School of Computing and Information Systems

Do LLMs Really Know What They Don't Know? Internal States Mainly Reflect Knowledge Recall Rather Than Truthfulness

Meeting room 4.4, Level 4. School of Computing and Information Systems 1,
Singapore Management University, 80 Stamford Road, Singapore 178902
Thursday, June 11, 2026, 11 – 11:30am

Recent work suggests that LLMs "know what they don't know”, positing that hallucinated and factually correct outputs arise from distinct internal processes and can therefore be distinguished using internal signals. However, hallucinations have multifaceted causes: beyond simple knowledge gaps, they can emerge from training incentives that encourage models to exploit statistical shortcuts or spurious associations learned during pretraining. In this paper, we argue that when LLMs rely on such learned associations to produce hallucinations, their internal processes are mechanistically similar to those of factual recall, as both stem from strong statistical correlations encoded in the model's parameters. To verify this, we propose a novel taxonomy categorizing hallucinations into Unassociated Hallucinations (UHs), where outputs lack parametric grounding, and Associated Hallucinations (AHs), which are driven by spurious associations. Through mechanistic analysis, we compare their computational processes and hidden-state geometries with factually correct outputs. Our results show that hidden states primarily reflect whether the model is recalling parametric knowledge rather than the truthfulness of the output itself. Consequently, AHs exhibit hidden-state geometries that largely overlap with factual outputs, rendering standard detection methods ineffective. In contrast, UHs exhibit distinctive, clustered representations that facilitate reliable detection.

This is a Pre-Conference talk for The 64th Annual Meeting of the Association for Computational Linguistics (ACL 2026).

Subtitle: Pre-Conference Talk by CHEANG Chi Seng
Contact: scisseminar@smu.edu.sg
Speaker Details: CHEANG Chi Seng, PhD Candidate, School of Computing and Information Systems, Singapore Management University, CHEANG Chi Seng is a first-year PhD student in Computer Science at Singapore Management University, supervised by Dr. DENG Yang. His research interests include natural language processing, trustworthy AI, and hallucination detection and mitigation in large language models.
RSVP:
Reserve a seat: forms.cloud.microsoft…
Type: Seminars & Workshops
Subject: Information Technology & Systems
Audience: Public
More info: computing.smu.edu.sg…

A Pruning-based Question-Answering for Interactive Video Search: A Simple Baseline

Meeting room 4.4, Level 4. School of Computing and Information Systems 1,
Singapore Management University, 80 Stamford Road, Singapore 178902
Thursday, June 11, 2026, 2 – 2:15pm

Video search systems frequently struggle with imprecise user queries, which retrieve excessively broad and noisy sets of candidate videos and reduce the effectiveness of ranking functions. This issue is compounded by the sheer volume of visually and semantically similar videos in large datasets, forcing users to browse extensively and increasing cognitive load. While interactive questioning offers a promising strategy to progressively narrow down candidate videos, dynamically generating these questions remains a significant challenge. Relying on large language models for this task often introduces high computational latency and the risk of hallucination. In this talk, I will present a simple yet highly effective alternative grounded in Shannon’s information theory. Rather than generating questions from scratch, our method mathematically selects the most discriminative concepts to compose targeted questions, allowing for rapid and reliable pruning of the search space. I will discuss the practical implementation mechanics of the system, specifically detailing dynamic search result updates and top-K rank list sampling. Furthermore, the presentation will address how our approach accounts for realistic system and user constraints, including imperfectly indexed video content and noisy or misleading user feedback. Finally, I will share our experimental findings, which demonstrate the strong retrieval performance of this method on the AVSD and TRECVid benchmarks.

This is a Pre-Conference talk for 16th ACM International Conference on Multimedia Retrieval (ICMR 2026).

