BZPEERPapers, people, evidence and research paths
Menu

Discover

SearchFind papers, people, topics, institutions and opportunities from one place.GraphSee how papers, authors, topics and evidence connect.

Research tools

HomeContinue from the papers, people and topics that matter now.CollectionsTurn saved papers into reading lists, evidence maps and shareable context.ResearchReconstruct the path that led you from search to paper to question.CollaborationMove from a paper, topic or expert request into a focused thread.PublishCreate full-text research articles with metadata, formulas and versions.

Network

ExpertsFind specialists whose work and reviews explain why they can help.CommunitiesJoin research groups organized around papers, topics and shared evidence.

Opportunities

ListingsBrowse roles, calls and collaborations connected to real research context.

Account

Log inReturn to your saved papers, graph views and active threads.Create accountStart saving papers, building a research trail and joining teams.
Log inCreate account

Source author record

Siu Ming Yiu

Siu Ming Yiu appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record
Artificial IntelligenceMachine LearningComputation and LanguageNeural and Evolutionary ComputingComputer Science and Game TheoryCryptography and SecurityDistributed, Parallel, and Cluster ComputingInformation RetrievalOperating Systemsq-fin.CPSound

Catalog footprint

What is connected

10works
11topics
4close collaborators

Actions

Connect this record

Log in to claim
Open graphBrowse works

Research graph

See the researcher in context

Open full explorer

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

10 published item(s)

preprint2026arXiv

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

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.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

Efficient Prompt Learning for Traffic Forecasting

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.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

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

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.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

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

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.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

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

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.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

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

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.

0 citations0 reviewsNo code linkedOpen access
preprint2026arXiv

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

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.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Deep Multiple Instance Learning For Forecasting Stock Trends Using Financial News

A major source of information can be taken from financial news articles, which have some correlations about the fluctuation of stock trends. In this paper, we investigate the influences of financial news on the stock trends, from a multi-instance view. The intuition behind this is based on the news uncertainty of varying intervals of news occurrences and the lack of annotation in every single financial news. Under the scenario of Multiple Instance Learning (MIL) where training instances are arranged in bags, and a label is assigned for the entire bag instead of instances, we develop a flexible and adaptive multi-instance learning model and evaluate its ability in directional movement forecast of Standard & Poors 500 index on financial news dataset. Specifically, we treat each trading day as one bag, with certain amounts of news happening on each trading day as instances in each bag. Experiment results demonstrate that our proposed multi-instance-based framework gains outstanding results in terms of the accuracy of trend prediction, compared with other state-of-art approaches and baselines.

0 citations0 reviewsNo code linkedOpen access
preprint2020arXiv

Optimal Procurement Auction for Cooperative Production of Virtual Products: Vickrey-Clarke-Groves Meet Cremer-McLean

We set up a supply-side game-theoretic model for the cooperative production of virtual products. In our model, a group of producers collaboratively produce a virtual product by contributing costly input resources to a production coalition. Producers are capacitated, i.e., they cannot contribute more resources than their capacity limits. Our model is an abstraction of emerging internet-based business models such as federated learning and crowd computing. To maintain an efficient and stable production coalition, the coordinator should share with producers the income brought by the virtual product. Besides the demand-side information asymmetry, another two sources of supply-side information asymmetry intertwined in this problem: 1) the capacity limit of each producer and 2) the cost incurred to each producer. In this paper, we rigorously prove that a supply-side mechanism from the VCG family, PVCG, can overcome such multiple information asymmetry and guarantee truthfulness. Furthermore, with some reasonable assumptions, PVCG simultaneously attains truthfulness, ex-post allocative efficiency, ex-post individual rationality, and ex-post weak budget balancedness on the supply side, easing the well-known tension between these four objectives in the mechanism design literature.

0 citations0 reviewsNo code linkedOpen access
preprint2016arXiv

A multi-task learning model for malware classification with useful file access pattern from API call sequence

Based on API call sequences, semantic-aware and machine learning (ML) based malware classifiers can be built for malware detection or classification. Previous works concentrate on crafting and extracting various features from malware binaries, disassembled binaries or API calls via static or dynamic analysis and resorting to ML to build classifiers. However, they tend to involve too much feature engineering and fail to provide interpretability. We solve these two problems with the recent advances in deep learning: 1) RNN-based autoencoders (RNN-AEs) can automatically learn low-dimensional representation of a malware from its raw API call sequence. 2) Multiple decoders can be trained under different supervisions to give more information, other than the class or family label of a malware. Inspired by the works of document classification and automatic sentence summarization, each API call sequence can be regarded as a sentence. In this paper, we make the first attempt to build a multi-task malware learning model based on API call sequences. The model consists of two decoders, one for malware classification and one for $\emph{file access pattern}$ (FAP) generation given the API call sequence of a malware. We base our model on the general seq2seq framework. Experiments show that our model can give competitive classification results as well as insightful FAP information.

0 citations0 reviewsNo code linkedOpen access