Showing 1–50 of 84 results for author: Yiu, S

Structured Latent Space Modeling over Multi-Scale Temporal Patches for Multivariate Time Series Forecasting

Authors: Xingsheng Chen, Deyu Yi, Siu-Ming Yiu

Abstract: Existing patching and multi-scale methods advance multivariate time series forecasting but treat learned representations as transient byproducts of prediction, lacking explicit mechanisms that enforce structural consistency across temporal scales. We propose M2Patch, a CNN-based architecture that organizes channel-independent observations into a structured latent space via two complementary differ… ▽ More Existing patching and multi-scale methods advance multivariate time series forecasting but treat learned representations as transient byproducts of prediction, lacking explicit mechanisms that enforce structural consistency across temporal scales. We propose M2Patch, a CNN-based architecture that organizes channel-independent observations into a structured latent space via two complementary differentiable penalties. Multi-scale patching decomposes the input into overlapping temporal granularities, depthwise separable CNN blocks with progressively growing dilation extracts scale-specific features at linear complexity, and per-scale learned projections compress these features into a compact latent representation. An intra-scale smoothness penalty enforces temporal continuity between adjacent patches, while an inter-scale alignment penalty restores cross-granularity interaction through learnable cross-scale mappings, so that all scales encode mutually consistent representations of the underlying dynamics. Extensive experiments on ten real-world benchmark datasets demonstrate that M2Patch significantly outperforms state-of-the-art baselines. Further analyses establish M2Patch as a structure-aware recognizer: it recovers channel functional groupings and remains robust under patch-level input corruption, confirming that the structured latent space captures the data's intrinsic dynamics. △ Less

Submitted 19 August, 2026; v1 submitted 7 July, 2026; originally announced July 2026.

The Interference Gap: Comparing Retrieval Bounds in Human Memory and RAG Systems

Authors: Dongxin Guo, Jikun Wu, Siu-Ming Yiu

Abstract: How do retrieval bounds compare between human episodic memory and Retrieval-Augmented Generation (RAG) systems under semantic interference? We present a unified signal detection theory (SDT) framework that applies to both, and use it to fit behavioral and computational data in matched paradigms. Both systems show logarithmic accuracy decline with association count (fan), but humans exhibit lower i… ▽ More How do retrieval bounds compare between human episodic memory and Retrieval-Augmented Generation (RAG) systems under semantic interference? We present a unified signal detection theory (SDT) framework that applies to both, and use it to fit behavioral and computational data in matched paradigms. Both systems show logarithmic accuracy decline with association count (fan), but humans exhibit lower interference sensitivity ($α/σ= 0.41$) than dense passage retrieval ($α/σ= 0.67$), with cognitively-inspired HippoRAG falling between the two ($α/σ= 0.44$). Behavioral experiments ($N = 112$) and simulations validate the framework; parameter recovery confirms identifiability ($r \geq .93$) and model comparison favors the logarithmic specification over a power-law alternative ($Δ$BIC $> 15$). We discuss encoding specificity, temporal context binding, and retrieval gating as candidate mechanisms whose causal role remains to be established. Six falsifiable predictions connect cognitive memory research with AI retrieval evaluation. △ Less

Submitted 9 May, 2026; originally announced June 2026.

Comments: 8 pages, 2 figures, 1 table. Accepted at CogSci 2026

ACM Class: H.3.3; I.2.7; J.4

SPDM: Geometry-Modulated State Space Modeling with Manifold Constraints for Time Series Forecasting

Authors: Xingsheng Chen, Siu-Ming Yiu

Abstract: Multivariate time series forecasting requires capturing the continuously evolving correlation structure among interacting variables. Existing state-space models process time series by scanning tokenized temporal or spatial sequences, discarding the evolutionary geometric structure. We address this limitation by introducing manifold constraints into state-space modeling: treating the cross-variable… ▽ More Multivariate time series forecasting requires capturing the continuously evolving correlation structure among interacting variables. Existing state-space models process time series by scanning tokenized temporal or spatial sequences, discarding the evolutionary geometric structure. We address this limitation by introducing manifold constraints into state-space modeling: treating the cross-variable correlation structure as a continuous trajectory on the symmetric positive definite manifold, whose Riemannian geometric features, tangent space linearity, and Frechet mean centrality act as a principled geometric regularizer that guides and stabilizes the selective scanning dynamics of SSMs. We propose SPDM, a geometry-aware SSM architecture that realizes this principle through two cooperating mechanisms: a manifold trajectory path that projects dynamically evolving covariance matrices from the SPD manifold to a Euclidean tangent space, and a geometric gating scheme that directly modulates SSM's internal selective parameters based on geometric signals derived from the manifold trajectory. The parameterization preserves the linear-time complexity of the Mamba parallel scan while embedding rich structural constraints, making the architecture preserve prediction accuracy and computational efficiency simultaneously. Extensive experiments on eleven real-world benchmark datasets establish state-of-the-art forecasting performance, and further studies confirm that geometrically constrained state-space dynamics are the dominant architectural factor behind its performance gains. △ Less

Submitted 6 June, 2026; originally announced June 2026.

The Deterministic Horizon: When Extended Reasoning Fails and Tool Delegation Becomes Necessary

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Extended chain-of-thought reasoning can degrade performance on deterministic state-tracking tasks, not solely because of preference biases but, on the evidence we present, because of information-theoretic limits in the capacity of decoder-only attention. We present: (1) an Attention Bottleneck analysis providing evidence that total state-tracking capacity in bits is bounded in terms of head count,… ▽ More Extended chain-of-thought reasoning can degrade performance on deterministic state-tracking tasks, not solely because of preference biases but, on the evidence we present, because of information-theoretic limits in the capacity of decoder-only attention. We present: (1) an Attention Bottleneck analysis providing evidence that total state-tracking capacity in bits is bounded in terms of head count, head dimension, and context length under stated modeling assumptions, and that total capacity is not the binding constraint; (2) a context-dependent error model with a depth-dependent quadratic term in the error exponent; (3) the State-Space Jaccard metric measuring state drift; and (4) a Deterministic Horizon $d^* \in [19, 31]$ (at $α= 0.5$) marking the depth at which unaided accuracy crosses 50%. Across twelve models and eight task domains (including SWE-Bench, WebArena, and SQL-Multi), tool-integrated reasoning reaches 76-94% accuracy versus 17-42% for neural chain-of-thought on PermutationProbe. Fine-tuning on optimal-length traces yields $<$3 percentage-point improvement, supporting an architectural ceiling. △ Less

Submitted 8 August, 2026; v1 submitted 29 May, 2026; originally announced June 2026.

Model Collapse as Cultural Evolution

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Model collapse, the progressive degradation of LLMs trained on their own outputs, has been characterized statistically but lacks a linguistic explanation for which structures degrade, in what order, and why. We show that iterated learning theory from cultural evolution fills this gap. We derive five falsifiable predictions, distinguish those uniquely discriminative for the theory from confirmatory… ▽ More Model collapse, the progressive degradation of LLMs trained on their own outputs, has been characterized statistically but lacks a linguistic explanation for which structures degrade, in what order, and why. We show that iterated learning theory from cultural evolution fills this gap. We derive five falsifiable predictions, distinguish those uniquely discriminative for the theory from confirmatory ones, and test them by self-training LLaMA-2-7B and Mistral-7B over 10 generations in English, German, and Turkish. The critical discriminative finding: compositionality follows a non-monotonic trajectory (initially rising, then falling) under unfiltered self-training. This signature persists with maximally regular seed data (ruling out noise removal) and is sustained only by task-grounded filtering, not random filtering, providing the first LLM-scale evidence for the compression-communication tradeoff. All predictions are confirmed with large effect sizes (Hedges' $g > 1.6$; $\mathrm{BF}_{10} > 100$), and LLM regularization gradients closely match human behavioral data ($R^2 = 0.94$). These results reframe model collapse as a cultural transmission phenomenon and yield concrete principles for self-training pipeline design. △ Less

Submitted 21 May, 2026; originally announced May 2026.

Comments: Accepted at CoNLL 2026. 18 pages, 3 figures, 2 tables

ACM Class: I.2.7; I.2.6

Do Language Models Know What Not to Say? Causal Evidence for Statistical Preemption in LLMs

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: How do learners acquire knowledge of what is unacceptable without negative evidence? Construction Grammar proposes statistical preemption: exposure to a conventional form (e.g., "donated the books to the library") preempts structurally possible but unattested alternatives ("*donated the library the books"). We present a computational study that, for the first time, directly dissociates statistical… ▽ More How do learners acquire knowledge of what is unacceptable without negative evidence? Construction Grammar proposes statistical preemption: exposure to a conventional form (e.g., "donated the books to the library") preempts structurally possible but unattested alternatives ("*donated the library the books"). We present a computational study that, for the first time, directly dissociates statistical preemption from the competing entrenchment hypothesis in large language models within a single converging design. Across four experiments spanning 120 English verb-construction pairings (dative, causative, locative), we show that (1) LLM surprisal patterns correlate strongly with human acceptability judgments ($r = 0.79$), validated against three independent behavioral datasets; (2) these patterns are driven by competing-form frequency rather than overall verb frequency, confirmed by non-circular partial correlations; (3) preemption sensitivity scales as a power law with model size; and (4) a controlled fine-tuning intervention causally demonstrates that manipulating competing-form frequencies shifts preemption behavior in the predicted direction, with reverse-direction controls ruling out frequency-sensitivity confounds. These results provide converging evidence that neural language models acquire negative linguistic knowledge through distributional competition, the core mechanism posited by Construction Grammar. △ Less

Submitted 21 May, 2026; originally announced May 2026.

Comments: Accepted at CoNLL 2026. 21 pages (9 main body + appendices and references); 4 figures, 14 tables

ACM Class: I.2.7; I.2.6

Sparse Autoencoders Map Brain-LLM Alignment onto Cortical Semantic Topography

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Intermediate layers of large language models (LLMs) best predict human brain responses to language, one of the most robust findings in computational neurolinguistics, yet why remains mechanistically unexplained. We address this gap by bridging sparse autoencoders (SAEs) from mechanistic interpretability with neural encoding models, decomposing GPT-2 XL and Llama-3.1-8B into 16K-32K interpretable f… ▽ More Intermediate layers of large language models (LLMs) best predict human brain responses to language, one of the most robust findings in computational neurolinguistics, yet why remains mechanistically unexplained. We address this gap by bridging sparse autoencoders (SAEs) from mechanistic interpretability with neural encoding models, decomposing GPT-2 XL and Llama-3.1-8B into 16K-32K interpretable features per layer. A human-validated taxonomy ($κ\geq 0.74$) reveals that semantic features alone recover 94% of peak encoding performance ($r=0.285$), substantially exceeding variance-matched baselines ($p<0.001$, $d=1.31$). Beyond this aggregate dominance, we test a novel cortical topography prediction: five semantic subcategories derived a priori from three independent neuroscience programs should map onto distinct brain regions. A formal convergence test confirms this alignment (Spearman $ρ=0.72$, $p<0.001$; hypergeometric $p=0.007$), demonstrating that SAE-discovered features recapitulate known cortical semantic organization at a granularity inaccessible to prior methods. SAE features further predict human reading times beyond lexical controls ($Δ\mathrm{logLik}=38.4$, $p<0.001$), and an exploratory prediction-error analysis provides preliminary evidence that the brain additionally encodes unexpected semantic content. Results generalize across English, Chinese, and French. △ Less

Submitted 21 May, 2026; originally announced May 2026.