Subtitle: Pre-Conference Talk by CHENG Yu Tong
Contact: scisseminar@smu.edu.sg
Speaker Details: CHENG Yu Tong, PhD Candidate, School of Computing and Information Systems, Singapore Management University, CHENG Yu Tong is a Ph.D. candidate supervised by Prof. NGO Chong Wah at Singapore Management University. His research primarily focuses on the field of multimedia retrieval, with a specialization in the known-item search task. He also has an extensive background in system design and competitive benchmarking, having participated in numerous iterations of the Video Browser Showdown (VBS). Notably, in the most recent VBS competition, his team's retrieval system, VIREO, achieved first place in the visual known-item search category. His current research investigates the development of smart conversational search systems, aiming to leverage interactive querying to streamline user interaction and improve retrieval accuracy in large-scale video datasets like TRECVid.
RSVP:
Reserve a seat: forms.cloud.microsoft…
Type: Seminars & Workshops
Subject: Information Technology & Systems
Audience: Public
More info: computing.smu.edu.sg…

Beyond Hard Constraints: Budget-Conditioned Reachability For Safe Offline Reinforcement Learning

Meeting room 5.1, Level 5. School of Computing and Information Systems 1,
Singapore Management University, 80 Stamford Road, Singapore 178902
Tuesday, June 16, 2026, 11:30 – 11:50am

Sequential decision-making using Markov Decision Process underpins many real-world applications. Both model-based and model-free methods have achieved strong results in these settings. However, real-world tasks must balance reward maximization with safety constraints, often conflicting objectives, that can lead to unstable min–max, adversarial optimization. A promising alternative is safety reachability analysis, which precomputes a forward-invariant safe state–action set, ensuring that an agent starting inside this set remains safe indefinitely. Yet, most reachability-based methods address only hard safety constraints, and little work extends reachability to cumulative cost constraints. To address this, first, we define a safety-conditioned reachability set that decouples reward maximization from cumulative safety cost constraints. Second, we show how this set enforces safety constraints without unstable min–max or Lagrangian optimization, yielding a novel offline safe RL algorithm that learns a safe policy from a fixed dataset without environment interaction. Finally, experiments on standard offline safe-RL benchmarks, and a real-world maritime navigation task demonstrate that our method matches or outperforms state-of-the-art baselines while maintaining safety.

This is a Pre-Conference talk for The 36th International Conference on Automated Planning and Scheduling (ICAPS 2026).

Subtitle: Pre-Conference Talk by Brahmanage Janaka Chathuranga THILAKARATHNA
Contact: scisseminar@smu.edu.sg
Speaker Details: Brahmanage Janaka Chathuranga THILAKARATHNA, PhD Candidate, School of Computing and Information Systems, Singapore Management University, Janaka Brahmanage is a fourth-year PhD candidate in Computer Science, conducting research under the guidance of Associate Prof. Akshat Kumar at the SMU School of Computing and Information Systems. His research focuses on safe reinforcement learning and multi-agent systems.
RSVP:
Reserve a seat: forms.cloud.microsoft…
Type: Seminars & Workshops
Subject: Information Technology & Systems
Audience: Public
More info: computing.smu.edu.sg…

Singapore HCI 2026 & HRI Workshop 2026

School of Computing and Information Systems 2, Level 3, Classroom 3-1, Singapore Management University, 90 Stamford Road, 178903
Thursday, June 18, 2026, 9am – 5pm

We cordially invite you to a one-day event—the Singapore HCI Meetup 2026 and HRI Workshop 2026—where HCI researchers across the island share ideas, celebrate achievements, and, most importantly, reconnect with old friends and meet new ones. This year's meetup will be hosted at the SMU campus on 18 June.

RSVP:
Reserve a seat: forms.office.com…
Type: Seminars & Workshops
Subject: Information Technology & Systems
Audience: Academic Community, Current Student, Public
More info: sghcimeetup26.notion.site

SMU Postgraduate Fast Track

Hybrid

Zoom details will be sent to you upon review of your profile.
Friday, June 19, 10am – Saturday, June 20, 2026, 2:30pm

