arXiv subjects

Yang Liu

Publications and source records attributed to Yang Liu.

At least 721 records · Page 40Linked to original sources

arXiv2026

Resisting Manipulative Bots in Meme Coin Copy Trading: A Multi-Agent Approach with Chain-of-Thought Reasoning

Copy trading has become the dominant entry strategy in meme coin markets. However, due to the market's extremely illiquid and volatile nature, the strategy exposes an exploitable attack surface: adversaries deploy manipulative bots to front-run trades, conceal positions, and fabricate sentiment, systematically extracting value from naïve copiers at scale. Despite its prevalence, bot-driven manipulation remains largely unexplored, and no robust defensive framework exists. We propose a manipulation-resistant copy-trading system based on a multi-agent architecture powered by a multi-modal large language model (LLM) and chain-of-thought (CoT) reasoning. Our approach outperforms zero-shot and most statistic-driven baselines in prediction accuracy as well as all baselines in economic performance, achieving an average copier return of 3% per meme coin investment under realistic market frictions. Overall, our results demonstrate the effectiveness of agent-based defenses and predictability of trader profitability in adversarial meme coin markets, providing a practical foundation for robust copy trading.

cs.AI↗

arXiv2026

From Helpfulness to Toxic Proactivity: Diagnosing Behavioral Misalignment in LLM Agents

The enhanced capabilities of LLM-based agents come with an emergency for model planning and tool-use abilities. Attributing to helpful-harmless trade-off from LLM alignment, agents typically also inherit the flaw of "over-refusal", which is a passive failure mode. However, the proactive planning and action capabilities of agents introduce another crucial danger on the other side of the trade-off. This phenomenon we term "Toxic Proactivity'': an active failure mode in which an agent, driven by the optimization for Machiavellian helpfulness, disregards ethical constraints to maximize utility. Unlike over-refusal, Toxic Proactivity manifests as the agent taking excessive or manipulative measures to ensure its "usefulness'' is maintained. Existing research pays little attention to identifying this behavior, as it often lacks the subtle context required for such strategies to unfold. To reveal this risk, we introduce a novel evaluation framework based on dilemma-driven interactions between dual models, enabling the simulation and analysis of agent behavior over multi-step behavioral trajectories. Through extensive experiments with mainstream LLMs, we demonstrate that Toxic Proactivity is a widespread behavioral phenomenon and reveal two major tendencies. We further present a systematic benchmark for evaluating Toxic Proactive behavior across contextual settings.

cs.CL↗

arXiv2026

LogPrism: Unifying Structure and Variable Encoding for Effective Log Compression

In the field of log compression, the prevailing "parse-then-compress" paradigm fundamentally limits effectiveness by treating log parsing and compression as isolated objectives. While parsers prioritize semantic accuracy (i.e., event identification), they often obscure deep correlations between static templates and dynamic variables that are critical for storage efficiency. In this paper, we investigate this misalignment through a comprehensive empirical study and propose LogPrism, a framework that bridges the gap via unified redundancy encoding. Rather than relying on a rigid pre-parsing step, LogPrism dynamically integrates structural extraction with variable encoding by constructing a Unified Redundancy Tree (URT). This hierarchical approach effectively mines "structure+variable" co-occurrence patterns, capturing deep contextual redundancies while accelerating processing through pre-emptive pattern encoding. Extensive experiments on 16 benchmark datasets confirm that LogPrism establishes a new state-of-the-art. It achieves the highest compression ratio on 14 datasets, surpassing existing baselines by margins of 6.12% to 83.34%, while delivering superior throughput at 29.87 MB/s (1.68$\times$~43.04$\times$ faster than competitors). Moreover, when configured in single-archive mode to maximize global pattern discovery, LogPrism boosts its compression ratio by 273.27%, outperforming the best baseline by 19.39% with a 2.62$\times$ speed advantage.

