Bytes

12 NUS Computing Recipients Honoured at 2026 National Awards

12 NUS Computing staff and faculty members have received the 2026 National Awards, for their dedication and service to NUS and Singapore.

Established in 1962, the National Awards recognise individuals for their contributions and service to the nation.

Public Administration Medal (Bronze)

• Associate Professor Soo Yuen Jien, Department of Computer Science, and formerly Director, Centre for Development of Teaching & Learning
• Professor Lee Mong Li, Director, NUS Centre for Trusted Internet and Community

Efficiency Medal

• Cheng Chek Keong, Laboratory Technologist

Long Service Medal

• Senior Lecturer Chia Wai Kit, Henry, Department of Computer Science
• Senior Lecturer Low Kok Lim, Department of Computer Science
• Professor Abhik Roychoudhury, Department of Computer Science
• Associate Professor Chan Chee Yong, Department of Computer Science
• Ang Hwee Ying Agnes, Senior Associate Director, Graduate Studies
• Teo Pei Pei, Management Support Officer, Office Operations and Events
• Associate Professor Tan Swee Lin, Sharon, Department of Information Systems and Analytics
• Professor Chan Mun Choon, Department of Computer Science 
• Professor Tulika Mitra, Dean of School of Computing
• Associate Professor Soo Yuen Jien, Department of Computer Science, and formerly Director, Centre for Development of Teaching & Learning

These honours recognise the contributions of colleagues across our academic, research, technical, and administrative community. Their work over the years has supported our students, strengthened the School, and contributed to the wider NUS community.

Congratulations to all our recipients on this well-deserved recognition, and thank you for your years of service to NUS Computing and Singapore.

A paper by Bai Yunpeng, a PhD candidate at NUS Computing, has been published in Nature Human Behaviour, one of the leading journals in the behavioural and cognitive sciences.

“Hierarchical resource rationality explains human reading behaviour” addresses a long-standing gap in reading research. Existing theories have explained either how the eyes move during reading or how comprehension is built, but not the link between the two. Bai Yunpeng, working with Associate Professor Antti Oulasvirta of Aalto University and Professor Zhao Shengdong of City University of Hong Kong, proposes that both are governed by a single principle: the brain selects eye movements to maximise expected understanding while minimising cognitive and time costs.

This principle operates across three nested timescales. Fixation decisions support word recognition. Sentence-level processing guides when the eyes skip ahead or move back to reread. Text-level comprehension goals shape memory and rereading over longer stretches of text. The team built this framework into a computational model, which reproduced a wide range of established findings in reading research, from effects at the level of individual words to outcomes as broad as overall comprehension.

The model offers a general account of how the brain coordinates perception, memory and action under limited resources, a principle that extends beyond reading. Potential applications include tools that support people with reading difficulties, educational technology that adapts to individual learners, and embodied AI systems designed to interpret visual information the way humans do.

The team plans to extend the model beyond text to more visual forms of reading, integrate it with embodied AI systems, and build simulators to support future studies of reading behaviour.

Assistant Professor Bian Yatao and NUS Presidential Young Professor Zhang Jiaheng have been awarded the National Research Foundation (NRF) Fellowship (Class of 2026), one of Singapore’s most competitive programmes for early-career researchers.

NRF Fellowship supports outstanding researchers to conduct five years of independent work in Singapore, with the aim of strengthening the country’s scientific and technological capabilities. Fellows are selected for the quality and ambition of their research, and their potential to deliver breakthrough outcomes.

Their projects: 

• Bian Yatao – Towards the Era of Reasoning AI for Science: Reasoning-Empowered Foundation Models for Atomic Systems
• Zhang Jiaheng – Efficient Zero-Knowledge Proofs for Large-Scale Computation and Applications

The awards reflect the range of research at NUS Computing, spanning AI for scientific discovery and the foundations of secure computation.

Sean Jean (Year 4, Computer Science) has won the Singapore National Finals of the International Quant Championship (IQC), organised by WorldQuant. He will represent Singapore at the Global Finals on 6-7 October, competing against champions from 12 other countries. 

Sean entered IQC this year after coming across it through a high school friend and later through Associate Professor Steven Halim’s CS3233 Competitive Programming course, which introduced him to quantitative finance. He submitted two alphas – trading signals or strategies used to predict market movements and generate returns – for the first stage during a break from his summer internship, without expecting to advance, and eventually qualified for stage two ranked outside the top 10,000. 

