arXiv subjects

Haochen Luo

Publications and source records attributed to Haochen Luo.

16 recordsLinked to original sources

arXiv2026

4DSynth: Controllable Procedural World Synthesis for Dynamic Embodied Simulation

Embodied agents need environments that are visually diverse, physically interactive, and changing over time. Procedural simulators can generate large interactive scene collections, and recent 4D generators produce compelling visual dynamics. Combining these properties in one environment, however, still demands extensive manual effort, and the result is rarely editable or controllable enough to reuse at scale. We present 4DSynth, a controllable procedural system that turns a natural-language description, a blueprint mask, or a single photograph into an editable 4D environment with explicit geometry, animated actors, collision-free trajectories, and physics-ready simulation state. Multiple scene routes share one geometry-grounded representation, so the same pipeline handles animation, camera planning, rendering, and task generation. To validate the full pipeline, we construct 4DSynth-Nav, an interactive navigation benchmark generated entirely from 4DSynth's procedural scenes. Two vision-language models evaluated across three difficulty tiers both fail the majority of tasks and stall after early subtasks. The same procedural controllability that produces these environments also makes each failure reproducible and each difficulty axis independently tunable. This paper presents both a controllable generation pipeline and the scalable benchmark it enables, offering a practical foundation for developing and evaluating embodied agents.

cs.RO↗

arXiv2026

Relay, Don't Route: Adaptive Population Handoff for Cost-Efficient LLM-Driven Evolution

Large language model (LLM)-driven evolution has shown promise for program search and algorithm discovery, but relying on strong models throughout long evolutionary runs is costly. A natural alternative is to combine cheap and strong models under a fixed inference budget. However, existing approaches typically allocate models at the level of individual queries or mutation steps, overlooking that evolutionary search is \textit{stateful}: each generated candidate changes the population from which subsequent mutations are produced. We empirically analyze LLM-driven evolutionary trajectories and find that search progress is strongly front-loaded, early trajectory performance is informative but noisy, and cheap models recover much of the early progress achieved by strong models at lower cost. Motivated by these findings, we propose \textbf{\model}, a training-free framework that shifts budget allocation from individual calls to evolving populations through adaptive \textit{population handoff}. A cheap model explores multiple trajectories in short blocks allocated by a bandit scheduler. Relay Gain, defined as the marginal improvement of a compact, quality-diverse candidate bank constructed for handoff, serves as the scheduler reward and determines when to hand off. The curated candidates initialize a shared strong model population for refinement. Across four benchmarks and three budgets, \model achieves the highest mean score in 11 of 12 settings, outperforming competitive baselines. Our results suggest that in stateful search, budget allocation should be organized around the population, not the individual call.

cs.CL↗

arXiv2026

Harness-Aware Self-Evolving: Co-Evolving Model Weights, Harness, and Task Solutions

Self-evolving frameworks usually optimize task solutions while treating the surrounding harness as fixed. We introduce Harness-Aware Self-Evolving (HASE), an agentic reinforcement-learning framework in which a single model can generate task solutions or edit selected harness components in a multi-turn action space. HASE enables a single Qwen3-8B model to match the text-classification performance of a GPT-OSS-120B model that uses Claude Code as the harness proposer. In alpha factor mining, HASE outperforms the reported GPT-OSS-120B baseline. HASE also repairs imperfect evaluation components and converges to state-of-the-art performance in circle-packing algorithm discovery. These results show that HASE improves the harness and the solution through one unified agentic process.

cs.AI↗

arXiv2026

SeaEvo: Advancing Algorithm Discovery with Strategy Space Evolution

Large Language Model (LLM)-guided evolutionary search is increasingly used for automated algorithm discovery, yet most current methods track search progress primarily through executable programs and scalar fitness. Even when natural-language reasoning is used through heuristic descriptions or reflection, it typically remains transient mutation context or unstructured memory, rather than organized as persistent population-level state over strategic directions. As a result, evolutionary search can struggle to distinguish syntactically different implementations of the same idea, preserve lower-fitness but strategically promising directions, or detect when an entire family of strategies has saturated. We introduce \model, a modular strategy-space layer that turns language-level strategic reasoning into first-class population-level evolutionary state in LLM-driven program search. \model represents each candidate program with an explicit natural-language strategy, clusters the archive by strategy semantics, retrieves behaviorally complementary inspirations, and periodically navigates the strategy landscape to avoid saturated directions. Without modifying the underlying evolutionary algorithms, \model improves existing evolutionary backbones across algorithm discovery, systems optimization, and agent-scaffold design tasks in most settings. Across four systems benchmarks, \model achieves a 20.6% average relative improvement, with the best single run on Prism scoring 3$\times$ higher. These results suggest that persistent strategy representations provide a practical mechanism for improving the effectiveness and cost-efficiency of LLM-guided evolutionary search, pointing toward compound AI systems whose search capabilities benefit from the structured accumulation and reuse of algorithmic strategies.