cs.SE↗

arXiv2026

LogicScan: An LLM-driven Framework for Detecting Business Logic Vulnerabilities in Smart Contracts

Business logic vulnerabilities have become one of the most damaging yet least understood classes of smart contract vulnerabilities. Unlike traditional bugs such as reentrancy or arithmetic errors, these vulnerabilities arise from missing or incorrectly enforced business invariants and are tightly coupled with protocol semantics. Existing static analysis techniques struggle to capture such high-level logic, while recent large language model based approaches often suffer from unstable outputs and low accuracy due to hallucination and limited verification. In this paper, we propose LogicScan, an automated contrastive auditing framework for detecting business logic vulnerabilities in smart contracts. The key insight behind LogicScan is that mature, widely deployed on-chain protocols implicitly encode well-tested and consensus-driven business invariants. LogicScan systematically mines these invariants from large-scale on-chain contracts and reuses them as reference constraints to audit target contracts. To achieve this, LogicScan introduces a Business Specification Language (BSL) to normalize diverse implementation patterns into structured, verifiable logic representations. It further combines noise-aware logic aggregation with contrastive auditing to identify missing or weakly enforced invariants while mitigating LLM-induced false positives. We evaluate LogicScan on three real-world datasets, including DeFiHacks, Web3Bugs, and a set of top-200 audited contracts. The results show that LogicScan achieves an F1 score of 85.2%, significantly outperforming state-of-the-art tools while maintaining a low false-positive rate on production-grade contracts. Additional experiments demonstrate that LogicScan maintains consistent performance across different LLMs and is cost-effective, and that its false-positive suppression mechanisms substantially improve robustness.

cs.CR↗

arXiv2026

From Pre- to Intra-operative MRI: Predicting Brain Shift in Temporal Lobe Resection for Epilepsy Surgery

Introduction: In neurosurgery, image-guided Neurosurgery Systems (IGNS) highly rely on preoperative brain magnetic resonance images (MRI) to assist surgeons in locating surgical targets and determining surgical paths. However, brain shift invalidates the preoperative MRI after dural opening. Updated intraoperative brain MRI with brain shift compensation is crucial for enhancing the precision of neuronavigation systems and ensuring the optimal outcome of surgical interventions. Methodology: We propose NeuralShift, a U-Net-based model that predicts brain shift entirely from pre-operative MRI for patients undergoing temporal lobe resection. We evaluated our results using Target Registration Errors (TREs) computed on anatomical landmarks located on the resection side and along the midline, and DICE scores comparing predicted intraoperative masks with masks derived from intraoperative MRI. Results: Our experimental results show that our model can predict the global deformation of the brain (DICE of 0.97) with accurate local displacements (achieve landmark TRE as low as 1.12 mm), compensating for large brain shifts during temporal lobe removal neurosurgery. Conclusion: Our proposed model is capable of predicting the global deformation of the brain during temporal lobe resection using only preoperative images, providing potential opportunities to the surgical team to increase safety and efficiency of neurosurgery and better outcomes to patients. Our contributions will be publicly available after acceptance in https://github.com/SurgicalDataScienceKCL/NeuralShift.

