National University of Singapore

Extreme heat is not just a risk factor; it amplifies fault predictions over time, impacting nearly 30% of EV charging posts and reshaping maintenance strategies for climate resilience.

IRIS can detect model substitutions and routing dilutions in LLM gateways with unprecedented accuracy using only the output text, challenging the reliability of commercial AI services.

Hallucinations in chemical reasoning models coexist with correct answers, revealing a complex relationship that challenges our understanding of model reliability.

PrefReward reveals how an explicit user preference matrix can drastically enhance personalization in text generation, outperforming traditional methods in both quality and interpretability.

Unveiling the hidden NUMA architecture of GPUs could revolutionize how we optimize memory efficiency in high-performance computing.

ATSplat achieves over 5.7 times reduction in Gaussian primitives while delivering state-of-the-art rendering quality and real-time performance.

Automating real-to-sim conversion with vision-language agents could revolutionize how we simulate robotic interactions, making it faster and cheaper than ever before.

NSM primes outperform traditional appraisal methods in controlling emotional responses in LLMs, revealing a more effective explanatory framework for AI emotion.

SGN enables effective data generation for shifted target domains without the need for retraining, transforming how we approach data augmentation in machine learning.

Test suites can dramatically enhance issue localization accuracy, bridging the gap between abstract descriptions and concrete code.

Landmark bias can lead to significant inaccuracies in geo-localization, but HoloGeo effectively mitigates this issue through evidence-driven reasoning, outperforming existing models.

Kaleidoscope reveals that a structured, context-aware evaluation process can significantly enhance the reliability of automated scoring in AI applications.

Small visual perturbations can cause World-Action Models to execute harmful actions while still predicting a plausible future, revealing a critical vulnerability in their design.

NodeImport reveals that strategically filtering nodes based on importance can dramatically enhance GNN performance in imbalanced settings.

Quantum topological data encoding reveals that quantum representations can unlock deeper insights from complex datasets, surpassing classical methods in capturing topological nuances.

RecRec reveals that decoupling reasoning from prediction can significantly enhance sequential recommendation performance, breaking free from the constraints of fixed-dimensional states.

EnCF outperforms traditional filters in complex observation scenarios, revealing a new frontier in data assimilation techniques.

LLM-generated bug reports often hinge on implicit assumptions, and this framework reveals how to validate their correctness through a novel witness-generation approach.

Induced anger can lock LLMs into poor decision patterns by reducing their sensitivity to penalties, unlike human decision-making.

TIGER achieves remarkable speedup in multimodal generation by intelligently routing visual tokens based on textual context, outpacing traditional methods.

Modern LLM performance hinges on dependency structures rather than individual instruction latencies, revealing a critical insight for GPU optimization.

Surpassing traditional Monte Carlo methods, this approach offers a stable and efficient alternative for learning neural set functions, dramatically cutting down computational costs.

CycleGRPO achieves simultaneous region understanding and localization in MLLMs without any reliance on textual ground truths, revolutionizing multimodal task integration.

TC-MAF achieves unprecedented anomaly detection performance by effectively fusing RGB and 3D evidence, setting a new benchmark in multimodal industrial applications.

Conveying depth as text rather than images can significantly boost spatial reasoning in vision-language models, challenging conventional approaches.

EasyOPD unifies on-policy distillation methods, enabling seamless integration and superior performance across diverse tasks in large language models.

LLM judges may misinterpret peer review quality, favoring superficial traits over genuine analytical depth, raising questions about their reliability in academic assessments.

Over 1,100 submissions reveal groundbreaking advancements in sports video understanding, with new methods pushing the boundaries of action prediction and localization.

nMAS can cut gastric biopsy report review time from over 83 hours to just 1.4 hours, unlocking substantial efficiency gains in clinical settings.

Navigation models trained in Image2Sim's synthetic environments outperform traditional methods, achieving zero-shot transfer to real-world settings.

Decoupling geometry from semantics in surgical scene understanding leads to unprecedented robustness in 4D reconstruction, even amidst drastic tissue changes.

ResearchStudio-Reel not only automates research dissemination but does so with unprecedented quality, outperforming both traditional methods and leading LLMs in aesthetic appeal and information accuracy.