cs.CL↗

arXiv2026

PDA: Text-Augmented Defense Framework for Robust Vision-Language Models against Adversarial Image Attacks

Vision-language models (VLMs) are vulnerable to adversarial image perturbations. Existing works based on adversarial training against task-specific adversarial examples are computationally expensive and often fail to generalize to unseen attack types. To address these limitations, we introduce Paraphrase-Decomposition-Aggregation (PDA), a training-free defense framework that leverages text augmentation to enhance VLM robustness under diverse adversarial image attacks. PDA performs prompt paraphrasing, question decomposition, and consistency aggregation entirely at test time, thus requiring no modification on the underlying models. To balance robustness and efficiency, we instantiate PDA as invariants that reduce the inference cost while retaining most of its robustness gains. Experiments on multiple VLM architectures and benchmarks for visual question answering, classification, and captioning show that PDA achieves consistent robustness gains against various adversarial perturbations while maintaining competitive clean accuracy, establishing a generic, strong and practical defense framework for VLMs during inference.

cs.CV↗

arXiv2026

From Natural Language to Executable Option Strategies via Large Language Models

Large Language Models (LLMs) excel at general code generation, yet translating natural-language trading intents into correct option strategies remains challenging. Real-world option design requires reasoning over massive, multi-dimensional option chain data with strict constraints, which often overwhelms direct generation methods. We introduce the Option Query Language (OQL), a domain-specific intermediate representation that abstracts option markets into high-level primitives under grammatical rules, enabling LLMs to function as reliable semantic parsers rather than free-form programmers. OQL queries are then validated and executed deterministically by an engine to instantiate executable strategies. We also present a new dataset for this task and demonstrate that our neuro-symbolic pipeline significantly improves execution accuracy and logical consistency over direct baselines.

cs.AI↗

arXiv2025

Evolutionary Factor Searching for Sparse Portfolio Optimization Using Large Language Models

Sparse portfolio optimization is a fundamental yet challenging problem in quantitative finance. Traditional approaches often use static objectives and thus adapt poorly to dynamic market regimes. In this work, we propose Evolutionary Factor Search, a framework that leverages large language models and evolutionary algorithms to automatically generate and evolve alpha factors for sparse portfolio construction. The framework recasts asset selection as a ranking task guided by the generated factors and uses an evolutionary feedback loop to iteratively refine the factor pool from portfolio performance. To handle the inter-factor redundancy that accumulates during this search, we further introduce a redundancy-aware weight allocation module that combines random-matrix-theory denoising of the factor correlation matrix with regularized quadratic programming, at negligible overhead and without additional tuning. Extensive experiments on four Fama-French benchmarks and three real-market datasets spanning the United States, Hong Kong, and Mainland China equity markets show that the proposed framework outperforms statistical and optimization-based baselines across diverse markets. Ablation studies further validate the importance of prompt composition, factor diversity, and language-model choice. These results highlight language-model-guided evolution as a robust and interpretable paradigm for portfolio optimization under structural constraints.

q-fin.PM↗

arXiv2025

DualOptim: Enhancing Efficacy and Stability in Machine Unlearning with Dual Optimizers

Existing machine unlearning (MU) approaches exhibit significant sensitivity to hyperparameters, requiring meticulous tuning that limits practical deployment. In this work, we first empirically demonstrate the instability and suboptimal performance of existing popular MU methods when deployed in different scenarios. To address this issue, we propose Dual Optimizer (DualOptim), which incorporates adaptive learning rate and decoupled momentum factors. Empirical and theoretical evidence demonstrates that DualOptim contributes to effective and stable unlearning. Through extensive experiments, we show that DualOptim can significantly boost MU efficacy and stability across diverse tasks, including image classification, image generation, and large language models, making it a versatile approach to empower existing MU algorithms.