cs.CV↗

arXiv2026

AudioJailbreak: Jailbreak Attacks against End-to-End Large Audio-Language Models

Jailbreak attacks to Large audio-language models (LALMs) are studied recently, but they exclusively focused on the attack scenario where the adversary can fully manipulate user prompts (named strong adversary) and limited in effectiveness, applicability, and practicability. In this work, we first conduct an extensive evaluation showing that advanced text jailbreak attacks cannot be easily ported to end-to-end LALMs via text-to-speech (TTS) techniques. We then propose AUDIOJAILBREAK, a novel audio jailbreak attack, featuring (1) asynchrony: the jailbreak audios do not need to align with user prompts in the time axis by crafting suffixal jailbreak audios; (2) universality: a single jailbreak perturbation is effective for different prompts by incorporating multiple prompts into the perturbation generation; (3) stealthiness: the malicious intent of jailbreak audios is concealed by proposing various intent concealment strategies; and (4) over-the-air robustness: the jailbreak audios remain effective when being played over the air by incorporating reverberation into the perturbation generation. In contrast, all prior audio jailbreak attacks cannot offer asynchrony, universality, stealthiness, and/or over-the-air robustness. Moreover, AUDIOJAILBREAK is also applicable to a more practical and broader attack scenario where the adversary cannot fully manipulate user prompts (named weak adversary). Extensive experiments with thus far the most LALMs demonstrate the high effectiveness of AUDIOJAILBREAK, in particular, it can jailbreak openAI's GPT-4o-Audio and bypass Meta's Llama-Guard-3 safeguard, in the weak adversary scenario. We highlight that our work peeks into the security implications of audio jailbreak attacks against LALMs, and realistically fosters improving their robustness, especially for the newly proposed weak adversary.

cs.CR↗

arXiv2026

Small-Margin Preferences Still Matter-If You Train Them Right

Preference optimization methods such as DPO align large language models (LLMs) using paired comparisons, but their effectiveness can be highly sensitive to the quality and difficulty of preference pairs. A common heuristic treats small-margin (ambiguous) pairs as noisy and filters them out. In this paper, we revisit this assumption and show that pair difficulty interacts strongly with the optimization objective: when trained with preference-based losses, difficult pairs can destabilize training and harm alignment, yet these same pairs still contain useful supervision signals when optimized with supervised fine-tuning (SFT). Motivated by this observation, we propose MixDPO, a simple yet effective difficulty-aware training strategy that (i) orders preference data from easy to hard (a curriculum over margin-defined difficulty), and (ii) routes difficult pairs to an SFT objective while applying a preference loss to easy pairs. This hybrid design provides a practical mechanism to leverage ambiguous pairs without incurring the optimization failures often associated with preference losses on low-margin data. Across three LLM-judge benchmarks, MixDPO consistently improves alignment over DPO and a range of widely-used variants, with particularly strong gains on AlpacaEval~2 length-controlled (LC) win rate.

cs.AI↗

arXiv2026

Cast: Automated Resilience Testing for Production Cloud Service Systems

The distributed nature of microservice architecture introduces significant resilience challenges. Traditional testing methods, limited by extensive manual effort and oversimplified test environments, fail to capture production system complexity. To address these limitations, we present Cast, an automated, end-to-end framework for microservice resilience testing in production. It achieves high test fidelity by replaying production traffic against a comprehensive library of application-level faults to exercise internal error-handling logic. To manage the combinatorial test space, Cast employs a complexity-driven strategy to systematically prune redundant tests and prioritize high-value tests targeting the most critical service execution paths. Cast automates the testing lifecycle through a three-phase pipeline (i.e., startup, fault injection, and recovery) and uses a multi-faceted oracle to automatically verify system resilience against nuanced criteria. Deployed in Huawei Cloud for over eight months, Cast has been adopted by many service teams to proactively address resilience vulnerabilities. Our analysis on four large-scale applications with millions of traces reveals 137 potential vulnerabilities, with 89 confirmed by developers. To further quantify its performance, Cast is evaluated on a benchmark set of 48 reproduced bugs, achieving a high coverage of 90%. The results show that Cast is a practical and effective solution for systematically improving the reliability of industrial microservice systems.

cs.SE↗

arXiv2026

Hard Constraints Meet Soft Generation: Guaranteed Feasibility for LLM-based Combinatorial Optimization