With 10 days left before the deadline, he built an automation pipeline that generates trading alphas using large language models and reinforcement learning, drawing on coursework from NUS’ AI-focused modules – CS3263, CS3264, and CS4248. His ranking rose to first place by the final presentation round, where placement was based on both score and presentation. 

IQC is a team-based competition, with teams of up to four members. Sean competed alone, under the name Sharpe Mind.

He is now refining his pipeline ahead of the Global Finals, drawing on conversations with WorldQuant researchers and other teams from the presentation round.

Three faculty members from NUS Computing have been awarded the prestigious National Research Foundation Investigatorship (NRF-I) under its 11th call, recognising their leadership and potential to drive high-impact research in artificial intelligence.

The NRF Investigatorship is one of Singapore’s most competitive research awards. It supports a small number of outstanding Principal Investigators to pursue ambitious, high-risk research that can lead to significant scientific breakthroughs. The programme is part of NRF’s broader mission to strengthen Singapore’s research ecosystem and develop world-class scientific talent.

The awardees are:

Abhik Roychoudhury - Agentic AI based Software of the future: from Scale to Trust

Angela Yao -  From Perception to Understanding: Contextual AI for Anticipatory and Proactive Assistance

Harold Soh - Learning Persistent Agent-Centric Representations for Embodied AI

Each award supports a five-year programme, enabling investigators to push the boundaries of their fields and contribute to Singapore’s position as a global hub for innovation.

Together, these projects reflect the strength and ambition of research at NUS Computing.

A PhD student from NUS Computing has picked up a Best Presentation Award at one of mobile computing's most competitive early-career showcases.

Bangjie Sun was recognised at the Rising Stars Forum, held alongside ACM MobiSys 2026 in Cambridge, UK. The forum invites PhD students and early-career postdoctoral researchers worldwide to present their work to a panel of established academics, with only a subset of submitted abstracts selected to present. Sun's talk, "Trustworthy Provenance for Physical and Digital Artifacts with Commodity Mobile Devices," was one of just three presentations honoured with the award, alongside researchers from Columbia University and Stanford University.

His research builds systems, deployable on everyday devices like smartphones, that can verify where a physical product or piece of digital content actually came from and whether it has been tampered with. Rather than requiring specialised lab equipment, the aim is to make this kind of verification something anyone can do with the device already in their pocket.

"I am honoured to receive this recognition at the ACM MobiSys 2026. Presenting alongside outstanding early-career researchers and engaging with leaders in the mobile computing community strengthened my commitment to pursuing an academic career in this field. The experience motivates me to further develop trustworthy, mobile-first systems that help people verify the origin and integrity of physical and digital artifacts. I look forward to building on this work and contributing to a future in which trustworthy technology has meaningful real-world impact,” said Bangjie.

A paper co-authored by Jialin Li, Sung Kah Kay Assistant Professor and alumnus Dehui Wei from the Department of Computer Science won the Outstanding Paper Award at NSDI (USENIX Symposium on Networked Systems Design and Implementation) '26, one of the top venues in computer networking and systems research.

The paper, "HyperEdge: An Edge CDN Infrastructure for Cost Efficient Video Streaming," was developed in collaboration with ByteDance. It addresses a costly problem faced by ByteDance's short video services: as ByteDance's platforms grew, delivering video through conventional content delivery networks became increasingly expensive. HyperEdge solves this by a new hybrid video deliver architecture that combines a network of under-utilised edge devices, a centralised tracker system, and an auxiliary CDN. HyperEdge also introduces a novel multi-path transmission protocol that keeps streaming reliable even when individual edge devices are unpredictable.

HyperEdge has been in production at ByteDance to support its popular short video platforms. It is now managing over a hundred thousand edge devices, serving about a hundred million users daily, and saving hundreds of millions of dollars in delivery costs annually.

HyperEdge was presented at NSDI '26's Distributed Data Systems session in May.

Assistant Professor Flavien Solt and collaborators at Microsoft Research has received a Distinguished Paper award at the 2026 IEEE Symposium on Security and Privacy, one of the field's most selective venues.

The paper, "Enter, Exit, Page Fault, Leak: Testing Isolation Boundaries for Microarchitectural Leaks", examines a basic promise of cloud computing: unrelated customers may share a physical server without gaining access to each other's data. Virtualisation is designed to keep their workloads separate. Yet features that make modern processors faster, including branch predictors, can leave traces of activity across this boundary. An attacker may exploit those traces to infer protected data, a threat security researchers have studied for years.