cs.LG↗

arXiv2024

ECDQC: Efficient Compilation for Distributed Quantum Computing with Linear Layout

In this paper, we propose an efficient compilation method for distributed quantum computing (DQC) using the Linear Nearest Neighbor (LNN) architecture. By exploiting the LNN topology's symmetry, we optimize quantum circuit compilation for High Local Connectivity, Sparse Full Connectivity (HLC-SFC) algorithms like Quantum Approximate Optimization Algorithm (QAOA) and Quantum Fourier Transform (QFT). We also utilize dangling qubits to minimize non-local interactions and reduce SWAP gates. Our approach significantly decreases compilation time, gate count, and circuit depth, improving scalability and robustness for large-scale quantum computations.

quant-ph↗

arXiv2024

Differentially Private Neural Network Training Under the Hidden State Assumption

Current differentially private learning paradigms face a severe utility bottleneck: DP-SGD degrades performance through noise accumulation over training steps, while aggregation-based approaches such as PATE suffer from data inefficiency due to disjoint data partitioning. We propose \textbf{Differentially Private Decoupled Training (DP-DT)}, a framework that decouples representation learning from privacy enforcement. DP-DT confines noise injection to the weight aggregation stage and employs auxiliary models, continuously synchronized with a global model, to perform noise-free feature extraction on private data shards. We prove that DP-DT converges globally to a limit point under non-convex objectives, via a Lyapunov potential analysis combined with the Kurdyka--\L{}ojasiewicz property. Under the Hidden State Assumption, where adversaries observe only the final published model, we further prove that DP-DT's privacy loss can converge to a constant bound rather than accumulating with iterations. Empirical results across vision and language benchmarks confirm that DP-DT significantly mitigates utility degradation, achieving state-of-the-art privacy-utility trade-offs.

cs.LG↗

arXiv2024

Which Model Generated This Image? A Model-Agnostic Approach for Origin Attribution

Recent progress in visual generative models enables the generation of high-quality images. To prevent the misuse of generated images, it is important to identify the origin model that generates them. In this work, we study the origin attribution of generated images in a practical setting where only a few images generated by a source model are available and the source model cannot be accessed. The goal is to check if a given image is generated by the source model. We first formulate this problem as a few-shot one-class classification task. To solve the task, we propose OCC-CLIP, a CLIP-based framework for few-shot one-class classification, enabling the identification of an image's source model, even among multiple candidates. Extensive experiments corresponding to various generative models verify the effectiveness of our OCC-CLIP framework. Furthermore, an experiment based on the recently released DALL-E 3 API verifies the real-world applicability of our solution.

cs.CV↗

arXiv2024

An Image Is Worth 1000 Lies: Adversarial Transferability across Prompts on Vision-Language Models

Different from traditional task-specific vision models, recent large VLMs can readily adapt to different vision tasks by simply using different textual instructions, i.e., prompts. However, a well-known concern about traditional task-specific vision models is that they can be misled by imperceptible adversarial perturbations. Furthermore, the concern is exacerbated by the phenomenon that the same adversarial perturbations can fool different task-specific models. Given that VLMs rely on prompts to adapt to different tasks, an intriguing question emerges: Can a single adversarial image mislead all predictions of VLMs when a thousand different prompts are given? This question essentially introduces a novel perspective on adversarial transferability: cross-prompt adversarial transferability. In this work, we propose the Cross-Prompt Attack (CroPA). This proposed method updates the visual adversarial perturbation with learnable prompts, which are designed to counteract the misleading effects of the adversarial image. By doing this, CroPA significantly improves the transferability of adversarial examples across prompts. Extensive experiments are conducted to verify the strong cross-prompt adversarial transferability of CroPA with prevalent VLMs including Flamingo, BLIP-2, and InstructBLIP in various different tasks. Our source code is available at \url{https://github.com/Haochen-Luo/CroPA}.

cs.CV↗

arXiv2023

Together We Make Sense -- Learning Meta-Sense Embeddings from Pretrained Static Sense Embeddings

Sense embedding learning methods learn multiple vectors for a given ambiguous word, corresponding to its different word senses. For this purpose, different methods have been proposed in prior work on sense embedding learning that use different sense inventories, sense-tagged corpora and learning methods. However, not all existing sense embeddings cover all senses of ambiguous words equally well due to the discrepancies in their training resources. To address this problem, we propose the first-ever meta-sense embedding method -- Neighbour Preserving Meta-Sense Embeddings, which learns meta-sense embeddings by combining multiple independently trained source sense embeddings such that the sense neighbourhoods computed from the source embeddings are preserved in the meta-embedding space. Our proposed method can combine source sense embeddings that cover different sets of word senses. Experimental results on Word Sense Disambiguation (WSD) and Word-in-Context (WiC) tasks show that the proposed meta-sense embedding method consistently outperforms several competitive baselines.