Large language models (LLMs) have emerged as promising general-purpose solvers for combinatorial optimization (CO), yet they fundamentally lack mechanisms to guarantee solution feasibility which is critical for real-world deployment. In this work, we introduce FALCON, a framework that ensures 100\% feasibility through three key innovations: (i) \emph{grammar-constrained decoding} enforces syntactic validity, (ii) a \emph{feasibility repair layer} corrects semantic constraint violations, and (iii) \emph{adaptive Best-of-$N$ sampling} allocates inference compute efficiently. To train the underlying LLM, we introduce the Best-anchored Objective-guided Preference Optimization (BOPO) in LLM training, which weights preference pairs by their objective gap, providing dense supervision without human labels. Theoretically, we prove convergence for BOPO and provide bounds on repair-induced quality loss. Empirically, across seven NP-hard CO problems, FALCON achieves perfect feasibility while matching or exceeding the solution quality of state-of-the-art neural and LLM-based solvers.

cs.AI↗

arXiv2026

Reflecting with Two Voices: A Co-Adaptive Dual-Strategy Framework for LLM-Based Agent Decision Making

Large language model (LLM) agents often rely on external demonstrations or retrieval-augmented planning, leading to brittleness, poor generalization, and high computational overhead. Inspired by human problem-solving, we propose DuSAR (Dual-Strategy Agent with Reflecting) -- a demonstration-free framework that enables a single frozen LLM to perform co-adaptive reasoning via two complementary strategies: a high-level holistic plan and a context-grounded local policy. These strategies interact through a lightweight reflection mechanism, where the agent continuously assesses progress via a Strategy Fitness Score and dynamically revises its global plan when stuck or refines it upon meaningful advancement, mimicking human metacognitive behavior. On both simulated household (ALFWorld) and real-world web (Mind2Web) environments, DuSAR achieves state-of-the-art performance using only open-source LLMs, substantially outperforming all prior methods without any demonstrations or fine-tuning. Remarkably, it also reduces per-step token consumption by a large margin while maintaining strong task success. Ablation studies confirm the necessity of dual-strategy coordination. Moreover, optional integration of expert demonstrations further boosts performance, highlighting DuSAR's flexibility and compatibility with external knowledge.

cs.AI↗

arXiv2026

NCP: Neighborhood-Preserving Non-Uniform Circle Packing for Visualization

Circle packing is widely used in visualization due to its aesthetic appeal and simplicity, particularly in tasks where the spatial arrangement and relationships between data are of interest, such as understanding proximity relationships (e.g., images with categories) or analyzing quantitative data (e.g., housing prices). Many applications require preserving neighborhood relationships while encoding a quantitative attribute using radii for data analysis. To meet these two requirements simultaneously, we present a neighborhood-preserving non-uniform circle packing method, NCP. This method preserves neighborhood relationships between the data represented by non-uniform circles to comprehensively analyze similar data and an attribute of interest. We formulate neighborhood-preserving non-uniform circle packing as a planar graph embedding problem based on the circle packing theorem. This formulation leads to a non-convex optimization problem, which can be solved by the continuation method. We conduct a quantitative evaluation and present two use cases to demonstrate that our NCP method can effectively generate non-uniform circle packing results.

cs.HC↗

arXiv2026

OMGEval: An Open Multilingual Generative Evaluation Benchmark for Large Language Models

Modern large language models (LLMs) should generally benefit individuals from various cultural backgrounds around the world. However, most recent advanced generative evaluation benchmarks tailed for LLMs mainly focus on English. To this end, we introduce OMGEval, the first Open-source Multilingual Generative test set that can assess the capability of LLMs in different languages. For each language, OMGEval provides 804 open-ended questions, covering a wide range of important capabilities of LLMs, such as general knowledge, logical reasoning, and so on. Each question is rigorously verified by human annotators. Notably, to sufficiently reflect the compatibility of LLMs in different cultural backgrounds, we perform localization for each non-English language. Specifically, the current version of OMGEval includes 5 languages (i.e., Zh, Ru, Fr, Es, Ar). Following AlpacaEval, we employ GPT-4 as the adjudicator to automatically score different model outputs, which is shown closely related to human evaluation. We evaluate several representative multilingual LLMs on the proposed OMGEval, which we believe will provide a valuable reference for the community to further understand and improve the multilingual capability of LLMs. OMGEval is available at https://github.com/blcuicall/OMGEval.