Researchers have traditionally looked for such weaknesses by hand-crafting test cases based on where they suspected a processor might fail. Led by co-first authors Oleksii Oleksenko and Flavien Solt, the team developed a more systematic approach. Their fuzzer generates short, randomised programs that exercise isolation boundaries, including those between a virtual machine and its host or between the kernel and a user-level process.

Tests on six x86 processors reproduced known leaks and showed that existing mitigations worked. They also revealed four previously undocumented vulnerabilities. One of the most serious, found on AMD Zen 4 processors, could expose another virtual machine's memory one bit at a time. Another could allow a user-level program to recover data recently written by the kernel. AMD confirmed both findings. The researchers kept them under embargo for roughly a year while mitigations were developed. This category of vulnerability is now known as Transient Scheduler Attacks.

The results matter beyond the four vulnerabilities. Cloud platforms routinely place customers who do not trust one another on shared hardware, making strong isolation essential. The fuzzer gives processor vendors a repeatable way to search for this class of flaw before an attacker does.

"The bugs matter, but the method is the real contribution. Testing for this class of leak has depended on intuition for years. Giving vendors a systematic way to find them is a forward-facing contribution of this work."

The work builds on Revizor, an earlier fuzzer created by members of the team, as well as broader research into the security guarantees that hardware must provide to software.

The team also introduces several avenues for future work. Generated programs follow a fixed structure, which may exclude some leak patterns. The fuzzer treats the processor as a black box, leaving it unable to observe what happens inside the chip during a test. Future versions could add this visibility and extend the technique to additional isolation boundaries.

A 2022 paper co-authored by two NUS Computing faculty members has received the Youth Achievement Award at the 10th Ministry of Education Award for Outstanding Scientific Research Achievements （第十届教育部科学研究优秀成果奖）– one of China's most significant national recognitions for research in the humanities and social sciences. 

The paper, "Rural-Urban Healthcare Access Inequality Challenge: Transformative Roles of Information Technology," was published in MIS Quarterly in 2022 by a team led by Yu Tong of Zhejiang University, with co-authors including Provost’s Chair Professor Teo Hock Hai and Professor Tan Chuan Hoo from the Department of Information Systems and Analytics, alongside collaborators from City University of Hong Kong and Southeast University.

The research looks at the divide in healthcare access between rural and urban China, where hospitals cluster in cities and rural residents often rely on village clinics or long journeys to reach proper care. The team examined how effective health information technology has been in narrowing that gap, and argued that the issue reaches beyond logistics into the social fabric of these communities: unresolved frustration among rural patients has, in some cases, spilled into protests, damage to medical facilities, and harassment of healthcare workers.

Getting to those findings meant fieldwork most researchers would shy away from. Data collection was done manually, with the team travelling to rural and urban sites directly and working through extensive policy documentation to understand the regulatory landscape shaping healthcare delivery on the ground.

This is the second honour for the paper, which previously received the Information Management Research Award from the Chinese Economic Society in 2024. That earlier recognition underscored the paper's relevance to universal access to quality healthcare, and its implications for how policymakers design and roll out healthcare initiatives.

The Youth Achievement Award sits among four award categories in this MOE scheme and recognises research led by early-career scholars at Chinese universities, drawn from 206 winning papers selected out of a national pool this year.

Congratulations to Prof Teo, Prof Tan and the full research team on this recognition! 

A research paper by Assistant Professor Lin Shao and collaborators has been selected as a finalist for the ICRA 2026 Best Paper Award, one of the highest honours at the IEEE International Conference on Robotics and Automation (ICRA).

The paper, Bi-Adapt: Few-shot Bimanual Adaptation for Novel Categories of 3D Objects via Semantic Correspondence, addresses a key challenge in robotics: scaling bimanual manipulation skills without relying on expensive human teleoperation data.

The framework learns transferable bimanual affordance representations that are inferred from diverse data sources –online videos, images, foundation models, and robot demonstrations – then adapts these representations across different robot embodiments. The result is a robot that can acquire complex two-handed manipulation skills to a wide range of objects and tasks it has not encountered before.

Bi-Adapt has been tested on a demanding set of real-world scenarios, from handling articulated objects and opening containers to tool use and other dexterous interactions. The work provides a practical pathway towards general-purpose robotic assistants in homes, warehouses, and manufacturing environments.