Hourly Sessions: Friday, 19 June 2026 , 10:00am - 8:00pm, Saturday, 20 June 2026 , 10:00am - 2:30pm *Refer to the registration link for specific slots of each programme
RSVP: https://postgraduate2.smu.edu.sg/acton/media/24273/FT2
Event Details: We are conducting Fast-Track Admission for the abovementioned SMU Postgraduate Programmes!, You will be able to expedite your application and be guided through the process by the Admissions Team. Eligible candidates* would be shortlisted for an interview with the admission committee. Application fee will be waived for all fast-track attendees., If you're ready to apply, register today!, You are welcome to bring a friend with you. Simply forward them this link:, postgraduate2.smu.edu.sg…, *Shortlisted candidates will be contacted to confirm their registration status.
Type: Information Sessions
Subject: Accounting, Analytics for Business, Consumer & Social Insights, Economics, Finance & Financial Markets, Operations Management, Social Sciences & Humanities
Audience: Prospective Student, Public

Designing for “Just Enough”: How Elegance Drives Adoption and Innovation

Seminar Room 2-5, Level 2. SMU SCIS 2, Singapore 178903
Friday, June 19, 2026, 11:15am – 12:15pm

In an environment saturated with information, products, and interfaces, users increasingly expect designs that are both simple and effective. Elegant Design: A Designer’s Guide to Harnessing Aesthetics examines how the tension between simplicity and complexity can be addressed by conceptualizing elegance as a form of ideal complexity: sufficiently rich to convey meaning, yet clear enough for immediate comprehension. Drawing on research at the intersection of cognitive science, information theory, and art, Luca Landoli demonstrates that aesthetic pleasure is not merely an optional enhancement, but a critical factor guiding decision-making in design, innovation, and product development. Utilizing over 100 visual examples and case studies, the book presents eight actionable strategies: four for enriching designs that lack engagement, and four for simplifying solutions that have become overly complex. This presentation is particularly relevant for scholars and students in information systems, human-computer interaction, innovation, and new product development who seek to understand how aesthetics can improve usability, sense-making, and adoption. Participants will gain a practical framework for iteratively balancing simplification and complexification to achieve elegant products, services, and user experiences.

Subtitle: Research Seminar by Dr Luca Iandoli
Contact: scisseminars@smu.edu.sg
Speaker Details: Luca Iandoli, Distinguished Chair and, Dean of the Lesley H. and William L. Collins, College of Professional Studies, St. John’s University, Luca Iandoli is Dean of the Collins College of Professional Studies at St. John’s University. A scholar of collective intelligence, interaction design, and product aesthetics, his work explores how people and organizations make sense of complexity through design. A former Fulbright Visiting Scholar at the MIT Center for Collective Intelligence, he has also served as associate professor at the University of Naples Federico II and visiting professor at Stevens Institute of Technology.Luca is the co-author of Elegant Design: A Designer’s Guide to Harnessing Aesthetics, which bridges cognitive science, art, and information systems to show how “ideal complexity” and aesthetic pleasure can drive better products, interfaces, and decisions.
RSVP:
Reserve a seat: forms.office.com…
Type: Seminars & Workshops
Subject: Information Technology & Systems
Audience: Academic Community, Current Student, Public
More info: computing.smu.edu.sg…

SMU Doctor of Engineering (EngD) Virtual Information Session

Virtual via Zoom
Details will be provided via email upon registration.
Thursday, July 2, 2026, 7 – 8pm

RSVP:
Reserve a seat: scispg.smu.edu.sg…
Type: Information Sessions, Webinar & Online Learning
Subject: Analytics for Business, Consumer & Social Insights, Business, Information Technology & Systems, Leadership, Operations Management, Organisational Behaviour, Strategic Management
Audience: Professionals, Prospective Student

Self-optimizing Static Program Analysis

Seminar Room 4-2, Level 4. SMU SCIS 2, Singapore 178903
Friday, July 3, 2026, 10 – 11am