cs.CL↗

arXiv2026

AgentIF-OneDay: A Task-level Instruction-Following Benchmark for General AI Agents in Daily Scenarios

The capacity of AI agents to effectively handle tasks of increasing duration and complexity continues to grow, demonstrating exceptional performance in coding, deep research, and complex problem-solving evaluations. However, in daily scenarios, the perception of these advanced AI capabilities among general users remains limited. We argue that current evaluations prioritize increasing task difficulty without sufficiently addressing the diversity of agentic tasks necessary to cover the daily work, life, and learning activities of a broad demographic. To address this, we propose AgentIF-OneDay, aimed at determining whether general users can utilize natural language instructions and AI agents to complete a diverse array of daily tasks. These tasks require not only solving problems through dialogue but also understanding various attachment types and delivering tangible file-based results. The benchmark is structured around three user-centric categories: Open Workflow Execution, which assesses adherence to explicit and complex workflows; Latent Instruction, which requires agents to infer implicit instructions from attachments; and Iterative Refinement, which involves modifying or expanding upon ongoing work. We employ instance-level rubrics and a refined evaluation pipeline that aligns LLM-based verification with human judgment, achieving an 80.1% agreement rate using Gemini-3-Pro. AgentIF-OneDay comprises 104 tasks covering 767 scoring points. We benchmarked four leading general AI agents and found that agent products built based on APIs and ChatGPT agents based on agent RL remain in the first tier simultaneously. Leading LLM APIs and open-source models have internalized agentic capabilities, enabling AI application teams to develop cutting-edge Agent products.

cs.CL↗

arXiv2026

InstructDiff: Domain-Adaptive Data Selection via Differential Entropy for Efficient LLM Fine-Tuning

Supervised fine-tuning (SFT) is fundamental to adapting large language models, yet training on complete datasets incurs prohibitive costs with diminishing returns. Existing data selection methods suffer from severe domain specificity: techniques optimized for general instruction-following fail on reasoning tasks, and vice versa. We observe that measuring entropy differences between base models and minimally instruction-tuned calibrated models reveals a pattern -- samples with the lowest differential entropy consistently yield optimal performance across domains, yet this principle manifests domain-adaptively: reasoning tasks favor entropy increase (cognitive expansion), while general tasks favor entropy decrease (cognitive compression). We introduce InstructDiff, a unified framework that operationalizes differential entropy as a domain-adaptive selection criterion through warmup calibration, bi-directional NLL filtering, and entropy-based ranking. Extensive experiments show that InstructDiff achieves 17\% relative improvement over full data training on mathematical reasoning and 52\% for general instruction-following, outperforming prior baselines while using only 10\% of the data.

cs.CL↗

arXiv2026

Infinite Magnetoresistance and Vortex Coupling in the Pb/BSCCO Heterostructure

Combining superconductivity with spintronics provides exciting opportunities to realize low-dissipation quantum devices. Here we report the synthesis, characterization and magnetotransport measurements of the Pb/Bi$_2$Sr$_2$CaCu$_2$O$_{8+δ}$ (BSCCO) superconducting heterostructures, where an insulating PbO$_{x}$ layer spontaneously forms at the interface. Non-volatile switching between superconducting (logical "0") and normal ("1") states in Pb films by an external field, i.e., infinite magnetoresistance (IMR), can be realized and are attributed to the strong trapping and pinning of vortices in BSCCO. Furthermore, butterfly-shaped hysteresis loops in magnetoresistance, pronounced resistance dips/jumps and thermal reset to superconducting states can be observed and are direct manifestations of the peculiar vortex dynamics in BSCCO and vortex coupling across the Pb/BSCCO interface. Our work demonstrates a simple and effective way to realize IMR through superconducting vortices and opens up new opportunities to study the vortex interactions across the superconducting interfaces.

cond-mat.supr-con↗

arXiv2026

Can Distillation Mitigate Backdoor Attacks in Pre-trained Encoders?