Comments: Accepted at CoNLL 2026. 20 pages (9 main + 1 limitations/acknowledgments + 3 references + 7 appendix), 5 figures, 20 tables

ACM Class: I.2.7; I.2.6; J.3

Brain-LLM Alignment Tracks Training Data, Not Typology

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Brain-LLM alignment is well established in English, yet the brain's language network is neuroanatomically universal across languages. Does alignment also generalize cross-linguistically, and what governs the variation? We test this using fMRI data from 112 participants across English, Chinese, and French (the Le Petit Prince corpus) and seven LLMs spanning English-dominant, Chinese-dominant, and m… ▽ More Brain-LLM alignment is well established in English, yet the brain's language network is neuroanatomically universal across languages. Does alignment also generalize cross-linguistically, and what governs the variation? We test this using fMRI data from 112 participants across English, Chinese, and French (the Le Petit Prince corpus) and seven LLMs spanning English-dominant, Chinese-dominant, and multilingual architectures. Our central finding is that training-language dominance, not an inherent property of English, drives the alignment pattern: a Chinese-dominant model (Baichuan2-7B), architecture-matched to LLaMA-2-7B, reverses the gradient entirely, aligning best with Chinese brains and worst with English. Beyond training dominance, formal typological distance independently covaries with alignment degradation, syntax-associated brain regions (IFG) show $2.3\times$ steeper typological gradients than lexico-semantic regions (PTL), and tokenization fertility accounts for $\sim$60% of a cross-linguistic shift in optimal encoding layer. These results reveal that the apparent "English advantage" in brain-LLM alignment is an artifact of training data composition, while the remaining variation reflects genuine typological structure concentrated in syntactic processing. △ Less

Submitted 21 May, 2026; originally announced May 2026.

Comments: Accepted to CoNLL 2026. 9 pages main content + 4 pages references + 6 pages appendix; 4 figures, 13 tables

ACM Class: I.2.7; I.2.6; J.3

Parameter-Efficient Neuroevolution for Diverse LLM Generation: Quality-Diversity Optimization via Prompt Embedding Evolution

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Large Language Models exhibit mode collapse, producing homogeneous outputs that fail to explore valid solution spaces. We present QD-LLM, a framework for parameter-efficient neuroevolution that evolves prompt embeddings, compact neural interfaces (~32K parameters) that steer generation in frozen LLMs (70B+ parameters), within a Quality-Diversity (QD) optimization framework. Our contributions: (1)… ▽ More Large Language Models exhibit mode collapse, producing homogeneous outputs that fail to explore valid solution spaces. We present QD-LLM, a framework for parameter-efficient neuroevolution that evolves prompt embeddings, compact neural interfaces (~32K parameters) that steer generation in frozen LLMs (70B+ parameters), within a Quality-Diversity (QD) optimization framework. Our contributions: (1) evolved prompt embeddings via gradient-free optimization enabling behavioral steering without model fine-tuning; (2) hybrid behavior characterization combining semantic and explicit features with formal coverage bounds (Theorem 1) under validated near-independence (NMI $= 0.08 \pm 0.02$); (3) co-evolutionary variation operators including targeted behavioral mutation via finite-difference gradient estimation. On HumanEval (164 problems), MBPP, and creative writing benchmarks, QD-LLM achieves 46.4% higher coverage and 41.4% higher QD-Score than QDAIF ($p<0.001$, 30 runs, Vargha-Delaney $A=0.94$). We demonstrate downstream utility: diverse archives improve test generation (34% more edge cases) and fine-tuning data quality (8.3% accuracy gain). We validate across open-source LLMs (Llama-3-70B, Mistral-Large) with full embedding access, establishing prompt embedding evolution as an effective paradigm bridging neuroevolution and modern LLMs. △ Less

Submitted 18 June, 2026; v1 submitted 10 May, 2026; originally announced May 2026.

EvoPref: Multi-Objective Evolutionary Optimization Discovers Diverse LLM Alignments Beyond Gradient Descent

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Gradient-based preference optimization methods for large language model (LLM) alignment suffer from preference collapse, converging to narrow behavioral modes while neglecting preference diversity. We introduce EvoPref, a multi-objective evolutionary algorithm that maintains populations of Low-Rank Adaptation (LoRA) adapters optimized across helpfulness, harmlessness, and honesty objectives using… ▽ More Gradient-based preference optimization methods for large language model (LLM) alignment suffer from preference collapse, converging to narrow behavioral modes while neglecting preference diversity. We introduce EvoPref, a multi-objective evolutionary algorithm that maintains populations of Low-Rank Adaptation (LoRA) adapters optimized across helpfulness, harmlessness, and honesty objectives using Non-dominated Sorting Genetic Algorithm II (NSGA-II) selection with archive-based diversity preservation. Our primary contribution is demonstrating that population-based methods discover substantially more diverse alignments than gradient descent. On standard benchmarks, EvoPref improves preference coverage by 18% (median 82.5% vs. 70.0% for ORPO, $p<0.001$, Wilcoxon, $n=30$) and reduces collapse rates by 47% (11.0% vs. 20.6%, $p<0.001$), while achieving competitive alignment quality (median 75.5% RewardBench vs. 75.0% for ORPO, $p<0.05$). We provide theoretical motivation extending recent multi-objective evolutionary algorithm (MOEA) runtime analysis (Dang et al., 2025) suggesting why archive-based methods escape collapse more effectively than single-trajectory optimization. Comprehensive comparisons against MOEA/D, SMS-EMOA, CMA-ES, and gradient baselines (DPO, IPO, KTO, ORPO) with rigorous statistical testing (Friedman with Holm correction, Vargha-Delaney effect sizes, median with IQR) confirm that multi-objective selection with diversity preservation is essential. This work establishes evolutionary optimization as a principled paradigm for diverse LLM alignment. △ Less

Submitted 18 June, 2026; v1 submitted 10 May, 2026; originally announced May 2026.

Bias by Necessity: Impossibility Theorems for Sequential Processing with Convergent AI and Human Validation

Authors: Jikun Wu, Dongxin Guo, Siu-Ming Yiu

Abstract: Are certain cognitive biases mathematically inevitable consequences of sequential information processing? We prove that primacy effects, anchoring, and order-dependence are architecturally necessary in autoregressive language models due to causal masking constraints. Our three impossibility theorems establish: (1) primacy bias arises from asymmetric attention accumulation; (2) anchoring emerges fr… ▽ More Are certain cognitive biases mathematically inevitable consequences of sequential information processing? We prove that primacy effects, anchoring, and order-dependence are architecturally necessary in autoregressive language models due to causal masking constraints. Our three impossibility theorems establish: (1) primacy bias arises from asymmetric attention accumulation; (2) anchoring emerges from sequential conditioning with provable information bounds; and (3) exact debiasing by permutation marginalization requires factorial-time computation, with Monte Carlo approximation feasible at constant per-tolerance overhead. We validate these bounds across 12 frontier LLMs ($R^2 = 0.89$; $Δ$BIC $= 16.6$ vs. next-best alternative). We then derive quantitative predictions from the framework and test them in two pre-registered human experiments ($N = 464$ analyzed). Study 1 confirms anchor position modulates anchoring magnitude ($d = 0.52$, BF$_{10} = 847$). Study 2 shows working memory load amplifies primacy bias ($d = 0.41$, BF$_{10} = 156$), with WM capacity predicting bias reduction ($r = -.38$). These convergent findings reframe cognitive biases as resource-rational responses to sequential processing. △ Less

Submitted 9 May, 2026; originally announced May 2026.

Comments: 6 pages, 3 figures, 5 tables. Accepted to CogSci 2026

ACM Class: I.2.7; I.2.6; J.4; F.1.3

When Can Human-AI Teams Outperform Individuals? Tight Bounds with Impossibility Guarantees

Authors: Dongxin Guo, Jikun Wu, Siu-Ming Yiu

Abstract: Human-AI teams fail to outperform their best member in 70% of studies, yet no theory specifies when complementarity is achievable. We derive tight bounds for the broad class of confidence-based aggregation rules by integrating signal detection theory with information-theoretic analysis, yielding four results: (1) a complementarity theorem (teams outperform individuals iff error correlation… ▽ More Human-AI teams fail to outperform their best member in 70% of studies, yet no theory specifies when complementarity is achievable. We derive tight bounds for the broad class of confidence-based aggregation rules by integrating signal detection theory with information-theoretic analysis, yielding four results: (1) a complementarity theorem (teams outperform individuals iff error correlation $ρ_{HM} < ρ^*$, with $ρ^* \approx a$ in the symmetric near-chance regime); (2) minimax bounds showing gains scale as $Θ(\sqrt{Δd})$ with metacognitive sensitivity difference; (3) an impossibility result proving no confidence-based aggregation rule achieves complementarity when $ρ_{HM} \geq ρ^*$; and (4) multi-class generalization $ρ^*_K \approx ρ^*/\sqrt{K-1}$. Predictions match observed team accuracy ($R = 0.94$ on ImageNet-16H, $R = 0.91$ on CIFAR-10H) and the multi-class threshold scaling holds on human data ($R = 0.93$, $K = 16$), with robustness under non-Gaussian distributions. The framework explains why complementarity is rare and provides actionable design formulas; results apply to aggregation, not to interactive deliberation that generates novel answers. △ Less

Submitted 9 May, 2026; originally announced May 2026.

Comments: 8 pages, 2 figures, 7 tables. Accepted at CogSci 2026

ACM Class: I.2.11; I.2.6; H.1.2

Efficient Prompt Learning for Traffic Forecasting

Authors: Qianru Zhang, Xinyi Gao, Alexander Zhou, Reynold Cheng, Siu-Ming Yiu, Hongzhi Yin

Abstract: Accurate traffic prediction is essential for optimizing transportation systems, enhancing resource allocation, and improving overall urban administration. Spatio-temporal graph neural networks (GNNs) have achieved state-of-the-art performance and have been widely used in various spatio-temporal prediction scenarios. However, these prediction methods often exhibit low generalization ability, strugg… ▽ More Accurate traffic prediction is essential for optimizing transportation systems, enhancing resource allocation, and improving overall urban administration. Spatio-temporal graph neural networks (GNNs) have achieved state-of-the-art performance and have been widely used in various spatio-temporal prediction scenarios. However, these prediction methods often exhibit low generalization ability, struggling with distribution shifts caused by spatio-temporal dynamics. To address this challenge, we propose an approach to enhance the generalization and adaptation of spatio-temporal GNNs through efficient prompting. Specifically, we introduce a lightweight and model-agnostic prompt tuning framework for spatio-temporal GNNs, named SimpleST. It facilitates adapting pre-trained spatio-temporal GNNs to novel distributions while keeping the model parameters fixed. This prompt mechanism reduces the overhead and complexity of adaptation, enabling efficient utilization of pre-trained models for out-of-distribution generalization. Extensive experiments conducted on five real-world urban spatio-temporal datasets demonstrate the superiority of our approach in terms of prediction accuracy and computational efficiency. △ Less

Submitted 7 May, 2026; originally announced May 2026.