Software systems pervade our personal and professional lives, yet their insecurity threaten our society. To assure that software systems are dependable and secure, one must reason about their code. Static program analysis enables such reasoning. It can be applied to individual software components, and it can show not only the presence but also prove the absence of bugs and vulnerabilities. Yet, to be useful to software developers, static analyses must be adapted to the context in which they are used. Studies show that poorly adapted analyses slow down rather than assist development. They report large sets of false warnings that distract developers from actual bugs, which the analyses often miss. They often run so long that results are reported when they are already outdated. In this talk I report on my ERC Advanced Grant on Self-optimizing Static Program Analysis (SOSA). Its research hypothesis is that one can generate precise and efficient static analyses of software systems by making static analysis self-aware and self-optimizing. With SOSA, a static analysis conducts analyses and optimizations not just of programs but of itself. The first ingredient to self-optimization in SOSA is a novel intermediate representation (IR) - not of the analyzed program - but the analysis of that program. Optimization potential for the analysis can be pinpointed directly on the level of this “Abstract IR”, and analysis optimizations are easily expressed as rewrites on the level of this Abstract IR. This mechanism, akin to JIT optimizations in general-purpose virtual machines, allows for instantaneous, online optimizations. The second ingredient becomes important in situation where the analysis struggles analyzing a given program for reasons unknown. In this case, we use Sequential Model-based Optimizations (SMO) to learn configurations of the analysis that will likely do better at least in subsequent analysis runs. In this talk I will report on a first design of our Abstract IR “FAIR”, which supports different kinds of analyses of Java programs, and on initial experiments with SMO applied to the off-the-shelve static analyzer FlowDroid.

Subtitle: Research Seminar by Dr Eric Bodden
Contact: scisseminars@smu.edu.sg
Speaker Details: Eric Bodden, Professor, Heinz Nixdorf Institute, Paderborn University, Eric Bodden is a leading expert on secure software engineering, especially in building highly precise tools for automated program analysis. He is Professor for Secure Software Engineering at Paderborn University and director for Software Engineering and IT-Security at Fraunhofer IEM.
RSVP:
Reserve a seat: forms.office.com…
Type: Seminars & Workshops
Subject: Information Technology & Systems
Audience: Academic Community, Current Student, Public
More info: computing.smu.edu.sg…

The Voice of the Flow: A Graph-based Approach for Step-Wise Explanations of Constraint Satisfaction Problems

Meeting room 4.4, Level 4. School of Computing and Information Systems 1,
Singapore Management University, 80 Stamford Road, Singapore 178902
Friday, July 3, 2026, 11:30 – 11:50am

Stepwise explanations are an important tool for interpreting constraint programs. Existing methods rely heavily on repeated minimal unsatisfiable subsets (MUS) extraction or intensive SAT-based propagation, which are computationally prohibitive for large-scale problems with complex constraints. We propose a novel framework leveraging Multi-Valued Decision Diagrams (MDDs) to overcome these bottlenecks. By decomposing CSPs into smaller, graph-based subproblems compactly represented as MDDs, we enable highly efficient constraint propagation through network-flow reformulation. To further enhance efficiency, we propose a Divide-and-Conquer approach to improve the search for minimal explanation steps. Furthermore, we utilize the MDD’s network-flow structure to generate nested explanations that break down complex derivation steps into granular, arc-level details, revealing specifically how variable-value assignments become infeasible. In this way, our approach equips the explainer agent with a voice that provides more intuitive and granular stepwise explanations. Evaluated on Graph Coloring and Nurse Rostering benchmarks, our framework significantly reduces explanation generation time while maintaining interpretability and conciseness compared to state-of-the-art methods.

This is a Pre-Conference talk for The 32nd International Conference on Principles and Practice of Constraint Programming (CP 2026).

Subtitle: Pre-Conference Talk by NGUYEN Minh Anh
Contact: scisseminar@smu.edu.sg
Speaker Details: NGUYEN Minh Anh, PhD Candidate, School of Computing and Information Systems, Singapore Management University, NGUYEN Minh Anh is a third-year PhD candidate in Computer Science at Singapore Management University, supervised by Prof. LAU Hoong Chuin. His research focuses on explainable optimization, where he develops methods such as step-wise and counterfactual explanations to provide transparency and trust in complex combinatorial optimization problems. His work explores the intersection of XAI and operations research, with a focus on practical applications like drone delivery and nurse scheduling.
RSVP:
Reserve a seat: forms.cloud.microsoft…
Type: Seminars & Workshops
Subject: Information Technology & Systems
Audience: Public
More info: computing.smu.edu.sg…