A unified taxonomy reveals how agentic architectures can transform recommender systems into more autonomous and interactive entities.

ResearchStudio-Idea transforms the ideation process by systematically grounding proposals in literature and identifying unresolved research bottlenecks, leading to more robust and traceable research directions.

LLMs exhibit a surprising mechanism-level routing ceiling, with accuracy jumping by over 10% when provided with specific cues, revealing critical misrouting issues in proof-mechanism classification.

Unlocking lossless parallel generation in dense video captioning could redefine efficiency standards in multimodal AI systems.

A single structural edit can drastically impair LLM performance in molecular tasks, highlighting the fragility of their generalization capabilities.

Projected quantum kernels can drastically improve sample efficiency in Gaussian process optimization, outperforming traditional high-dimensional quantum kernels.

BrainFIBRE reveals neurobiologically interpretable representations that outperform existing methods in predicting critical brain health markers from microstructural data.

Label Influence Propagation reveals that dynamically adjusting label influences can significantly enhance multi-label node classification performance, outperforming existing methods.

Selecting the right Pauli strings can cut quantum error mitigation errors by over a third, using far less data than traditional methods.

Mainstream video models consistently falter in fine-grained understanding, revealing critical vulnerabilities in their hallucination capabilities.

Untrusted AI agents can now operate at unprecedented speeds while maintaining rigorous safety guarantees, achieving a 2.96x speedup with only 2.1% regret.

T2LDM++ generates realistic LiDAR scenes with rich geometric details, overcoming the limitations of existing models that struggle with insufficient training data and controllability.

Flow Splatting achieves superior image quality and faster rendering speeds by efficiently modeling dynamic scenes with 4D Gaussian representations.

Large-scale human motion data can now be effectively repurposed to teach diverse non-humanoid robots, unlocking new capabilities in locomotion and manipulation.

WorldEvolver redefines LLM agent planning by achieving unprecedented prediction accuracy and decision-making success through self-evolving memory mechanisms.

DOPD reveals that intelligently routing supervision based on advantage gaps can significantly enhance capability transfer in distillation, outperforming conventional methods.

HTT enables tactile learning across diverse sensors, achieving adaptability that was previously unattainable in contact-rich manipulation tasks.

Manipulative betting ads on social media are not just misleading; they can significantly harm users' mental well-being, and our new dataset equips researchers to combat this issue effectively.

Metacognitive regulation emerges as a key differentiator for deeper collaboration in human-AI interactions, reshaping our understanding of dialogue dynamics.

Translating C interpreters to safe Rust can be done with minimal human intervention while completely eliminating memory vulnerabilities.

Distinct model capabilities reveal that relational context significantly influences mental health assessments, with Claude-3-Haiku and GPT-4o leading in classification and trigger detection, respectively.

Visual aimbots can be effectively countered with a system that achieves over 85% success in real-time defense while maintaining negligible overhead.

GROVE transforms pedestrian simulation by enabling customizable, realistic scenarios that challenge social robots in ways traditional methods cannot.

PolicyAlign enables LLMs to adapt to rapidly changing safety policies without relying on expensive supervision data, achieving significant safety improvements across diverse applications.

SVP-IL boosts success rates on ambiguous language tasks by over 60% with minimal training data, revolutionizing data efficiency in robotic manipulation.

GRA reveals that grounding VLA models in geometric representations from generated videos can outperform traditional methods that attempt to extract control signals from the same data.

CineCap achieves a new state of the art in cinematographic video captioning by effectively balancing descriptive completeness with factual accuracy through innovative structured reasoning techniques.

RaDaR can identify rare diseases 1.87 months earlier than traditional methods, revolutionizing diagnostic timelines for patients.

LLMs can effectively decompose interactions into phases and roles, but struggle to generate dynamic, realistic motion without a structured approach.

SIES achieves rapid synchronization and adaptive coordination across diverse tasks and scales, outperforming traditional methods without retraining.

Real-time cloth manipulation success rates soar as robots learn to refine actions using a simulator-in-the-loop approach, outperforming traditional methods.

FedOT not only verifies ownership but also traces model leakage back to malicious clients, a critical advancement in federated learning security.

Evo-RAD achieves a groundbreaking +21.04% improvement in diagnosing rare retinal diseases by dynamically refining evidence retrieval, challenging the limitations of static models.