"We are honoured that Bi-Adapt has been selected as a finalist for the ICRA Best Paper Award. This recognition highlights the importance of developing scalable learning frameworks that enable robots to acquire complex bimanual manipulation skills from diverse sources of experience,” said Assistant Professor Lin Shao. “By allowing robots to learn and transfer skills at scale, Bi-Adapt brings us one step closer to general-purpose robotic intelligence."

Project website: https://biadapt-project.github.io/

A team from NUS School of Computing took the Best Paper and Presentation Award at PPMisDet, a misinformation detection workshop held at CVPR 2026 in Denver in June.

The team is led by Provost’s Chair Professor Mohan Kankanhalli, together with NUS Computing PhD student Xu Danni, NUS Computing Research Fellow Harry Cheng, and Dr Fan Shaojing, former Senior Research Fellow at NUS Computing and current Senior Lecturer in the Department of Electrical and Computer Engineering at College of Design and Engineering.

Their paper, "RW-Post: Auditable Evidence-Grounded Multimodal Fact-Checking in the Wild" addresses a practical gap in automated fact-checking: most existing tools take only a summarised claim as input, which can strip away meaning and context. RW-Post instead preserves social media post content, paired with human-written fact-checking articles annotated with evidence and reasoning chains.

On top of that dataset sits AgentFact, a framework of five specialised agents handling evidence retrieval, visual analysis, reasoning, and explanation. The system is designed to show how it reached a conclusion, not just what it concluded, which the team argues matters for user trust, system improvement, and accountability in high-stakes settings like news verification and public health.

"This recognition is a testimony to the fact that we have very talented students and researchers at NUS who are passionate about tackling the huge problem of misinformation in society,” said Prof Kankanhalli. "It is most gratifying to see such progress being made in AI for Social Good."

The paper is available at arxiv.org/abs/2512.22933.

Today’s AI agents can follow instructions, write code, and answer questions. But ask one to adapt to a new workflow or learn from its own mistakes over time, and the limitations surface quickly. Most agents are static: they perform well on the tasks they were designed for, but adjusting them to new environments still requires significant human engineering. 

Associate Professor Bryan Low’s project on “Self-Configurable Agentic Learning via Co-Optimization” has been selected for the AWS Agentic AI Amazon Research Awards (ARA). His research explores how to unify the strengths of Context Engineering and Reinforcement Learning into a single coherent approach. The research proposes a continuous co-optimisation loop that jointly updates the agent’s architecture and its underlying model parameters together, enabling agents to learn from their environment without the need for manually designed reward signals. 

The idea is that as each agent iterates, its capabilities compound – creating a self-improving cycle that reduces the engineering effort needed to deploy high-performance agents in real-world settings. 

The project builds on a series of recent results from Prof Low’s group of PhD researchers. MEM1, a reinforcement learning method for long-horizon agents developed by Zhou Zijian in collaboration with MIT's SMART programme, won the Best Paper Award at the NeurIPS 2025 Workshop on Multi-Turn Interactions and was subsequently accepted to ICLR 2026. 

A follow-up framework, MeMo (Memory as a Model), developed by Ryan Quek, Alfred Leong and Arun Verma, introduced a modular approach to encode new knowledge required for multi-hop and long-context understanding tasks without modifying the main agent’s parameters. The work was covered by VentureBeat and drew attention from the wider AI community.

A third project, CORAL, developed with Shao Yong Ong and Zhou Zijian, tackles multi-agent evolution, demonstrating that populations of autonomous agents can collaborate and outperform tightly structured systems on complex optimisation tasks.

Together, these three threads –  knowledge compression, knowledge storage, and knowledge utilisation – form the foundation for the self-configurable agents that the AWS-funded project aims to realise. Among the applications the team is most excited about: enterprise-scale deployment, where agents could adapt to custom workflows with minimal engineering, and continuously learning coding, search, and research agents.

Congratulations to Prof Low and team on this well-deserved recognition! 

Every major AI chatbot today runs on powerful cloud servers, often thousands of kilometres from the person using it. For a casual conversation, that works fine. But for a wearable monitoring your heart rate, a robot navigating a warehouse, or a sensor detecting hand gestures – applications where decisions need to happen instantly, privately, and with minimal power – the cloud is not a viable option.

This is the challenge at the heart of Physical AI: embedding intelligence directly into the small, resource-constrained devices that populate our physical world. Most of these devices have a fraction of the computing power needed to run today's language models, and sending data back to the cloud introduces latency, privacy risks, and bandwidth costs that defeat the purpose.