Comments: 24 pages. This paper is accepted by VLDBJ

Journal ref: The VLDB Journal of 2026

SAGA: Workflow-Atomic Scheduling for AI Agent Inference on GPU Clusters

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: AI agents execute tens to hundreds of chained LLM calls per task, yet GPU schedulers treat each call as independent, discarding gigabytes of intermediate state between steps and inflating end-to-end latency by 3-8x. We argue that this request-level abstraction is fundamentally mismatched to compound AI workloads, and propose a shift to program-level scheduling: treating the entire agent workflow (… ▽ More AI agents execute tens to hundreds of chained LLM calls per task, yet GPU schedulers treat each call as independent, discarding gigabytes of intermediate state between steps and inflating end-to-end latency by 3-8x. We argue that this request-level abstraction is fundamentally mismatched to compound AI workloads, and propose a shift to program-level scheduling: treating the entire agent workflow (not individual inference calls) as the first-class schedulable unit. We present SAGA, a distributed scheduler that implements this abstraction through three mechanisms: (1) Agent Execution Graphs that capture workflow structure to predict KV cache reuse across tool-call boundaries, achieving within 1.31x of Bélády's optimal offline policy; (2) session-affinity batching with work stealing that co-locates correlated requests while maintaining global load balance; and (3) Agent Fair Share, a task-completion-time fairness metric with provable bounded-deviation guarantees. On a 64-GPU cluster serving SWE-bench coding agents and WebArena browser tasks, SAGA reduces task completion time by 1.64x (geometric mean, p < 0.001) over vLLM v0.15.1 with prefix caching and affinity routing, while improving GPU memory utilization by 1.22x and achieving 99.2% SLO attainment under multi-tenant interference. These latency gains come at a quantified cost: approximately 30% lower peak throughput than throughput-optimal batch scheduling, a tradeoff appropriate for the latency-sensitive interactive deployments that dominate compound AI usage. Our results demonstrate that workflow-aware scheduling is essential for efficient compound AI serving. △ Less

Submitted 18 June, 2026; v1 submitted 1 May, 2026; originally announced May 2026.

When to Retrieve During Reasoning: Adaptive Retrieval for Large Reasoning Models

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Large reasoning models such as DeepSeek-R1 and OpenAI o1 generate extended chains of thought spanning thousands of tokens, yet their integration with retrieval-augmented generation (RAG) remains fundamentally misaligned. Current RAG systems optimize for providing context before reasoning begins, while reasoning models require evidence injection during multi-step inference chains. We introduce ReaL… ▽ More Large reasoning models such as DeepSeek-R1 and OpenAI o1 generate extended chains of thought spanning thousands of tokens, yet their integration with retrieval-augmented generation (RAG) remains fundamentally misaligned. Current RAG systems optimize for providing context before reasoning begins, while reasoning models require evidence injection during multi-step inference chains. We introduce ReaLM-Retrieve, a reasoning-aware retrieval framework that addresses this mismatch through three key innovations: (1) a step-level uncertainty detector that identifies knowledge gaps at reasoning-step granularity rather than token or sentence level; (2) a retrieval intervention policy that learns when external evidence maximally benefits ongoing reasoning; and (3) an efficiency-optimized integration mechanism that reduces per-retrieval overhead by 3.2x compared to naive integration. Experiments on MuSiQue, HotpotQA, and 2WikiMultiHopQA demonstrate that ReaLM-Retrieve achieves on average 10.1% absolute improvement in answer F1 over standard RAG (range: 9.0-11.8% across the three benchmarks) while reducing retrieval calls by 47% compared to fixed-interval approaches like IRCoT (all improvements significant at p<0.01, paired bootstrap). On the challenging MuSiQue benchmark requiring 2-4 hop reasoning, our method achieves 71.2% F1 with an average of only 1.8 retrieval calls per question. Analysis shows that ReaLM-Retrieve also improves retrieval quality itself, achieving 81.3% Recall@5 with consistently higher precision and MRR than fixed-interval baselines on supporting evidence, establishing new state-of-the-art efficiency-accuracy trade-offs for reasoning-intensive retrieval tasks. △ Less

Submitted 29 April, 2026; originally announced April 2026.

Comments: 12 pages, 3 figures, 9 tables. Accepted at SIGIR 2026 (49th International ACM SIGIR Conference on Research and Development in Information Retrieval), Melbourne, Australia

ACM Class: H.3.3; I.2.7

FinGround: Detecting and Grounding Financial Hallucinations via Atomic Claim Verification

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Financial AI systems must produce answers grounded in specific regulatory filings, yet current LLMs fabricate metrics, invent citations, and miscalculate derived quantities. These errors carry direct regulatory consequences as the EU AI Act's high-risk enforcement deadline approaches (August 2026). Existing hallucination detectors treat all claims uniformly, missing 43% of computational errors tha… ▽ More Financial AI systems must produce answers grounded in specific regulatory filings, yet current LLMs fabricate metrics, invent citations, and miscalculate derived quantities. These errors carry direct regulatory consequences as the EU AI Act's high-risk enforcement deadline approaches (August 2026). Existing hallucination detectors treat all claims uniformly, missing 43% of computational errors that require arithmetic re-verification against structured tables. We present FinGround, a three-stage verify-then-ground pipeline for financial document QA. Stage 1 performs finance-aware hybrid retrieval over text and tables. Stage 2 decomposes answers into atomic claims classified by a six-type financial taxonomy and verified with type-routed strategies including formula reconstruction. Stage 3 rewrites unsupported claims with paragraph- and table-cell-level citations. To cleanly isolate verification value from retrieval quality, we propose retrieval-equalized evaluation as standard methodology for RAG verification research: when all systems receive identical retrieval, FinGround still reduces hallucination rates by 68% over the strongest baseline ($p < 0.01$). The full pipeline achieves a 78% reduction relative to GPT-4o. An 8B distilled detector retains 91.4% F1 at 18x lower per-claim latency, enabling $0.003/query deployment, supported by qualitative signals from a four-week analyst pilot. △ Less

Submitted 26 April, 2026; originally announced April 2026.

Comments: Accepted to ACL 2026 Industry Track. 14 pages, 1 figure, 14 tables

ACM Class: I.2.7; H.3.3; I.2.6; J.4

ComplianceNLP: Knowledge-Graph-Augmented RAG for Multi-Framework Regulatory Gap Detection

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Financial institutions must track over 60,000 regulatory events annually, overwhelming manual compliance teams; the industry has paid over USD 300 billion in fines and settlements since the 2008 financial crisis. We present ComplianceNLP, an end-to-end system that automatically monitors regulatory changes, extracts structured obligations, and identifies compliance gaps against institutional polici… ▽ More Financial institutions must track over 60,000 regulatory events annually, overwhelming manual compliance teams; the industry has paid over USD 300 billion in fines and settlements since the 2008 financial crisis. We present ComplianceNLP, an end-to-end system that automatically monitors regulatory changes, extracts structured obligations, and identifies compliance gaps against institutional policies. The system integrates three components: (1) a knowledge-graph-augmented RAG pipeline grounding generations in a regulatory knowledge graph of 12,847 provisions across SEC, MiFID II, and Basel III; (2) multi-task obligation extraction combining NER, deontic classification, and cross-reference resolution over a shared LEGAL-BERT encoder; and (3) compliance gap analysis that maps obligations to internal policies with severity-aware scoring. On our benchmark, ComplianceNLP achieves 87.7 F1 on gap detection, outperforming GPT-4o+RAG by +3.5 F1, with 94.2% grounding accuracy ($r=0.83$ vs. human judgments) and 83.4 F1 under realistic end-to-end error propagation. Ablations show that knowledge-graph re-ranking contributes the largest marginal gain (+4.6 F1), confirming that structural regulatory knowledge is critical for cross-reference-heavy tasks. Domain-specific knowledge distillation (70B $\to$ 8B) combined with Medusa speculative decoding yields $2.8\times$ inference speedup; regulatory text's low entropy ($H=2.31$ bits vs. $3.87$ general text) produces 91.3% draft-token acceptance rates. In four months of parallel-run deployment processing 9,847 updates at a financial institution, the system achieved 96.0% estimated recall and 90.7% precision, with a $3.1\times$ sustained analyst efficiency gain. We report deployment lessons on trust calibration, GRC integration, and distributional shift monitoring for regulated-domain NLP. △ Less

Submitted 26 April, 2026; originally announced April 2026.

Comments: Accepted at ACL 2026 Industry Track. 19 pages, 15 tables, 1 figure

ACM Class: I.2.7; H.3.3; I.2.4; I.2.6

AgentEval: DAG-Structured Step-Level Evaluation for Agentic Workflows with Error Propagation Tracking

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Agentic systems that chain reasoning, tool use, and synthesis into multi-step workflows are entering production, yet prevailing evaluation practices like end-to-end outcome checks and ad-hoc trace inspection systematically mask the intermediate failures that dominate real-world error budgets. We present AgentEval, a framework that formalizes agent executions as evaluation directed acyclic graphs (… ▽ More Agentic systems that chain reasoning, tool use, and synthesis into multi-step workflows are entering production, yet prevailing evaluation practices like end-to-end outcome checks and ad-hoc trace inspection systematically mask the intermediate failures that dominate real-world error budgets. We present AgentEval, a framework that formalizes agent executions as evaluation directed acyclic graphs (DAGs), where each node carries typed quality metrics assessed by a calibrated LLM judge (GPT-4o), classified through a hierarchical failure taxonomy (3 levels, 21 subcategories), and linked to upstream dependencies for automated root cause attribution. An ablation study isolates the impact of DAG-based dependency modeling: it alone contributes +22 percentage points to failure detection recall and +34 pp to root cause accuracy over flat step-level evaluation with identical judges and rubrics. Across three production workflows (450 test cases, two agent model families, predominantly sequential architectures with a 12% non-DAG trace rate), AgentEval achieves 2.17x higher failure detection recall than end-to-end evaluation (0.89 vs. 0.41), Cohen's kappa = 0.84 agreement with human experts, and 72% root cause accuracy against an 81% human ceiling. Cross-system evaluation on tau-bench and SWE-bench traces confirms transferability (failure detection recall >= 0.78) without taxonomy or rubric modification. A 4-month pilot with 18 engineers detected 23 pre-release regressions through CI/CD-integrated regression testing, reducing median root-cause identification time from 4.2 hours to 22 minutes and driving measurable failure rate reductions in two workflows. △ Less

Submitted 26 April, 2026; originally announced April 2026.

Comments: Accepted at ACL 2026 Industry Track. 14 pages, 3 figures, 21 tables

ACM Class: I.2.7; D.2.5; D.2.4

RouteNLP: Closed-Loop LLM Routing with Conformal Cascading and Distillation Co-Optimization

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Serving diverse NLP workloads with large language models is costly: at one enterprise partner, inference costs exceeded $200K/month despite over 70% of queries being routine tasks well within the capability of smaller models. We present RouteNLP, a closed-loop framework that routes queries across a tiered model portfolio to minimize cost while satisfying per-task quality constraints. The framework… ▽ More Serving diverse NLP workloads with large language models is costly: at one enterprise partner, inference costs exceeded $200K/month despite over 70% of queries being routine tasks well within the capability of smaller models. We present RouteNLP, a closed-loop framework that routes queries across a tiered model portfolio to minimize cost while satisfying per-task quality constraints. The framework integrates three components: a difficulty-aware router with shared task-conditioned representations trained on preference data and quality signals; confidence-calibrated cascading that uses conformal prediction for distribution-free threshold initialization; and a distillation-routing co-optimization loop that clusters escalation failures, applies targeted knowledge distillation to cheaper models, and automatically retrains the router, yielding over twice the cost improvement of untargeted distillation. In an 8-week pilot deployment processing ~5K queries/day at an enterprise customer-service division, RouteNLP reduced inference costs by 58% while maintaining 91% response acceptance and reducing p99 latency from 1,847 ms to 387 ms. On a six-task benchmark spanning finance, customer service, and legal domains, the framework achieves 40-85% cost reduction while retaining 96-100% quality on structured tasks and 96-98% on generation tasks, with human evaluation confirming that 74.5% of routed generation outputs match or exceed frontier-model quality. △ Less

Submitted 26 April, 2026; originally announced April 2026.

