arXiv subjects

Yanming Liu

Publications and source records attributed to Yanming Liu.

At least 19 recordsLinked to original sources

arXiv2026

SkillForge: Compositional Skill Synthesis with Verification-in-the-Loop for Generating Formally Verified Dafny Programs

Generating formally verified programs from natural language remains challenging: existing approaches either produce code in a single pass without recourse when verification fails, or rely on open-ended agentic reasoning that is non-deterministic and opaque. We introduce SKILLFORGE, a framework that decomposes formal code synthesis into a library of atomic, reusable skills, each targeting a specific subtask such as specification inference, body synthesis, invariant generation, error diagnosis, or targeted repair, and defined by a prompt template, tool binding, and decidable success criterion. A verification-driven harness orchestrates these skills: it submits candidates to the Dafny verifier, diagnoses failures into structured categories, deterministically routes to the appropriate repair skill, and iterates until formal correctness is proved or a budget is exhausted. On a curated benchmark of natural language to Dafny specification pairs, SKILLFORGE substantially outperforms both state-of-the-art agentic approaches (including ReAct-style agents, MCTS-based repair, and RL-guided verification) and traditional iterative baselines, while requiring fewer tokens and lower latency. Ablation studies confirm that every skill contributes measurably, and the harness converges rapidly with the majority of programs verified on the first attempt.

cs.CL↗

arXiv2026

AttriMem: Attribution-Guided Process Feedback for Agent Memory Construction

Effective memory is crucial for LLM agents, yet constructing it effectively remains challenging. A memory-construction policy decides what information to extract, store, update, compress, or discard as interactions accumulate. Heuristic memory methods rely on subjective, task-specific rules, which can misalign with downstream objectives and limit cross-task adaptability. RL-based methods, by contrast, learn from task feedback but mainly use outcome- or module-level rewards. These coarse signals indicate task success but cannot identify which intermediate memory contents support the final answer, creating a fine-grained credit-assignment bottleneck. However, constructing such process feedback is prohibitively difficult because intermediate memory decisions lack unique ground-truth targets, while the appropriate credit varies with the agent's uncertain reasoning trajectory and therefore cannot be specified in advance. We propose AttriMem, an attribution-guided process-feedback framework for learning memory-construction policies with RL. AttriMem augments the global outcome reward with local rewards derived from token-level contributions to the final answer. Experiments on long-horizon dialogue question answering show that AttriMem outperforms retrieval-based, heuristic, and RL-based baselines, generalizes across benchmarks and answer models, stabilizes RL optimization.

cs.AI↗

arXiv2026

OASIF: An Efficient Obfuscation-Aware Self-Improving Framework for LLM-Based Assembly Code Instruction Following and Comprehension

Large Language Models (LLMs) have recently shown promise in automated binary analysis, yet they remain brittle under commercial-grade obfuscation. We present OASIF, an Obfuscation-Aware Self-evolving Instruction-Following framework for obfuscated assembly comprehension. OASIF couples a token-efficient assembly encoder with a lightweight projector to expose long obfuscated code to a pretrained code LLM under a bounded context budget and follows a three-phase training: (i) feature-space alignment, (ii) supervised instruction fine-tuning, and (iii) online self-evolving reinforcement learning with hybrid rewards, enabling continual adaptation with minimal manual verification. On VMISA-Bench, a challenging out-of-distribution suite featuring three commercial VM-based obfuscators, OASIF consistently improves open-source backbones; Qwen2.5-Coder-Instruct-14B attains Success Rate gains of +15.9, +5.8, and +16.9 percentage points (pp) on Code Virtualizer, Themida (v3.0.7), and VMProtect (v3.5), respectively, and improves the OASIF-Bench average by +9.8. OASIF further delivers stable gains across seven standard BCSD benchmarks while preserving general and domain-relevant capabilities on HumanEval, VulBench, and HumanEval-Decompile.

cs.SE↗

arXiv2026

PADD: Path-Aligned Decompression Distillation for Non-Router Teacher to Guide MoE Student Learning