P4IR reduces code compliance errors by up to 38.6% while outperforming top LLMs in accuracy and reliability.

Discretizing reward models can significantly enhance policy performance by reducing oversensitivity without sacrificing discriminative ability.

Performance of large reasoning models drops significantly as logical complexity rises, revealing critical gaps in current evaluation benchmarks.

ASYS reveals a novel way to automate the discovery of analytical forms for PDEs, producing interpretable solutions where none existed before.

CARE transforms the approach to reasoning length in video-MLLMs, enabling models to adaptively balance exploration and efficiency based on their evolving competence.

WAMs are evolving beyond mere video generators, revealing a critical trade-off between representational richness and computational efficiency in predictive-action modeling.

SPOT-E transforms frozen VLMs into more reliable evidence readers by dynamically spotlighting critical visual information during inference.

Semantic features derived from Visual Question Answering can dramatically enhance the robustness of cancer prognosis models across diverse clinical settings.

Compositionality in neural networks only emerges in a narrow depth-connectivity regime, with specific architectural constraints dictating success or failure.

Quantum ring all-reduce slashes communication costs in distributed training while delivering privacy guarantees unattainable by classical methods.

Achieving full-stack fidelity in live simulations without sacrificing performance could revolutionize how we evaluate distributed systems before deployment.

MAFP reveals that treating stakeholder stances as agents in a game-theoretic framework can drastically improve decision quality in complex scenarios.

Current memory agents fail to provide reliable governance in shared settings, with no method achieving a balance between utility, access control, and forgetting.

CoEV not only detects hallucinations in medical VLMs but also corrects them in real-time, enhancing diagnostic reliability without retraining.

MAST achieves targeted forgetting in RLVR-induced reasoning with minimal impact on performance, preserving critical task accuracy while effectively unlearning unwanted knowledge.

VideoCFR not only boosts performance in video reasoning tasks but also reveals the critical visual evidence driving model decisions without relying on human annotations.

A new benchmark and model reveal that even minimal cross-scale supervision can dramatically enhance pathological image interpretation.

Coding agents struggle to create complete and engaging games, with top performers barely reaching 41.46% success in end-to-end game generation.

OPD-Evolver outperforms traditional memory systems by up to 11.5%, showcasing a new paradigm in agent evolution that transcends mere memory storage.

Fine-tuning neural operators with PhysGuard can reduce low-frequency error by up to 32% under severe domain shifts, preserving essential physics while adapting to real-world data.

Joycent synthesizes accented speech directly from standard phone sequences, eliminating the need for error-prone accented phone predictions.

AURA achieves 95.47% accuracy in identifying active antibiotics, revealing that traditional methods misidentify the active set over 60% of the time.

Achieving robot navigation policy training in under 20 seconds could revolutionize the deployment of DRL in robotics.

Roken can generate coordinated multi-robot trajectories in real-time, achieving higher success rates than traditional sequential planning methods.

Achieving up to 10x weight compression in LLMs without altering weights could revolutionize GPU memory usage and model deployment efficiency.

Cross-lingual evidence significantly hampers the performance of deep research agents, revealing critical integration challenges that go beyond mere retrieval failures.

CFALR outperforms traditional methods by seamlessly integrating collaborative filtering with large language models for personalized fashion recommendations.

CHOP achieves superior performance on OOD tasks by utilizing a frozen ICON, revealing that interpretability and generalization can coexist in operator learning.

Latent Memory slashes token usage by up to 10x while maintaining competitive performance in multimodal question answering.

Object detection frameworks can nearly double the accuracy of bird call localization in noisy environments, transforming wildlife monitoring practices.

Algorithmic complexity can reveal insights that class-wide minimax certificates miss, reshaping our understanding of kernel bandits.

TacForeSight enables robots to anticipate contact changes in real-time, outperforming traditional methods in dynamic manipulation tasks.

READER reveals that even frozen LLMs can expose rich authorship signals, achieving up to 84% accuracy in identifying model sources from black-box outputs.

ASP can rival traditional ILP methods in e-graph extraction efficiency while uncovering previously hidden optimal solutions.

Uncovering how VR motion sensors can reconstruct brainwave data reveals a startling new vector for privacy breaches in the Metaverse.