Comments: Accepted at ACL 2026 Industry Track. 13 pages, 2 figures, 15 tables, 1 algorithm

ACM Class: I.2.7; I.2.6; H.3.3; C.4

GeoCert: Certified Geometric AI for Reliable Forecasting

Authors: Regina Zhang, Zongru Li, Honggang Wen, Xiaofeng Liu, Siu-Ming Yiu, Pietro Liò, Kwok-Yan Lam

Abstract: Forecasting systems in science must be accurate, physically consistent, and certifiably reliable. Most existing models address prediction, constraint enforcement, and verification separately, limiting scalability and interpretability. We introduce GeoCert, a geometric AI framework that unifies forecasting, physical reasoning, and formal verification within a single differentiable computation. GeoC… ▽ More Forecasting systems in science must be accurate, physically consistent, and certifiably reliable. Most existing models address prediction, constraint enforcement, and verification separately, limiting scalability and interpretability. We introduce GeoCert, a geometric AI framework that unifies forecasting, physical reasoning, and formal verification within a single differentiable computation. GeoCert formulates forecasting as evolution along a hyperbolic manifold, where negative curvature induces contraction dynamics, intrinsic robustness, and logarithmic-time certification. A hierarchical constraint architecture separates universal physical laws from domain-specific dynamics, enabling certified generalization across energy, climate, finance, and transportation systems. GeoCert achieves state-of-the-art accuracy while reducing computational cost by 97.5% and maintaining better certification rates. By embedding verification into the geometry of learning, GeoCert transforms forecasting from empirical approximation to formally verified inference, offering a scalable foundation for trustworthy, reproducible, and physically grounded scientific AI. △ Less

Submitted 25 April, 2026; originally announced April 2026.

Comments: 15 pages, 4 figures

SafeAnchor: Preventing Cumulative Safety Erosion in Continual Domain Adaptation of Large Language Models

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Safety alignment in large language models is remarkably shallow: it is concentrated in the first few output tokens and reversible by fine-tuning on as few as 100 adversarial examples. This fragility becomes critical in real-world deployment, where models undergo sequential adaptation across domains such as medicine, law, and code, causing safety guardrails to erode cumulatively. Yet all existing s… ▽ More Safety alignment in large language models is remarkably shallow: it is concentrated in the first few output tokens and reversible by fine-tuning on as few as 100 adversarial examples. This fragility becomes critical in real-world deployment, where models undergo sequential adaptation across domains such as medicine, law, and code, causing safety guardrails to erode cumulatively. Yet all existing safety-preserving methods target only single-task fine-tuning, leaving the multi-domain sequential setting entirely unaddressed. We introduce SafeAnchor, a framework that anchors safety in place throughout continual adaptation. SafeAnchor first identifies low-rank safety subspaces in LoRA parameter space via Fisher Information eigendecomposition, then constrains domain-specific gradient updates to the orthogonal complement of these subspaces, and finally monitors for residual safety drift with threshold-triggered corrective replay. Evaluated on Llama-2-7B-Chat and Mistral-7B-Instruct across a three-domain pipeline and eight benchmarks, SafeAnchor retains 93.2% of original safety alignment, outperforming all baselines by 18-42 points, while matching unconstrained fine-tuning to within 1.5 points on domain tasks. △ Less

Submitted 19 April, 2026; originally announced April 2026.

Comments: 16 pages (12 main + 4 appendix), 2 figures, 12 tables

MSC Class: 68T07; 68T50; 68T05 ACM Class: I.2.7; I.2.6; I.2.0

Capacity-Controlled Global Attention for Graph Transformers

Authors: Yang Liu, Dongxin Guo, Tom Zheng, Siu Ming Yiu, Liam Ning, Jikun Wu

Abstract: Global self-attention drives modern graph transformers, yet the softmax at its core imposes a structural constraint rarely examined directly: every attention row is non-negative and sums to one, so each per-head output is a mass-conserving convex combination of value vectors. A node can never "attend to nothing." We argue this conservation constraint is a single root cause behind three pathologies… ▽ More Global self-attention drives modern graph transformers, yet the softmax at its core imposes a structural constraint rarely examined directly: every attention row is non-negative and sums to one, so each per-head output is a mass-conserving convex combination of value vectors. A node can never "attend to nothing." We argue this conservation constraint is a single root cause behind three pathologies usually studied in isolation: the collapse of node representations with depth (over-smoothing), a low-rank bottleneck on per-head outputs, and brittle optimization in deep stacks. Drawing on how sigmoid gating removes analogous attention sinks in language models, we introduce SigGate-GT, a graph transformer that applies a learned, per-head, input-conditioned sigmoid gate to the attention output inside the GraphGPS framework. The gate is a smooth, per-dimension "volume control" that can drive head outputs toward zero, relaxing the constraint without abandoning attention's probabilistic interpretation. Analytically and through synthetic experiments, we show the gate strictly increases the stable rank of per-head outputs, and connect this rank gain to all three manifestations. On five molecular and long-range benchmarks, SigGate-GT matches the prior best on ZINC (0.059 MAE), records the strongest result among the graph-transformer baselines we evaluate on ogbg-molhiv (82.47% ROC-AUC), and is competitive on ogbg-molpcba and the Long-Range Graph Benchmark, with statistically significant gains over GraphGPS on all five datasets (p < 0.05). Mechanism analyses confirm the diagnosis: gating slows over-smoothing (a 30% mean relative gain in representation diversity across 4-16 layers), keeps attention entropy from collapsing, and stabilizes training across a 10x learning-rate range, at about 1% parameter overhead on OGB and under 3% wall-clock cost. △ Less

Submitted 7 June, 2026; v1 submitted 19 April, 2026; originally announced April 2026.

Comments: 13 pages, 2 figures, 15 tables

ACM Class: I.2.6; I.5.1; G.2.2

A Systematic Survey and Benchmark of Deep Learning for Molecular Property Prediction in the Foundation Model Era

Authors: Zongru Li, Xingsheng Chen, Honggang Wen, Regina Qianru Zhang, Ming Li, Xiaojin Zhang, Hongzhi Yin, Qiang Yang, Kwok-Yan Lam, Pietro Lio, Siu-Ming Yiu

Abstract: Molecular property prediction integrates quantum chemistry, cheminformatics, and deep learning to connect molecular structure with physicochemical and biological behavior. This survey traces four complementary paradigms, including Quantum, Descriptor Machine Learning, Geometric Deep Learning, and Foundation Models, and outlines a unified taxonomy linking molecular representations, model architectu… ▽ More Molecular property prediction integrates quantum chemistry, cheminformatics, and deep learning to connect molecular structure with physicochemical and biological behavior. This survey traces four complementary paradigms, including Quantum, Descriptor Machine Learning, Geometric Deep Learning, and Foundation Models, and outlines a unified taxonomy linking molecular representations, model architectures, and interdisciplinary applications. Benchmark analyses integrate evidence from both widely used datasets and datasets reflecting industry perspectives, encompassing quantum, physicochemical, physiological, and biophysical domains. The survey examines current standards in data curation, splitting strategies, and evaluation protocols, highlighting challenges including inconsistent stereochemistry, heterogeneous assay sources, and reproducibility limitations under random or poorly defined splits. These observations motivate the modernization of benchmark design toward more transparent, time- and scaffold-aware methodologies. We further propose three forward-looking directions: (i) physics-aware learning embedding quantum consistency, (ii) uncertainty-calibrated foundation models for trustworthy inference, and (iii) realistic multimodal benchmark ecosystems integrating computational and experimental data. Repository: https://github.com/Zongru-Li/Survey-and-Benchmarks-of-DL-for-Molecular-Property-Prediction-in-the-Foundation-Model-Era. △ Less

Submitted 17 April, 2026; originally announced April 2026.

Comments: 32 pages. It is just accepted by Journal of Chemical Theory and Computation 2026

Journal ref: Journal of Chemical Theory and Computation 2026

UniMamba: A Unified Spatial-Temporal Modeling Framework with State-Space and Attention Integration

Authors: Xingsheng Chen, Xianpei Mu, Deyu Yi, Yilin Yuan, Xingwei He, Bo Gao, Regina Zhang, Pietro Lio, Siu-Ming Yiu

Abstract: Multivariate time series forecasting is fundamental to numerous domains such as energy, finance, and environmental monitoring, where complex temporal dependencies and cross-variable interactions pose enduring challenges. Existing Transformer-based methods capture temporal correlations through attention mechanisms but suffer from quadratic computational cost, while state-space models like Mamba ach… ▽ More Multivariate time series forecasting is fundamental to numerous domains such as energy, finance, and environmental monitoring, where complex temporal dependencies and cross-variable interactions pose enduring challenges. Existing Transformer-based methods capture temporal correlations through attention mechanisms but suffer from quadratic computational cost, while state-space models like Mamba achieve efficient long-context modeling yet lack explicit temporal pattern recognition. Therefore we introduce UniMamba, a unified spatial-temporal forecasting framework that integrates efficient state-space dynamics with attention-based dependency learning. UniMamba employs a Mamba Variate-Channel Encoding Layer enhanced with FFT-Laplace Transform and TCN to capture global temporal dependencies, and a Spatial Temporal Attention Layer to jointly model inter-variate correlations and temporal evolution. A Feedforward Temporal Dynamics Layer further fuses continuous and discrete contexts for accurate forecasting. Comprehensive experiments on eight public benchmark datasets demonstrate that UniMamba consistently outperforms state-of-the-art forecasting models in both forecasting accuracy and computational efficiency, establishing a scalable and robust solution for long-sequence multivariate time-series prediction. △ Less

Submitted 27 June, 2026; v1 submitted 6 March, 2026; originally announced April 2026.

Closing the Theory-Practice Gap in Spiking Transformers via Effective Dimension

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Spiking transformers achieve competitive accuracy with conventional transformers while offering $38$-$57\times$ energy efficiency on neuromorphic hardware, yet no theoretical framework guides their design. This paper establishes the first comprehensive expressivity theory for spiking self-attention. We prove that spiking attention with Leaky Integrate-and-Fire neurons is a universal approximator o… ▽ More Spiking transformers achieve competitive accuracy with conventional transformers while offering $38$-$57\times$ energy efficiency on neuromorphic hardware, yet no theoretical framework guides their design. This paper establishes the first comprehensive expressivity theory for spiking self-attention. We prove that spiking attention with Leaky Integrate-and-Fire neurons is a universal approximator of continuous permutation-equivariant functions, providing explicit spike circuit constructions including a novel lateral inhibition network for softmax normalization with proven $O(1/\sqrt{T})$ convergence. We derive tight spike-count lower bounds via rate-distortion theory: $\varepsilon$-approximation requires $Ω(L_f^2 nd/\varepsilon^2)$ spikes, with rigorous information-theoretic derivation. Our key insight is input-dependent bounds using measured effective dimensions ($d_{\text{eff}}=47$--$89$ for CIFAR/ImageNet), explaining why $T=4$ timesteps suffice despite worst-case $T \geq 10{,}000$ predictions. We provide concrete design rules with calibrated constants ($C=2.3$, 95\% CI: $[1.9, 2.7]$). Experiments on Spikformer, QKFormer, and SpikingResformer across vision and language benchmarks validate predictions with $R^2=0.97$ ($p<0.001$). Our framework provides the first principled foundation for neuromorphic transformer design. △ Less

Submitted 17 April, 2026; originally announced April 2026.