As large language models (LLMs) continue to scale, it becomes increasingly challenging to grow model capacity under fixed computation budgets. We propose Path-Aligned Decompression Distillation (PADD), a framework for distilling knowledge from dense teachers without explicit routing into mixture-of-experts (MoE) students while learning high-quality routing policies. PADD organizes knowledge distillation into four stages in two phases: an initialization phase (Stage I) that builds diverse functionality in the student's experts through teacher neuron clustering and student-expert warmup, and a training phase (Stages II--IV) that integrates online adaptive distillation, path-refined policy optimization, and reward-augmented load balancing in a single training pipeline. Experiments on mathematical reasoning benchmarks demonstrate that PADD yields substantial gains over strong baselines at the same inference cost and that the MoE student can match or surpass its dense teacher. They also demonstrate effective teacher-to-student knowledge distillation and stable routing behavior.

cs.CL↗

arXiv2026

Dep-Search: Learning Dependency-Aware Reasoning Traces with Persistent Memory

Large Language Models (LLMs) have demonstrated remarkable capabilities in complex reasoning tasks, particularly when augmented with search mechanisms that enable systematic exploration of external knowledge bases. The field has evolved from traditional retrieval-augmented generation (RAG) frameworks to more sophisticated search-based frameworks that orchestrate multi-step reasoning through explicit search strategies. However, existing search frameworks still rely heavily on implicit natural language reasoning to determine search strategies and how to leverage retrieved information across reasoning steps. This reliance on implicit reasoning creates fundamental challenges for managing dependencies between sub-questions, efficiently reusing previously retrieved knowledge, and learning optimal search strategies through reinforcement learning. To address these limitations, we propose Dep-Search, a dependency-aware search framework that advances beyond existing search frameworks by integrating structured reasoning, retrieval, and persistent memory through GRPO. Dep-Search introduces explicit control mechanisms that enable the model to decompose questions with dependency relationships, retrieve information when needed, access previously stored knowledge from memory, and summarize long reasoning contexts into reusable memory entries. Through extensive experiments on seven diverse question answering datasets, we demonstrate that Dep-Search significantly enhances LLMs' ability to tackle complex multi-hop reasoning tasks, achieving substantial improvements over strong baselines across different model scales.

cs.CL↗

arXiv2026

ToolGate: Contract-Grounded and Verified Tool Execution for LLMs

Large Language Models (LLMs) augmented with external tools have demonstrated remarkable capabilities in complex reasoning tasks. However, existing frameworks rely heavily on natural language reasoning to determine when tools can be invoked and whether their results should be committed, lacking formal guarantees for logical safety and verifiability. We present \textbf{ToolGate}, a forward execution framework that provides logical safety guarantees and verifiable state evolution for LLM tool calling. ToolGate maintains an explicit symbolic state space as a typed key-value mapping representing trusted world information throughout the reasoning process. Each tool is formalized as a Hoare-style contract consisting of a precondition and a postcondition, where the precondition gates tool invocation by checking whether the current state satisfies the required conditions, and the postcondition determines whether the tool's result can be committed to update the state through runtime verification. Our approach guarantees that the symbolic state evolves only through verified tool executions, preventing invalid or hallucinated results from corrupting the world representation. Experimental validation demonstrates that ToolGate significantly improves the reliability and verifiability of tool-augmented LLM systems while maintaining competitive performance on complex multi-step reasoning tasks. This work establishes a foundation for building more trustworthy and debuggable AI systems that integrate language models with external tools.

cs.CL↗

arXiv2026

NC2C: Automated Convexification of Generic Non-Convex Optimization Problems

Non-convex optimization problems are pervasive across mathematical programming, engineering design, and scientific computing, often posing intractable challenges for traditional solvers due to their complex objective functions and constrained landscapes. To address the inefficiency of manual convexification and the over-reliance on expert knowledge, we propose NC2C, an LLM-based end-to-end automated framework designed to transform generic non-convex optimization problems into solvable convex forms using large language models. NC2C leverages LLMs' mathematical reasoning capabilities to autonomously detect non-convex components, select optimal convexification strategies, and generate rigorous convex equivalents. The framework integrates symbolic reasoning, adaptive transformation techniques, and iterative validation, equipped with error correction loops and feasibility domain correction mechanisms to ensure the robustness and validity of transformed problems. Experimental results on a diverse dataset of 100 generic non-convex problems demonstrate that NC2C achieves an 89.3\% execution rate and a 76\% success rate in producing feasible, high-quality convex transformations. This outperforms baseline methods by a significant margin, highlighting NC2C's ability to leverage LLMs for automated non-convex to convex transformation, reduce expert dependency, and enable efficient deployment of convex solvers for previously intractable optimization tasks.