cs.CL↗

arXiv2022

A Novel Markov Model for Near-Term Railway Delay Prediction

Predicting the near-future delay with accuracy for trains is momentous for railway operations and passengers' traveling experience. This work aims to design prediction models for train delays based on Netherlands Railway data. We first develop a chi-square test to show that the delay evolution over stations follows a first-order Markov chain. We then propose a delay prediction model based on non-homogeneous Markov chains. To deal with the sparsity of the transition matrices of the Markov chains, we propose a novel matrix recovery approach that relies on Gaussian kernel density estimation. Our numerical tests show that this recovery approach outperforms other heuristic approaches in prediction accuracy. The Markov chain model we propose also shows to be better than other widely-used time series models with respect to both interpretability and prediction accuracy. Moreover, our proposed model does not require a complicated training process, which is capable of handling large-scale forecasting problems.

cs.LG↗

arXiv2021

A new parsimonious method for classifying Cancer Tissue-of-Origin Based on DNA Methylation 450K data

DNA methylation is a well-studied genetic modification that regulates gene transcription of Eukaryotes. Its alternations have been recognized as a significant component of cancer development. In this study, we use the DNA methylation 450k data from The Cancer Genome Atlas to evaluate the efficacy of DNA methylation data on cancer classification for 30 cancer types. We propose a new method for gene selection in high dimensional data(over 450 thousand). Variance filtering is first introduced for dimension reduction and Recursive feature elimination (RFE) is then used for feature selection. We address the problem of selecting a small subsets of genes from large number of methylated sites, and our parsimonious model is demonstrated to be efficient, achieving an accuracy over 91%, outperforming other studies which use DNA micro-arrays and RNA-seq Data . The performance of 20 models, which are based on 4 estimators (Random Forest, Decision Tree, Extra Tree and Support Vector Machine) and 5 classifiers (k-Nearest Neighbours, Support Vector Machine, XGboost, Light GBM and Multi-Layer Perceptron), is compared and robustness of the RFE algorithm is examined. Results suggest that the combined model of extra tree plus catboost classifier offers the best performance in cancer identification, with an overall validation accuracy of 91% , 92.3%, 93.3% and 93.5% for 20, 30, 40 and 50 features respectively. The biological functions in cancer development of 50 selected genes is also explored through enrichment analysis and the results show that 12 out of 16 of our top features have already been identified to be specific with cancer and we also propose some more genes to be tested for future studies. Therefore, our method may be utilzed as an auxiliary diagnostic method to determine the actual clinicopathological status of a specific cancer.

q-bio.TO↗

arXiv2017

Optimum Search Schemes for Approximate String Matching Using Bidirectional FM-Index

Finding approximate occurrences of a pattern in a text using a full-text index is a central problem in bioinformatics and has been extensively researched. Bidirectional indices have opened new possibilities in this regard allowing the search to start from anywhere within the pattern and extend in both directions. In particular, use of search schemes (partitioning the pattern and searching the pieces in certain orders with given bounds on errors) can yield significant speed-ups. However, finding optimal search schemes is a difficult combinatorial optimization problem. Here for the first time, we propose a mixed integer program (MIP) capable to solve this optimization problem for Hamming distance with given number of pieces. Our experiments show that the optimal search schemes found by our MIP significantly improve the performance of search in bidirectional FM-index upon previous ad-hoc solutions. For example, approximate matching of 101-bp Illumina reads (with two errors) becomes 35 times faster than standard backtracking. Moreover, despite being performed purely in the index, the running time of search using our optimal schemes (for up to two errors) is comparable to the best state-of-the-art aligners, which benefit from combining search in index with in-text verification using dynamic programming. As a result, we anticipate a full-fledged aligner that employs an intelligent combination of search in the bidirectional FM-index using our optimal search schemes and in-text verification using dynamic programming outperforms today's best aligners. The development of such an aligner, called FAMOUS (Fast Approximate string Matching using OptimUm search Schemes), is ongoing as our future work.

cs.DS↗