Self-Supervised Learning (SSL) has become a prominent paradigm for pre-training encoders to learning general-purpose representations from unlabeled data and releasing them on third-party platforms for broad downstream deep learning tasks. However, SSL is vulnerable to backdoor attacks, where an adversary may train and distribute poisoned pre-training encoders to contaminate the downstream models. In this paper, we study a defense mechanism based on distillation against poisoned encoders in SSL. Traditionally, distillation transfers knowledge from a pre-trained teacher model to a student model, enabling the student to replicate or refine the teacher's learned representations. We repurpose distillation to extract benign knowledge and remove backdoors from a poisoned pre-trained encoder to produce a clean and reliable pre-trained model. We conduct extensive experiments to evaluate the effectiveness of distillation in mitigating backdoor attacks on pre-trained encoders. Based on two state-of-the-art backdoor attacks and four widely adopted image classification datasets, our results demonstrate that distillation reduces the attack success rate from 80.87% to 27.51%, with only a 6.35% drop in model accuracy. Furthermore, by comparing four teacher architectures, three student models, and six loss functions, we find that the distillation with fine-tuned teacher networks, warm-up-based student training, and attention-based distillation losses yield the best performance.

cs.LG↗

arXiv2026

A Survey of LLM Alignment: Instruction Understanding, Intention Reasoning, and Reliable Generation

Large language models have demonstrated exceptional capabilities in understanding and generation. However, in real-world scenarios, users' natural language expressions are often inherently fuzzy, ambiguous, and uncertain, leading to challenges such as vagueness, polysemy, and contextual ambiguity. This paper focuses on three challenges in LLM-based text generation tasks: instruction understanding, intention reasoning, and reliable dialog generation. Regarding human complex instruction, LLMs have deficiencies in understanding long contexts and instructions in multi-round conversations. For intention reasoning, LLMs may have inconsistent command reasoning, difficulty reasoning about commands containing incorrect information, difficulty understanding user ambiguous language commands, and a weak understanding of user intention in commands. Besides, In terms of Reliable Dialog Generation, LLMs may have unstable generated content and unethical generation. To this end, we classify and analyze the performance of LLMs in challenging scenarios and conduct a comprehensive evaluation of existing solutions. Furthermore, we introduce benchmarks and categorize them based on the aforementioned three core challenges. Finally, we explore potential directions for future research to enhance the reliability and adaptability of LLMs in real-world applications.

cs.HC↗

arXiv2026

Towards Transparent RAG: Fostering Evidence Traceability in LLM Generation via Reinforcement Learning

Retrieval-Augmented Generation (RAG) delivers substantial value in knowledge-intensive applications. However, its generated responses often lack transparent reasoning paths that trace back to source evidence from retrieved documents. This opacity not only compromises the interpretability of the output but also limits the model's ability to fully exploit the provided context. To address this, we propose TRACE (Transparent RAG with evidenCE tracing), a framework designed to enhance evidence traceability in Large Language Models (LLMs) through reinforcement learning (RL). TRACE guides LLMs to produce structured outputs with explicit evidence citations by prompting and rewarding evidence relevance and proper formatting, alongside accuracy, to optimize structured traceability. To ensure training stability with multiple reward signals, we further introduce an adaptive strategy for merging rewards and adopt a stabilized KL-divergence estimator. Experiments on three multi-hop QA datasets using Qwen2.5-7B-Instruct and Llama-3.1-8B-Instruct show that TRACE achieves both transparent, evidence-attributed outputs and accuracy improvements of 10-30%. The resulting performance is comparable to advanced commercial LLMs (e.g., OpenAI o1, DeepSeek-R1). Further analyses demonstrate strong generalization capabilities to unseen tasks. Our code is publicly available now.

cs.CL↗