The 2026 Singapore Symposium on Natural Language Processing (SSNLP 2026)

Abstract: This presentation charts the evolution of Speech Language Models (SLMs) through the lens of our recent research. We describe the process of integrating speech encoders with text-based LLMs, emphasizing critical training strategies to prevent catastrophic forgetting. Next, we address the challenge of enabling LLMs to speak. To overcome the computational bottlenecks of conventional waveform-to-token autoregressive models, we introduce novel speech representation techniques that optimize sequence length and efficiency. The talk concludes with an approach to empowering SLMs with reasoning capabilities, advancing toward models that can "think while speaking."

Bio: Hung-yi Lee is a professor of the Department of Electrical Engineering at National Taiwan University (NTU), with a joint appointment at the Department of Computer Science & Information Engineering of the university. His recent research focuses on developing technology that can reduce the requirement of annotated data for speech processing (including voice conversion and speech recognition) and natural language processing (including abstractive summarization and question answering). He won Salesforce Research Deep Learning Grant in 2019, AWS ML Research Award in 2020, Outstanding Young Engineer Award from The Chinese Institute of Electrical Engineering in 2018, Young Scholar Innovation Award from Foundation for the Advancement of Outstanding Scholarship in 2019, Ta-You Wu Memorial Award from Ministry of Science and Technology of Taiwan in 2019, and The 59th Ten Outstanding Young Person Award in Science and Technology Research & Development of Taiwan. He is a Fellow of International Speech Communication Association (ISCA). He owns a YouTube channel teaching deep learning technology in Mandarin, which has more than 350,000 subscribers.

Bio: Seok Min leads the technical research team in Infocomm Media Development Authority (IMDA), driving initiatives in trust-related technologies, such as AI testing and digital watermarking. She began her career in cybersecurity research and has more recently focused on AI governance testing (AI Verify) and the evaluation of generative models (Project Moonshot).

Bio: Jian Gang is the Head of Applied Research in AI Singapore, and leads the teams that develop the SEA-LION family of models, the SEA-HELM evaluation suite and leaderboard, as well as various research publication efforts.

Abstract: Despite the superior performance demonstrated by Transformer-based LLMs across numerous applications involving natural languages, their high computational cost, energy consumption, and limited accessibility underscore the need for efficient, interpretable, and adaptable small language models (SLMs). This talk highlights methods to develop economical and interpretable SLMs that rival their larger counterparts in performance without significant computational requirements. Our research emphasizes three key dimensions: economical resource usage, adaptability to diverse and low-resource tasks, and enhanced interpretability. Techniques like competitive knowledge distillation, leveraging student-teacher dynamics, and activation sparsity in manifold-preserving transformers demonstrate significant efficiency gains without compromising performance. We formulate novel decomposer components for LLMs for modularizing problem decomposition and solution generation, allowing smaller models to excel in complex reasoning tasks. We also propose innovative prompt construction and alignment strategies that boost in-context knowledge adaptation in low-resource settings for SLMs. Our findings demonstrate that SLMs can achieve scalability, interpretability, and adaptability, paving the way for broader and sustainable AI accessibility.

Bio: Tanmoy Chakraborty is a Rajiv Khemani Young Faculty Chair Professor in AI and an Associate Professor in the Dept. of Electrical Engineering and the School of AI at IIT Delhi. He leads the Laboratory for Computational Social Systems (LCS2), a research group that primarily focuses on building economical, adaptable and interpretable language models. He served as the DAAD visiting professor at MPI Saarbrucken, PECFAR visiting professor at TU Munich and Humboldt visiting professor at TU Darmstadt. Tanmoy has received numerous recognitions, including the Indian National Academic of Science Young Associate, Ramanujan Fellowship, ACL '23 Outstanding Paper Award, IJCAI'23 AI for Social Good Award, and several faculty awards from industries like Microsoft, IBM, Google, LinkedIn, JP Morgan, and Adobe. He has authored two textbooks -- "Social Network Analysis" and "Introduction to Large Language Models". Tanmoy earned his PhD from IIT Kharagpur in 2015 as a Google PhD Scholar. He served as the PC Chair of EMNLP'25 and the WebConf'26 (Web4Good Track), and local organizing chair of AACL'25. More details may be found at tanmoychak.com.

Abstract: Large language models can generate fluent texts on demand, yet fluency is not the same as accessibility: the same text may be trivial for one reader and incomprehensible for another. This keynote argues that text simplification and complexity control should be grounded in standardized proficiency definitions rather than ad-hoc heuristics. I introduce a CEFR-grounded framework that treats "simplicity" as a target proficiency level, enabling controlled adaptation of syntax and vocabulary while preserving meaning. Central to the framework is CEFR-SP, a corpus of 17k English sentences with expert CEFR annotations, which supports proficiency-aware evaluation and learning signals. Building on CEFR-SP, I present reinforcement-learning approaches that steer LLM rewriting toward a specified CEFR level via lexical- and sentence-level rewards, and I discuss error analyses that reveal where current LLM-based simplifiers still fail. I conclude with open challenges in document-level simplification, including disentangling linguistic difficulty from topical expertise and discourse structure, toward inclusive NLP.

Bio: Yuki Arase is a professor at the School of Computing, Institute of Science Tokyo (aka Tokyo Institute of Technology), Japan. After obtaining her PhD in Information Science from Osaka University in 2010, she worked for Microsoft Research Asia, where she started NLP research that continues to captivate her to this day. Her research interests focus on paraphrasing and NLP technology for language education and healthcare.

Abstract: Modern large language models (LLMs) are increasingly expected not only to generate responses but also to evaluate their own outputs. This talk presents a unified perspective on transforming LLMs into reliable evaluators and demonstrates how such judging capability can, in turn, strengthen their reasoning performance. We begin by exploring the evaluative knowledge inherently embedded within large models and introduce methods to elicit and apply this capacity in tasks such as translation evaluation and solution verification. We then illustrate how these evaluative abilities can be integrated into reinforcement learning pipelines, allowing models to critique their own chain‑of‑thought and guide optimization toward deeper and more consistent reasoning. Collectively, these insights outline a practical framework: elicit the model's latent evaluative skills, calibrate its reasoning, structure clear assessment criteria, rigorously verify outcomes, and ultimately leverage these self‑generated judgments to train models that reason, and learn, to be better judges of their own performance.

Bio: Derek F. Wong is a Full Professor at the University of Macau, where he leads the Natural Language Processing and Chinese–Portuguese Machine Translation Laboratory (NLP2CT Lab). He serves on the boards and committees of CIPS, CCF, and AFNLP, and holds editorial roles with IEEE/ACM TASLP, ACM TALLIP, TACL, and the ACL Rolling Review. His research has earned multiple honors, including the Macao Science and Technology Awards (2012, 2022), the FST Research Excellence Award, and both the Outstanding Academic Staff Incentive (2022) and Teaching Excellence Award (2024) from UM. He has also contributed to major NLP conferences such as ACL, NeurIPS, ICML, IJCAI, AAAI, EMNLP, NAACL, COLING, AACL, and IJCNLP in various program leadership roles.