Comments: 6 pages, 3 figures, 7 tables

MSC Class: 68T07; 92B20 ACM Class: I.2.6; I.5.1

When Do Early-Exit Networks Generalize? A PAC-Bayesian Theory of Adaptive Depth

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Early-exit neural networks enable adaptive computation by allowing confident predictions to exit at intermediate layers, achieving 2-8$\times$ inference speedup. Despite widespread deployment, their generalization properties lack theoretical understanding -- a gap explicitly identified in recent surveys. This paper establishes a unified PAC-Bayesian framework for adaptive-depth networks. (1) Novel… ▽ More Early-exit neural networks enable adaptive computation by allowing confident predictions to exit at intermediate layers, achieving 2-8$\times$ inference speedup. Despite widespread deployment, their generalization properties lack theoretical understanding -- a gap explicitly identified in recent surveys. This paper establishes a unified PAC-Bayesian framework for adaptive-depth networks. (1) Novel Entropy-Based Bounds: We prove the first generalization bounds depending on exit-depth entropy $H(D)$ and expected depth $\mathbb{E}[D]$ rather than maximum depth $K$, with sample complexity $\mathcal{O}((\mathbb{E}[D] \cdot d + H(D))/ε^2)$. (2) Explicit Constructive Constants: Our analysis yields the leading coefficient $\sqrt{2\ln 2} \approx 1.177$ with complete derivation. (3) Provable Early-Exit Advantages: We establish sufficient conditions under which adaptive-depth networks strictly outperform fixed-depth counterparts. (4) Extension to Approximate Label Independence: We relax the label-independence assumption to $ε$-approximate policies, broadening applicability to learned routing. (5) Comprehensive Validation: Experiments across 6 architectures on 7 benchmarks demonstrate tightness ratios of 1.52-3.87$\times$ (all $p < 0.001$) versus $>$100$\times$ for classical bounds. Bound-guided threshold selection matches validation-tuned performance within 0.1-0.3%. △ Less

Submitted 17 April, 2026; originally announced April 2026.

Comments: 6 pages, 1 figure, 7 tables, 1 algorithm

MSC Class: 68T07; 62C12 ACM Class: I.2.6; F.2.2

Geometric Metrics for MoE Specialization: From Fisher Information to Early Failure Detection

Authors: Dongxin Guo, Jikun Wu, Siu Ming Yiu

Abstract: Expert specialization is fundamental to Mixture-of-Experts (MoE) model success, yet existing metrics (cosine similarity, routing entropy) lack theoretical grounding and yield inconsistent conclusions under reparameterization. We present an information-geometric framework providing the first rigorous characterization of MoE specialization dynamics. Our key insight is that expert routing distributio… ▽ More Expert specialization is fundamental to Mixture-of-Experts (MoE) model success, yet existing metrics (cosine similarity, routing entropy) lack theoretical grounding and yield inconsistent conclusions under reparameterization. We present an information-geometric framework providing the first rigorous characterization of MoE specialization dynamics. Our key insight is that expert routing distributions evolve on the probability simplex equipped with the Fisher information metric, enabling formal analysis via Riemannian geometry. We prove that standard heuristic metrics violate parameterization invariance (Theorem 1), establish that specialization corresponds to geodesic flow with quantified approximation bounds (Theorem 2), and derive a failure predictor with theoretical threshold justification (Theorem 3). The framework introduces two principled metrics: Fisher Specialization Index (FSI) achieving r=0.91+/-0.02 correlation with downstream performance, and Fisher Heterogeneity Score (FHS) predicting training failure at 10% completion with AUC=0.89+/-0.03 -- outperforming validation-loss-based early stopping by 23% while requiring 40x fewer compute cycles. We validate intervention protocols achieving 87% recovery rate when FHS>1 is detected. Comprehensive experiments across language modeling (WikiText-103, C4), vision MoE (ImageNet), and scaling studies (8-64 experts, 125M-2.7B parameters) validate our theoretical predictions. △ Less

Submitted 15 April, 2026; originally announced April 2026.

Comments: 6 pages, 2 figures, 7 tables

MSC Class: 62B10; 68T07 ACM Class: I.2.6; G.3

Coalition Formation in LLM Agent Networks: Stability Analysis and Convergence Guarantees

Authors: Dongxin Guo, Jikun Wu, Siu-Ming Yiu

Abstract: Large Language Model (LLM) agents are increasingly deployed in multi-agent systems requiring strategic coordination. While recent work has analyzed LLM behavior in two-player games, coalition formation, where $n$ agents dynamically form cooperative groups, remains theoretically uncharacterized. We present the first framework grounding coalition formation in LLM agent networks in hedonic game theor… ▽ More Large Language Model (LLM) agents are increasingly deployed in multi-agent systems requiring strategic coordination. While recent work has analyzed LLM behavior in two-player games, coalition formation, where $n$ agents dynamically form cooperative groups, remains theoretically uncharacterized. We present the first framework grounding coalition formation in LLM agent networks in hedonic game theory with formal stability guarantees. We introduce the LLM Coalition Formation Game (LCFG), establish sufficient conditions for Nash-stable partitions, and prove complexity results. Our analysis reveals that LLM agents exhibit bounded rationality characterized by $ε$-rational preferences; we provide both deterministic existence guarantees and consistency-driven stability bounds whose predictions are consistent with empirical outcomes. Experiments with GPT-4, Claude-3, and Llama-3 across 2,400 episodes validate our framework: LLM coalitions achieve Nash stability in 73.2% of cases under our Coalition-of-Thought (CoalT) protocol, compared to 58.4% under chain-of-thought and 41.8% under standard prompting ($p < 0.001$). Our framework provides theoretical foundations for designing stable multi-agent LLM systems. △ Less

Submitted 15 April, 2026; originally announced April 2026.

Comments: 15 pages including supplementary material, 2 figures, 5 tables

ACM Class: I.2.11; J.4

UniBioTransfer: A Unified Framework for Multiple Biometrics Transfer

Authors: Caiyi Sun, Yujing Sun, Xiangyu Li, Yuhang Zheng, Yiming Ren, Jiamin Wang, Yuexin Ma, Siu-Ming Yiu

Abstract: Deepface generation has traditionally followed a task-driven paradigm, where distinct tasks (e.g., face transfer and hair transfer) are addressed by task-specific models. Nevertheless, this single-task setting severely limits model generalization and scalability. A unified model capable of solving multiple deepface generation tasks in a single pass represents a promising and practical direction, y… ▽ More Deepface generation has traditionally followed a task-driven paradigm, where distinct tasks (e.g., face transfer and hair transfer) are addressed by task-specific models. Nevertheless, this single-task setting severely limits model generalization and scalability. A unified model capable of solving multiple deepface generation tasks in a single pass represents a promising and practical direction, yet remains challenging due to data scarcity and cross-task conflicts arising from heterogeneous attribute transformations. To this end, we propose UniBioTransfer, the first unified framework capable of handling both conventional deepface tasks (e.g., face transfer and face reenactment) and shape-varying transformations (e.g., hair transfer and head transfer). Besides, UniBioTransfer naturally generalizes to unseen tasks, like lip, eye, and glasses transfer, with minimal fine-tuning. Generally, UniBioTransfer addresses data insufficiency in multi-task generation through a unified data construction strategy, including a swapping-based corruption mechanism designed for spatially dynamic attributes like hair. It further mitigates cross-task interference via an innovative BioMoE, a mixture-of-experts based model coupled with a novel two-stage training strategy that effectively disentangles task-specific knowledge. Extensive experiments demonstrate the effectiveness, generalization, and scalability of UniBioTransfer, outperforming both existing unified models and task-specific methods across a wide range of deepface generation tasks. Project page is at https://scy639.github.io/UniBioTransfer.github.io/ △ Less

Submitted 20 March, 2026; originally announced March 2026.

ASDA: Automated Skill Distillation and Adaptation for Financial Reasoning

Authors: Tik Yu Yim, Wenting Tan, Sum Yee Chan, Tak-Wah Lam, Siu Ming Yiu

Abstract: Adapting large language models (LLMs) to specialized financial reasoning typically requires expensive fine-tuning that produces model-locked expertise. Training-free alternatives have emerged, yet our experiments show that leading methods (GEPA and ACE) achieve only marginal gains on the FAMMA financial reasoning benchmark, exposing the limits of unstructured text optimization for complex, multi-s… ▽ More Adapting large language models (LLMs) to specialized financial reasoning typically requires expensive fine-tuning that produces model-locked expertise. Training-free alternatives have emerged, yet our experiments show that leading methods (GEPA and ACE) achieve only marginal gains on the FAMMA financial reasoning benchmark, exposing the limits of unstructured text optimization for complex, multi-step domain reasoning. We introduce Automated Skill Distillation and Adaptation (ASDA), a framework that automatically generates structured skill artifacts through iterative error-corrective learning without modifying model weights. A teacher model analyzes a student model's failures on financial reasoning tasks, clusters errors by subfield and error type, and synthesizes skill files containing reasoning procedures, code templates, and worked examples, which are dynamically injected during inference. Evaluated on FAMMA, ASDA achieves up to +17.33% improvement on arithmetic reasoning and +5.95% on non-arithmetic reasoning, substantially outperforming all training-free baselines. The resulting skill artifacts are human-readable, version-controlled, and compatible with the Agent Skills open standard, offering any organization with a labeled domain dataset a practical and auditable path to domain adaptation without weight access or retraining. △ Less

Submitted 17 March, 2026; originally announced March 2026.

Driving Like Yourself: A Benchmark for Closed-Loop Personalized End-to-End Autonomous Driving

Authors: Xiaoru Dong, Ruiqin Li, Xiao Han, Zhenxuan Wu, Jiamin Wang, Jian Chen, Qi Jiang, SM Yiu, Xinge Zhu, Yuexin Ma

Abstract: Human driving behavior is inherently diverse, yet most end-to-end autonomous driving (E2E-AD) systems learn a single average driving style, neglecting individual differences. Achieving personalized E2E-AD faces challenges across three levels: limited real-world datasets with individual-level annotations, a lack of quantitative metrics for evaluating personal driving styles, and the absence of algo… ▽ More Human driving behavior is inherently diverse, yet most end-to-end autonomous driving (E2E-AD) systems learn a single average driving style, neglecting individual differences. Achieving personalized E2E-AD faces challenges across three levels: limited real-world datasets with individual-level annotations, a lack of quantitative metrics for evaluating personal driving styles, and the absence of algorithms that can learn stylized representations from users' trajectories. To address these gaps, we propose Person2Drive, a comprehensive personalized E2E-AD platform and benchmark. It includes an open-source, flexible data collection system that simulates realistic scenarios to generate scalable, diverse personalized driving datasets; style vector-based evaluation metrics with Maximum Mean Discrepancy and KL divergence to comprehensively quantify individual driving behaviors; and a personalized E2E-AD framework with a style reward model that efficiently adapts E2E models for safe and individualized driving. Crucially, our framework enables plug-and-play personalization by fine-tuning only the trajectory prediction head, preserving the pretrained base model and ensuring safety. Extensive experiments demonstrate that Person2Drive enables fine-grained analysis and effective personalization, while preserving driving performance and success rate even in challenging scenarios. △ Less

Submitted 13 July, 2026; v1 submitted 21 February, 2026; originally announced February 2026.

