editing

Published in Preprint, Arxiv, 2025

We introduce FoR-SALE, a diffusion editing framework that incorporates spatial Frame of Reference reasoning to improve text-to-image generation. By detecting and correcting FoR misalignment between language and vision, FoR-SALE enhances spatial accuracy and improves state-of-the-art model performance by up to 5.3\%.

Recommended citation: T. Premsri, P. Kordjamshidi. "FoR-SALE: Frame of Reference-guided Spatial Adjustment in LLM-based Diffusion Editing" Preprint. 2025.
Download Paper

FoREST: Frame of Reference Evaluation in Spatial Reasoning Tasks

Published in Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing (EMNLP 2025 Main), 2025

We introduce FoREST, a benchmark designed to evaluate Frame-of-Reference (FoR) comprehension in spatial reasoning tasks. Results show major gaps in FoR understanding across LLMs and text-to-image systems, and our Spatial-Guided prompting method improves their spatial reasoning performance.🏆 SAC Highlight Award.

Recommended citation: T. Premsri, P. Kordjamshidi. "FoREST: Frame of Reference Evaluation in Spatial Reasoning Tasks." In Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.
Download Paper

talks

teaching

Teaching experience 1

Undergraduate course, University 1, Department, 2014

This is a description of a teaching experience. You can use markdown like any other post.

Teaching experience 2

Workshop, University 1, Department, 2015

This is a description of a teaching experience. You can use markdown like any other post.