Assistant Professor Ambuj Varshney and his WEISER research group are building a way around this. His project, “TinyLLM: A Framework for Training and Deploying Language Models at the Edge Computers”, has been selected for the 2026 Google Awards for Machine Learning Research and Education with TPUs. The team is designing custom language models – ranging from tens to hundreds of millions of parameters – trained from scratch on curated datasets and optimised to run directly on edge devices.

The group has already developed early prototypes based on optimised GPT-2 architectures, demonstrating that these TinyLLMs can interpret sensor data for hand gesture detection, robot localisation, and vital sign monitoring. The next phase pushes into two domains: wearable health technology, where a local language model can process continuous sensor data without ever leaving the device, preserving patient privacy and enabling immediate feedback; and wireless communications, where TinyLLMs could optimise lower-level network tasks like signal modulation, demodulation, and error recovery.

The project also has a strong educational dimension. Prof Varshney is introducing a new hands-on component in the graduate course CS5272 (Embedded Software Design), where students will learn to design, train, and deploy their own TinyLLMs from scratch for embedded sensing applications.

Working alongside Prof Varshney are PhD students Pramuka Sooriyapatabandige, Rajashekar Reddy Chinthalapani, Kandala Savitha Viswanadh, and Dhairya Shah. The TinyLLM framework is already publicly available as an open-source project. The encourage broader adoption in teaching and research, the team will continue releasing all future models and advances under permissive licenses. 

For more information, visit the project page at tinyllm.org or Prof Varshney's homepage at ambuj.se.

NUS Presidential Young Professor Umang Mathur has been awarded the Temasek-Presidential Young Professorship (T-PYP) grant for his research on safer and more secure hardware design. The project, conducted in collaboration with defence research stakeholders, brings programming-language theory and formal methods into hardware design – making implicit design assumptions explicit and machine-checkable. 

His research tackles a real gap: software engineers have powerful tools to catch bugs early, but hardware designers – work on systems just as critical – largely don’t. The languages and workflows used to design hardware still leave fundamental assumptions about timing, coordination, and module interaction unstated. As hardware grows more complex and more critical to domains such as defence, when a chip has a timing bug, you can’t just push a fix.

Central to the project is Anvil, a hardware description language Assistant Prof Mathur’s group have developed, in collaboration with NUS faculty members Associate Professors Prateek Saxena and Trevor Carlson. The initial version of Anvil allows hardware developers to express timing intent as a first-class element in RTL designs. With the T-PYP grant, Prof Mathur's team will extend this work into new type systems, compilation techniques, and tooling for verifying timing, communication, and security properties – while remaining compatible with existing hardware design workflows. The project begins in June 2026.

Training a large language model is not just about writing good code and pressing run. These models are spread across hundreds or thousands of processors – such as Google's Tensor Processing Units (TPUs), specialised accelerators designed to train and serve large-scale machine learning models – that must learn to work in lockstep: exchanging data, splitting tasks, staying synchronised. When something goes wrong, the whole system slows down, and expensive hardware sits idle.

The problem? Finding and fixing these bottlenecks is still a craft. Engineers pore over profiling traces, adjust configurations by trial and error, and lean on deep systems expertise that most research teams simply do not have.

Professor He Bingsheng wants to change that. His project, selected for the 2026 Google Awards for Machine Learning Research and Education with TPUs, is building tools that automatically diagnose and resolve performance bottlenecks in distributed AI workloads, turning what has been a manual, expert-driven process into an automated, reproducible one.

The project, Lightweight and Automated Performance Optimization of Training and Inference on TPUs, rethinks profiling itself as the foundation of a self-optimising system.

Rather than simply flagging problems for a human to interpret, the tools capture bottlenecks with minimal disruption and generate actionable optimisation strategies on their own. Working alongside Professor He on the project are Visiting Research Fellow Weihao Cui and PhD researchers Feng Yu and Junyi Hou, with a shared long-term vision: making efficient large-scale ML accessible to any researcher, not just a small pool of systems specialists.

The most immediate impact will be on large language model workloads – both dense transformers and mixture-of-experts (MoE) variants – across pre-training, fine-tuning, and inference. MoE models, which route different inputs to specialised sub-networks, are particularly tricky: their sparse, irregular routing patterns make manual performance tuning especially painful. Recommendation systems and embedding-heavy applications face similar challenges, relying on high-dimensional sparse lookups that stress distributed infrastructure in much the same way.

The project is being developed as an open-source initiative under Medusa Compute, with publications planned at top-tier ML and systems venues.