Comments: Accepted to ECCV 2026. Camera-ready version

MODE: Efficient Time Series Prediction with Mamba Enhanced by Low-Rank Neural ODEs

Authors: Xingsheng Chen, Regina Zhang, Bo Gao, Xingwei He, Xiaofeng Liu, Pietro Lio, Kwok-Yan Lam, Siu-Ming Yiu

Abstract: Time series prediction plays a pivotal role across diverse domains such as finance, healthcare, energy systems, and environmental modeling. However, existing approaches often struggle to balance efficiency, scalability, and accuracy, particularly when handling long-range dependencies and irregularly sampled data. To address these challenges, we propose MODE, a unified framework that integrates Low… ▽ More Time series prediction plays a pivotal role across diverse domains such as finance, healthcare, energy systems, and environmental modeling. However, existing approaches often struggle to balance efficiency, scalability, and accuracy, particularly when handling long-range dependencies and irregularly sampled data. To address these challenges, we propose MODE, a unified framework that integrates Low-Rank Neural Ordinary Differential Equations (Neural ODEs) with an Enhanced Mamba architecture. As illustrated in our framework, the input sequence is first transformed by a Linear Tokenization Layer and then processed through multiple Mamba Encoder blocks, each equipped with an Enhanced Mamba Layer that employs Causal Convolution, SiLU activation, and a Low-Rank Neural ODE enhancement to efficiently capture temporal dynamics. This low-rank formulation reduces computational overhead while maintaining expressive power. Furthermore, a segmented selective scanning mechanism, inspired by pseudo-ODE dynamics, adaptively focuses on salient subsequences to improve scalability and long-range sequence modeling. Extensive experiments on benchmark datasets demonstrate that MODE surpasses existing baselines in both predictive accuracy and computational efficiency. Overall, our contributions include: (1) a unified and efficient architecture for long-term time series modeling, (2) integration of Mamba's selective scanning with low-rank Neural ODEs for enhanced temporal representation, and (3) substantial improvements in efficiency and scalability enabled by low-rank approximation and dynamic selective scanning. △ Less

Submitted 11 January, 2026; v1 submitted 1 January, 2026; originally announced January 2026.

Comments: 12 pages, 6 figures, and 3 tables. Updated description and explanations, and correct some typos

ConInstruct: Evaluating Large Language Models on Conflict Detection and Resolution in Instructions

Authors: Xingwei He, Qianru Zhang, Pengfei Chen, Guanhua Chen, Linlin Yu, Yuan Yuan, Siu-Ming Yiu

Abstract: Instruction-following is a critical capability of Large Language Models (LLMs). While existing works primarily focus on assessing how well LLMs adhere to user instructions, they often overlook scenarios where instructions contain conflicting constraints-a common occurrence in complex prompts. The behavior of LLMs under such conditions remains under-explored. To bridge this gap, we introduce ConIns… ▽ More Instruction-following is a critical capability of Large Language Models (LLMs). While existing works primarily focus on assessing how well LLMs adhere to user instructions, they often overlook scenarios where instructions contain conflicting constraints-a common occurrence in complex prompts. The behavior of LLMs under such conditions remains under-explored. To bridge this gap, we introduce ConInstruct, a benchmark specifically designed to assess LLMs' ability to detect and resolve conflicts within user instructions. Using this dataset, we evaluate LLMs' conflict detection performance and analyze their conflict resolution behavior. Our experiments reveal two key findings: (1) Most proprietary LLMs exhibit strong conflict detection capabilities, whereas among open-source models, only DeepSeek-R1 demonstrates similarly strong performance. DeepSeek-R1 and Claude-4.5-Sonnet achieve the highest average F1-scores at 91.5% and 87.3%, respectively, ranking first and second overall. (2) Despite their strong conflict detection abilities, LLMs rarely explicitly notify users about the conflicts or request clarification when faced with conflicting constraints. These results underscore a critical shortcoming in current LLMs and highlight an important area for future improvement when designing instruction-following LLMs. △ Less

Submitted 19 November, 2025; v1 submitted 18 November, 2025; originally announced November 2025.

Comments: Accepted to AAAI 2026

A Comprehensive Survey on Benchmarks and Solutions in Software Engineering of LLM-Empowered Agentic System

Authors: Jiale Guo, Suizhi Huang, Mei Li, Dong Huang, Xingsheng Chen, Regina Zhang, Zhijiang Guo, Han Yu, Siu-Ming Yiu, Pietro Lio, Kwok-Yan Lam

Abstract: The integration of Large Language Models (LLMs) into software engineering has driven a transition from traditional rule-based systems to autonomous agentic systems capable of solving complex problems. However, systematic progress is hindered by a lack of comprehensive understanding of how benchmarks and solutions interconnect. This survey addresses this gap by providing the first holistic analysis… ▽ More The integration of Large Language Models (LLMs) into software engineering has driven a transition from traditional rule-based systems to autonomous agentic systems capable of solving complex problems. However, systematic progress is hindered by a lack of comprehensive understanding of how benchmarks and solutions interconnect. This survey addresses this gap by providing the first holistic analysis of LLM-powered software engineering, offering insights into evaluation methodologies and solution paradigms. We review over 150 recent papers and propose a taxonomy along two key dimensions: (1) Solutions, categorized into prompt-based, fine-tuning-based, and agent-based paradigms, and (2) Benchmarks, including tasks such as code generation, translation, and repair. Our analysis highlights the evolution from simple prompt engineering to sophisticated agentic systems incorporating capabilities like planning, reasoning, memory mechanisms, and tool augmentation. To contextualize this progress, we present a unified pipeline illustrating the workflow from task specification to deliverables, detailing how different solution paradigms address various complexity levels. Unlike prior surveys that focus narrowly on specific aspects, this work connects 50+ benchmarks to their corresponding solution strategies, enabling researchers to identify optimal approaches for diverse evaluation criteria. We also identify critical research gaps and propose future directions, including multi-agent collaboration, self-evolving systems, and formal verification integration. This survey serves as a foundational guide for advancing LLM-driven software engineering. We maintain a GitHub repository that continuously updates the reviewed and related papers at https://github.com/lisaGuojl/LLM-Agent-SE-Survey. △ Less

Submitted 23 October, 2025; v1 submitted 10 October, 2025; originally announced October 2025.

Comments: 22 pages

FLDmamba: Integrating Fourier and Laplace Transform Decomposition with Mamba for Enhanced Time Series Prediction

Authors: Qianru Zhang, Chenglei Yu, Haixin Wang, Yudong Yan, Yuansheng Cao, Siu-Ming Yiu, Tailin Wu, Hongzhi Yin

Abstract: Time series prediction, a crucial task across various domains, faces significant challenges due to the inherent complexities of time series data, including non-stationarity, multi-scale periodicity, and transient dynamics, particularly when tackling long-term predictions. While Transformer-based architectures have shown promise, their quadratic complexity with sequence length hinders their efficie… ▽ More Time series prediction, a crucial task across various domains, faces significant challenges due to the inherent complexities of time series data, including non-stationarity, multi-scale periodicity, and transient dynamics, particularly when tackling long-term predictions. While Transformer-based architectures have shown promise, their quadratic complexity with sequence length hinders their efficiency for long-term predictions. Recent advancements in State-Space Models, such as Mamba, offer a more efficient alternative for long-term modeling, but they cannot capture multi-scale periodicity and transient dynamics effectively. Meanwhile, they are susceptible to data noise issues in time series. This paper proposes a novel framework, FLDmamba (Fourier and Laplace Transform Decomposition Mamba), addressing these limitations. FLDmamba leverages the strengths of both Fourier and Laplace transforms to effectively capture both multi-scale periodicity, transient dynamics within time series data, and improve the robustness of the model to the data noise issue. Our extensive experiments demonstrate that FLDmamba achieves superior performance on time series prediction benchmarks, outperforming both Transformer-based and other Mamba-based architectures. To promote the reproducibility of our method, we have made both the code and data accessible via the following URL:{\href{https://github.com/AI4Science-WestlakeU/FLDmamba}{https://github.com/AI4Science-WestlakeU/\model}. △ Less

Submitted 17 July, 2025; originally announced July 2025.

Comments: 12 pages

AutoHFormer: Efficient Hierarchical Autoregressive Transformer for Time Series Prediction

Authors: Qianru Zhang, Honggang Wen, Ming Li, Dong Huang, Siu-Ming Yiu, Christian S. Jensen, Pietro Liò

Abstract: Time series forecasting requires architectures that simultaneously achieve three competing objectives: (1) strict temporal causality for reliable predictions, (2) sub-quadratic complexity for practical scalability, and (3) multi-scale pattern recognition for accurate long-horizon forecasting. We introduce AutoHFormer, a hierarchical autoregressive transformer that addresses these challenges throug… ▽ More Time series forecasting requires architectures that simultaneously achieve three competing objectives: (1) strict temporal causality for reliable predictions, (2) sub-quadratic complexity for practical scalability, and (3) multi-scale pattern recognition for accurate long-horizon forecasting. We introduce AutoHFormer, a hierarchical autoregressive transformer that addresses these challenges through three key innovations: 1) Hierarchical Temporal Modeling: Our architecture decomposes predictions into segment-level blocks processed in parallel, followed by intra-segment sequential refinement. This dual-scale approach maintains temporal coherence while enabling efficient computation. 2) Dynamic Windowed Attention: The attention mechanism employs learnable causal windows with exponential decay, reducing complexity while preserving precise temporal relationships. This design avoids both the anti-causal violations of standard transformers and the sequential bottlenecks of RNN hybrids. 3) Adaptive Temporal Encoding: a novel position encoding system is adopted to capture time patterns at multiple scales. It combines fixed oscillating patterns for short-term variations with learnable decay rates for long-term trends. Comprehensive experiments demonstrate that AutoHFormer 10.76X faster training and 6.06X memory reduction compared to PatchTST on PEMS08, while maintaining consistent accuracy across 96-720 step horizons in most of cases. These breakthroughs establish new benchmarks for efficient and precise time series modeling. Implementations of our method and all baselines in hierarchical autoregressive mechanism are available at https://github.com/lizzyhku/Autotime. △ Less

Submitted 22 November, 2025; v1 submitted 18 June, 2025; originally announced June 2025.

Comments: 14 pages

Journal ref: ICDE'2026

HUMOF: Human Motion Forecasting in Interactive Social Scenes

Authors: Caiyi Sun, Yujing Sun, Xiao Han, Zemin Yang, Jiawei Liu, Xinge Zhu, Siu Ming Yiu, Yuexin Ma

Abstract: Complex scenes present significant challenges for predicting human behaviour due to the abundance of interaction information, such as human-human and humanenvironment interactions. These factors complicate the analysis and understanding of human behaviour, thereby increasing the uncertainty in forecasting human motions. Existing motion prediction methods thus struggle in these complex scenarios. I… ▽ More Complex scenes present significant challenges for predicting human behaviour due to the abundance of interaction information, such as human-human and humanenvironment interactions. These factors complicate the analysis and understanding of human behaviour, thereby increasing the uncertainty in forecasting human motions. Existing motion prediction methods thus struggle in these complex scenarios. In this paper, we propose an effective method for human motion forecasting in interactive scenes. To achieve a comprehensive representation of interactions, we design a hierarchical interaction feature representation so that high-level features capture the overall context of the interactions, while low-level features focus on fine-grained details. Besides, we propose a coarse-to-fine interaction reasoning module that leverages both spatial and frequency perspectives to efficiently utilize hierarchical features, thereby enhancing the accuracy of motion predictions. Our method achieves state-of-the-art performance across four public datasets. The source code will be available at https://github.com/scy639/HUMOF. △ Less

Submitted 1 April, 2026; v1 submitted 4 June, 2025; originally announced June 2025.

Comments: Accepted by ICLR 2026

EffiBench-X: A Multi-Language Benchmark for Measuring Efficiency of LLM-Generated Code

Authors: Yuhao Qing, Boyu Zhu, Mingzhe Du, Zhijiang Guo, Terry Yue Zhuo, Qianru Zhang, Jie M. Zhang, Heming Cui, Siu-Ming Yiu, Dong Huang, See-Kiong Ng, Luu Anh Tuan

Abstract: Existing code generation benchmarks primarily evaluate functional correctness, with limited focus on code efficiency and often restricted to a single language like Python. To address this gap, we introduce EffiBench-X, the first multi-language benchmark designed to measure the efficiency of LLM-generated code. EffiBench-X supports Python, C++, Java, JavaScript, Ruby, and Golang. It comprises compe… ▽ More Existing code generation benchmarks primarily evaluate functional correctness, with limited focus on code efficiency and often restricted to a single language like Python. To address this gap, we introduce EffiBench-X, the first multi-language benchmark designed to measure the efficiency of LLM-generated code. EffiBench-X supports Python, C++, Java, JavaScript, Ruby, and Golang. It comprises competitive programming tasks with human-expert solutions as efficiency baselines. Evaluating state-of-the-art LLMs on EffiBench-X reveals that while models generate functionally correct code, they consistently underperform human experts in efficiency. Even the most efficient LLM-generated solutions (Qwen3-32B) achieve only around \textbf{62\%} of human efficiency on average, with significant language-specific variations. LLMs show better efficiency in Python, Ruby, and JavaScript than in Java, C++, and Golang. For instance, DeepSeek-R1's Python code is significantly more efficient than its Java code. These results highlight the critical need for research into LLM optimization techniques to improve code efficiency across diverse languages. The dataset and evaluation infrastructure are submitted and available at https://github.com/EffiBench/EffiBench-X.git and https://huggingface.co/datasets/EffiBench/effibench-x. △ Less

Submitted 19 May, 2025; originally announced May 2025.

Comments: Under Review

HMamba: Hyperbolic Mamba for Sequential Recommendation

Authors: Qianru Zhang, Honggang Wen, Wei Yuan, Crystal Chen, Menglin Yang, Siu-Ming Yiu, Hongzhi Yin

Abstract: Sequential recommendation systems have become a cornerstone of personalized services, adept at modeling the temporal evolution of user preferences by capturing dynamic interaction sequences. Existing approaches predominantly rely on traditional models, including RNNs and Transformers. Despite their success in local pattern recognition, Transformer-based methods suffer from quadratic computational… ▽ More Sequential recommendation systems have become a cornerstone of personalized services, adept at modeling the temporal evolution of user preferences by capturing dynamic interaction sequences. Existing approaches predominantly rely on traditional models, including RNNs and Transformers. Despite their success in local pattern recognition, Transformer-based methods suffer from quadratic computational complexity and a tendency toward superficial attention patterns, limiting their ability to infer enduring preference hierarchies in sequential recommendation data. Recent advances in Mamba-based sequential models introduce linear-time efficiency but remain constrained by Euclidean geometry, failing to leverage the intrinsic hyperbolic structure of recommendation data. To bridge this gap, we propose Hyperbolic Mamba, a novel architecture that unifies the efficiency of Mamba's selective state space mechanism with hyperbolic geometry's hierarchical representational power. Our framework introduces (1) a hyperbolic selective state space that maintains curvature-aware sequence modeling and (2) stabilized Riemannian operations to enable scalable training. Experiments across four benchmarks demonstrate that Hyperbolic Mamba achieves 3-11% improvement while retaining Mamba's linear-time efficiency, enabling real-world deployment. This work establishes a new paradigm for efficient, hierarchy-aware sequential modeling. △ Less

Submitted 14 May, 2025; originally announced May 2025.

M2Rec: Multi-scale Mamba for Efficient Sequential Recommendation

Authors: Qianru Zhang, Liang Qu, Honggang Wen, Dong Huang, Siu-Ming Yiu, Nguyen Quoc Viet Hung, Hongzhi Yin

Abstract: Sequential recommendation systems aim to predict users' next preferences based on their interaction histories, but existing approaches face critical limitations in efficiency and multi-scale pattern recognition. While Transformer-based methods struggle with quadratic computational complexity, recent Mamba-based models improve efficiency but fail to capture periodic user behaviors, leverage rich se… ▽ More Sequential recommendation systems aim to predict users' next preferences based on their interaction histories, but existing approaches face critical limitations in efficiency and multi-scale pattern recognition. While Transformer-based methods struggle with quadratic computational complexity, recent Mamba-based models improve efficiency but fail to capture periodic user behaviors, leverage rich semantic information, or effectively fuse multimodal features. To address these challenges, we propose \model, a novel sequential recommendation framework that integrates multi-scale Mamba with Fourier analysis, Large Language Models (LLMs), and adaptive gating. First, we enhance Mamba with Fast Fourier Transform (FFT) to explicitly model periodic patterns in the frequency domain, separating meaningful trends from noise. Second, we incorporate LLM-based text embeddings to enrich sparse interaction data with semantic context from item descriptions. Finally, we introduce a learnable gate mechanism to dynamically balance temporal (Mamba), frequency (FFT), and semantic (LLM) features, ensuring harmonious multimodal fusion. Extensive experiments demonstrate that \model\ achieves state-of-the-art performance, improving Hit Rate@10 by 3.2\% over existing Mamba-based models while maintaining 20\% faster inference than Transformer baselines. Our results highlight the effectiveness of combining frequency analysis, semantic understanding, and adaptive fusion for sequential recommendation. Code and datasets are available at: https://anonymous.4open.science/r/M2Rec. △ Less

Submitted 7 May, 2025; originally announced May 2025.

Time Series Analysis in Frequency Domain: A Survey of Open Challenges, Opportunities and Benchmarks

Authors: Qianru Zhang, Yuting Sun, Honggang Wen, Peng Yang, Xinzhu Li, Ming Li, Kwok-Yan Lam, Siu-Ming Yiu, Hongzhi Yin

Abstract: Frequency-domain analysis has emerged as a powerful paradigm for time series analysis, offering unique advantages over traditional time-domain approaches while introducing new theoretical and practical challenges. This survey provides a comprehensive examination of spectral methods from classical Fourier analysis to modern neural operators, systematically summarizing three open challenges in curre… ▽ More Frequency-domain analysis has emerged as a powerful paradigm for time series analysis, offering unique advantages over traditional time-domain approaches while introducing new theoretical and practical challenges. This survey provides a comprehensive examination of spectral methods from classical Fourier analysis to modern neural operators, systematically summarizing three open challenges in current research: (1) causal structure preservation during spectral transformations, (2) uncertainty quantification in learned frequency representations, and (3) topology-aware analysis for non-Euclidean data structures. Through rigorous reviewing of over 100 studies, we develop a unified taxonomy that bridges conventional spectral techniques with cutting-edge machine learning approaches, while establishing standardized benchmarks for performance evaluation. Our work identifies key knowledge gaps in the field, particularly in geometric deep learning and quantum-enhanced spectral analysis. The survey offers practitioners a systematic framework for method selection and implementation, while charting promising directions for future research in this rapidly evolving domain. △ Less

Submitted 23 September, 2025; v1 submitted 11 February, 2025; originally announced April 2025.

Comments: 35 pages

Better Bill GPT: Comparing Large Language Models against Legal Invoice Reviewers

Authors: Nick Whitehouse, Nicole Lincoln, Stephanie Yiu, Lizzie Catterson, Rivindu Perera

Abstract: Legal invoice review is a costly, inconsistent, and time-consuming process, traditionally performed by Legal Operations, Lawyers or Billing Specialists who scrutinise billing compliance line by line. This study presents the first empirical comparison of Large Language Models (LLMs) against human invoice reviewers - Early-Career Lawyers, Experienced Lawyers, and Legal Operations Professionals-asses… ▽ More Legal invoice review is a costly, inconsistent, and time-consuming process, traditionally performed by Legal Operations, Lawyers or Billing Specialists who scrutinise billing compliance line by line. This study presents the first empirical comparison of Large Language Models (LLMs) against human invoice reviewers - Early-Career Lawyers, Experienced Lawyers, and Legal Operations Professionals-assessing their accuracy, speed, and cost-effectiveness. Benchmarking state-of-the-art LLMs against a ground truth set by expert legal professionals, our empirically substantiated findings reveal that LLMs decisively outperform humans across every metric. In invoice approval decisions, LLMs achieve up to 92% accuracy, surpassing the 72% ceiling set by experienced lawyers. On a granular level, LLMs dominate line-item classification, with top models reaching F-scores of 81%, compared to just 43% for the best-performing human group. Speed comparisons are even more striking - while lawyers take 194 to 316 seconds per invoice, LLMs are capable of completing reviews in as fast as 3.6 seconds. And cost? AI slashes review expenses by 99.97%, reducing invoice processing costs from an average of $4.27 per invoice for human invoice reviewers to mere cents. These results highlight the evolving role of AI in legal spend management. As law firms and corporate legal departments struggle with inefficiencies, this study signals a seismic shift: The era of LLM-powered legal spend management is not on the horizon, it has arrived. The challenge ahead is not whether AI can perform as well as human reviewers, but how legal teams will strategically incorporate it, balancing automation with human discretion. △ Less

Submitted 2 April, 2025; originally announced April 2025.

UniDemoiré: Towards Universal Image Demoiréing with Data Generation and Synthesis

Authors: Zemin Yang, Yujing Sun, Xidong Peng, Siu Ming Yiu, Yuexin Ma

Abstract: Image demoiréing poses one of the most formidable challenges in image restoration, primarily due to the unpredictable and anisotropic nature of moiré patterns. Limited by the quantity and diversity of training data, current methods tend to overfit to a single moiré domain, resulting in performance degradation for new domains and restricting their robustness in real-world applications. In this pape… ▽ More Image demoiréing poses one of the most formidable challenges in image restoration, primarily due to the unpredictable and anisotropic nature of moiré patterns. Limited by the quantity and diversity of training data, current methods tend to overfit to a single moiré domain, resulting in performance degradation for new domains and restricting their robustness in real-world applications. In this paper, we propose a universal image demoiréing solution, UniDemoiré, which has superior generalization capability. Notably, we propose innovative and effective data generation and synthesis methods that can automatically provide vast high-quality moiré images to train a universal demoiréing model. Our extensive experiments demonstrate the cutting-edge performance and broad potential of our approach for generalized image demoiréing. △ Less

Submitted 10 February, 2025; originally announced February 2025.

Comments: Accepted by AAAI 2025

An Attack-Driven Incident Response and Defense System (ADIRDS)

Authors: Anthony Cheuk Tung Lai, Siu Ming Yiu, Ping Fan Ke, Alan Ho

Abstract: One of the major goals of incident response is to help an organization or a system owner to quickly identify and halt the attacks to minimize the damages (and financial loss) to the system being attacked. Typical incident responses rely very much on the log information captured by the system during the attacks and if needed, may need to isolate the victim from the network to avoid further destruct… ▽ More One of the major goals of incident response is to help an organization or a system owner to quickly identify and halt the attacks to minimize the damages (and financial loss) to the system being attacked. Typical incident responses rely very much on the log information captured by the system during the attacks and if needed, may need to isolate the victim from the network to avoid further destructive attacks. However, there are real cases that there are insufficient log records/information for the incident response team to identify the attacks and their origins while the attacked system cannot be stopped due to service requirements (zero downtime online systems) such as online gaming sites. Typical incident response procedures and industrial standards do not provide an adequate solution to address this scenario. In this paper, being motivated by a real case, we propose a solution, called "Attack-Driven Incident Response and Defense System (ADIRDS)" to tackle this problem. ADIRDS is an online monitoring system to run with the real system. By modeling the real system as a graph, critical nodes/assets of the system are closely monitored. Instead of relying on the original logging system, evidence will be collected from the attack technique perspectives. To migrate the risks, realistic honeypots with very similar business context as the real system are deployed to trap the attackers. We successfully apply this system to a real case. Based on our experiments, we verify that our new approach of designing the realistic honeypots is effective, 38 unique attacker's IP addresses were captured. We also compare the performance of our realistic honey with both low and high interactive honeypots proposed in the literature, the results found that our proposed honeypot can successfully cheat the attackers to attack our honeypot, which verifies that our honeypot is more effective. △ Less

Submitted 4 February, 2025; originally announced February 2025.

Comments: 18 pages, 3 figures, 4 tables

Efficient Traffic Prediction Through Spatio-Temporal Distillation

Authors: Qianru Zhang, Xinyi Gao, Haixin Wang, Siu-Ming Yiu, Hongzhi Yin

Abstract: Graph neural networks (GNNs) have gained considerable attention in recent years for traffic flow prediction due to their ability to learn spatio-temporal pattern representations through a graph-based message-passing framework. Although GNNs have shown great promise in handling traffic datasets, their deployment in real-life applications has been hindered by scalability constraints arising from hig… ▽ More Graph neural networks (GNNs) have gained considerable attention in recent years for traffic flow prediction due to their ability to learn spatio-temporal pattern representations through a graph-based message-passing framework. Although GNNs have shown great promise in handling traffic datasets, their deployment in real-life applications has been hindered by scalability constraints arising from high-order message passing. Additionally, the over-smoothing problem of GNNs may lead to indistinguishable region representations as the number of layers increases, resulting in performance degradation. To address these challenges, we propose a new knowledge distillation paradigm termed LightST that transfers spatial and temporal knowledge from a high-capacity teacher to a lightweight student. Specifically, we introduce a spatio-temporal knowledge distillation framework that helps student MLPs capture graph-structured global spatio-temporal patterns while alleviating the over-smoothing effect with adaptive knowledge distillation. Extensive experiments verify that LightST significantly speeds up traffic flow predictions by 5X to 40X compared to state-of-the-art spatio-temporal GNNs, all while maintaining superior accuracy. △ Less

Submitted 11 March, 2025; v1 submitted 14 January, 2025; originally announced January 2025.

Comments: 9 pages

Journal ref: AAAI'2025

Generalizable Single-view Object Pose Estimation by Two-side Generating and Matching

Authors: Yujing Sun, Caiyi Sun, Yuan Liu, Yuexin Ma, Siu Ming Yiu

Abstract: In this paper, we present a novel generalizable object pose estimation method to determine the object pose using only one RGB image. Unlike traditional approaches that rely on instance-level object pose estimation and necessitate extensive training data, our method offers generalization to unseen objects without extensive training, operates with a single reference image of the object, and eliminat… ▽ More In this paper, we present a novel generalizable object pose estimation method to determine the object pose using only one RGB image. Unlike traditional approaches that rely on instance-level object pose estimation and necessitate extensive training data, our method offers generalization to unseen objects without extensive training, operates with a single reference image of the object, and eliminates the need for 3D object models or multiple views of the object. These characteristics are achieved by utilizing a diffusion model to generate novel-view images and conducting a two-sided matching on these generated images. Quantitative experiments demonstrate the superiority of our method over existing pose estimation techniques across both synthetic and real-world datasets. Remarkably, our approach maintains strong performance even in scenarios with significant viewpoint changes, highlighting its robustness and versatility in challenging conditions. The code will be re leased at https://github.com/scy639/Gen2SM. △ Less

Submitted 24 November, 2024; originally announced November 2024.

Comments: Accepted by WACV 2025, not published yet

Geminio: Language-Guided Gradient Inversion Attacks in Federated Learning

Authors: Junjie Shan, Ziqi Zhao, Jialin Lu, Rui Zhang, Siu Ming Yiu, Ka-Ho Chow

Abstract: Foundation models that bridge vision and language have made significant progress. While they have inspired many life-enriching applications, their potential for abuse in creating new threats remains largely unexplored. In this paper, we reveal that vision-language models (VLMs) can be weaponized to enhance gradient inversion attacks (GIAs) in federated learning (FL), where an FL server attempts to… ▽ More Foundation models that bridge vision and language have made significant progress. While they have inspired many life-enriching applications, their potential for abuse in creating new threats remains largely unexplored. In this paper, we reveal that vision-language models (VLMs) can be weaponized to enhance gradient inversion attacks (GIAs) in federated learning (FL), where an FL server attempts to reconstruct private data samples from gradients shared by victim clients. Despite recent advances, existing GIAs struggle to reconstruct high-resolution images when the victim has a large local data batch. One promising direction is to focus reconstruction on valuable samples rather than the entire batch, but current methods lack the flexibility to target specific data of interest. To address this gap, we propose Geminio, the first approach to transform GIAs into semantically meaningful, targeted attacks. It enables a brand new privacy attack experience: attackers can describe, in natural language, the data they consider valuable, and Geminio will prioritize reconstruction to focus on those high-value samples. This is achieved by leveraging a pretrained VLM to guide the optimization of a malicious global model that, when shared with and optimized by a victim, retains only gradients of samples that match the attacker-specified query. Geminio can be launched at any FL round and has no impact on normal training (i.e., the FL server can steal clients' data while still producing a high-utility ML model as in benign scenarios). Extensive experiments demonstrate its effectiveness in pinpointing and reconstructing targeted samples, with high success rates across complex datasets and large batch sizes with resilience against defenses. △ Less

Submitted 2 August, 2025; v1 submitted 22 November, 2024; originally announced November 2024.

Comments: ICCV2025 camera-ready version

HGAurban: Heterogeneous Graph Autoencoding for Urban Spatial-Temporal Learning

Authors: Qianru Zhang, Xinyi Gao, Haixin Wang, Dong Huang, Siu-Ming Yiu, Hongzhi Yin

Abstract: Spatial-temporal graph representations play a crucial role in urban sensing applications, including traffic analysis, human mobility behavior modeling, and citywide crime prediction. However, a key challenge lies in the noisy and sparse nature of spatial-temporal data, which limits existing neural networks' ability to learn meaningful region representations in the spatial-temporal graph. To overco… ▽ More Spatial-temporal graph representations play a crucial role in urban sensing applications, including traffic analysis, human mobility behavior modeling, and citywide crime prediction. However, a key challenge lies in the noisy and sparse nature of spatial-temporal data, which limits existing neural networks' ability to learn meaningful region representations in the spatial-temporal graph. To overcome these limitations, we propose HGAurban, a novel heterogeneous spatial-temporal graph masked autoencoder that leverages generative self-supervised learning for robust urban data representation. Our framework introduces a spatial-temporal heterogeneous graph encoder that extracts region-wise dependencies from multi-source data, enabling comprehensive modeling of diverse spatial relationships. Within our self-supervised learning paradigm, we implement a masked autoencoder that jointly processes node features and graph structure. This approach automatically learns heterogeneous spatial-temporal patterns across regions, significantly improving the representation of dynamic temporal correlations. Comprehensive experiments across multiple spatiotemporal mining tasks demonstrate that our framework outperforms state-of-the-art methods and robustly handles real-world urban data challenges, including noise and sparsity in both spatial and temporal dimensions. △ Less

Submitted 14 August, 2025; v1 submitted 14 October, 2024; originally announced October 2024.

Comments: 10 pages

Journal ref: CIKM 2025

TUBench: Benchmarking Large Vision-Language Models on Trustworthiness with Unanswerable Questions

Authors: Xingwei He, Qianru Zhang, A-Long Jin, Yuan Yuan, Siu-Ming Yiu

Abstract: Large Vision-Language Models (LVLMs) have achieved remarkable progress on visual perception and linguistic interpretation. Despite their impressive capabilities across various tasks, LVLMs still suffer from the issue of hallucination, which involves generating content that is incorrect or unfaithful to the visual or textual inputs. Traditional benchmarks, such as MME and POPE, evaluate hallucinati… ▽ More Large Vision-Language Models (LVLMs) have achieved remarkable progress on visual perception and linguistic interpretation. Despite their impressive capabilities across various tasks, LVLMs still suffer from the issue of hallucination, which involves generating content that is incorrect or unfaithful to the visual or textual inputs. Traditional benchmarks, such as MME and POPE, evaluate hallucination in LVLMs within the scope of Visual Question Answering (VQA) using answerable questions. However, some questions are unanswerable due to insufficient information in the images, and the performance of LVLMs on such unanswerable questions remains underexplored. To bridge this research gap, we propose TUBench, a benchmark specifically designed to evaluate the reliability of LVLMs using unanswerable questions. TUBench comprises an extensive collection of high-quality, unanswerable questions that are meticulously crafted using ten distinct strategies. To thoroughly evaluate LVLMs, the unanswerable questions in TUBench are based on images from four diverse domains as visual contexts: screenshots of code snippets, natural images, geometry diagrams, and screenshots of statistical tables. These unanswerable questions are tailored to test LVLMs' trustworthiness in code reasoning, commonsense reasoning, geometric reasoning, and mathematical reasoning related to tables, respectively. We conducted a comprehensive quantitative evaluation of 28 leading foundational models on TUBench, with Gemini-1.5-Pro, the top-performing model, achieving an average accuracy of 69.2%, and GPT-4o, the third-ranked model, reaching 66.7% average accuracy, in determining whether questions are answerable. TUBench is available at https://github.com/NLPCode/TUBench. △ Less

Submitted 5 October, 2024; originally announced October 2024.

A Survey on Point-of-Interest Recommendation: Models, Architectures, and Security

Authors: Qianru Zhang, Peng Yang, Junliang Yu, Haixin Wang, Xingwei He, Siu-Ming Yiu, Hongzhi Yin

Abstract: The widespread adoption of smartphones and Location-Based Social Networks has led to a massive influx of spatio-temporal data, creating unparalleled opportunities for enhancing Point-of-Interest (POI) recommendation systems. These advanced POI systems are crucial for enriching user experiences, enabling personalized interactions, and optimizing decision-making processes in the digital landscape. H… ▽ More The widespread adoption of smartphones and Location-Based Social Networks has led to a massive influx of spatio-temporal data, creating unparalleled opportunities for enhancing Point-of-Interest (POI) recommendation systems. These advanced POI systems are crucial for enriching user experiences, enabling personalized interactions, and optimizing decision-making processes in the digital landscape. However, existing surveys tend to focus on traditional approaches and few of them delve into cutting-edge developments, emerging architectures, as well as security considerations in POI recommendations. To address this gap, our survey stands out by offering a comprehensive, up-to-date review of POI recommendation systems, covering advancements in models, architectures, and security aspects. We systematically examine the transition from traditional models to advanced techniques such as large language models. Additionally, we explore the architectural evolution from centralized to decentralized and federated learning systems, highlighting the improvements in scalability and privacy. Furthermore, we address the increasing importance of security, examining potential vulnerabilities and privacy-preserving approaches. Our taxonomy provides a structured overview of the current state of POI recommendation, while we also identify promising directions for future research in this rapidly advancing field. △ Less

Submitted 9 March, 2025; v1 submitted 3 October, 2024; originally announced October 2024.

Comments: 20 pages

Report number: 20 pages

Journal ref: TKDE 2025