cs.CL↗

arXiv2025

Deep learning with hybrid frequency differencing and principal component analysis for 21-cm foreground and beam mitigation

Twenty-one-centimeter intensity mapping is a powerful probe of the large-scale distribution of neutral hydrogen (HI) and cosmological observables such as baryon acoustic oscillations. A major challenge is contamination from bright foregrounds and frequency-dependent beam effects, which can lead to signal loss in traditional methods such as principal component analysis (PCA). We develop a hybrid approach that trains a U-shaped convolutional neural network (UNet) on two input channels derived from frequency differencing (FD) and PCA cleaning, enabling it to exploit their complementary behavior across different scales. This two-channel strategy achieves improved performance, maintaining the cross-correlation power spectrum close to unity on large scales under a cosine beam and improving by 5\%-8\% relative to either FD- or PCA-based UNet alone. We further show that the method can robustly recover the HI signal even when the beam model is imperfect and differs between training and testing, with the large-scale cross-correlation remaining close to unity within the $1\sigma$ level. These results demonstrate that the proposed approach provides a robust framework for HI signal reconstruction under realistic observational conditions.

astro-ph.CO↗

arXiv2025

Movable Antenna-Enhanced Secure Communication: Opportunities, Challenges, and Solutions

The broadcast nature of wireless communication renders it inherently vulnerable to security threats such as jamming and eavesdropping. While traditional array beamforming techniques help to mitigate these threats, they usually incur high hardware and processing costs, particularly in large-scale arrays with fixed-position antennas (FPAs). In contrast, movable antenna (MA) arrays can fully exploit the channel variation in spatial regions by enabling flexible antenna movement, which has emerged as a promising technology for secure communications. This article provides a magazine-type overview of MA-aided secure communications. Specifically, we first illuminate the promising application scenarios for MA-enhanced secure communication systems. Then, we examine the security advantages of MAs over conventional FPA systems, fundamentally stemming from their ability to adjust channel correlations between legitimate users, eavesdroppers, and jammers. Furthermore, we discuss important technical challenges and their potential solutions related to MA hardware architecture, channel acquisition, and antenna position optimization to realize secure transmissions. Finally, several promising directions for MA-aided secure communications are presented to inspire future research.

cs.IT↗

arXiv2025

Movable Antenna Empowered Secure Near-Field MIMO Communications

This paper investigates movable antenna (MA) empowered secure transmission in near-field multiple-input multiple-output (MIMO) communication systems, where the base station (BS) equipped with an MA array transmits confidential information to a legitimate user under the threat of a potential eavesdropper. To enhance physical layer security (PLS) of the considered system, we aim to maximize the secrecy rate by jointly designing the hybrid digital and analog beamformers, as well as the positions of MAs at the BS. To solve the formulated non-convex problem with highly coupled variables, an alternating optimization (AO)-based algorithm is introduced by decoupling the original problem into two separate subproblems. Specifically, for the subproblem of designing hybrid beamformers, a semi-closed-form solution for the fully-digital beamformer is first derived by a weighted minimum mean-square error (WMMSE)-based algorithm. Subsequently, the digital and analog beamformers are determined by approximating the fully-digital beamformer through the manifold optimization (MO) technique. For the MA positions design subproblem, we utilize the majorization-minimization (MM) algorithm to iteratively optimize each MA's position while keeping others fixed. Extensive simulation results validate the considerable benefits of the proposed MA-aided near-field beam focusing approach in enhancing security performance compared to the traditional far-field and/or the fixed position antenna (FPA)-based systems. In addition, the proposed scheme can realize secure transmission even if the eavesdropper is located in the same direction as the user and closer to the BS.

cs.IT↗

arXiv2025

LLM-OptiRA: LLM-Driven Optimization of Resource Allocation for Non-Convex Problems in Wireless Communications

Solving non-convex resource allocation problems poses significant challenges in wireless communication systems, often beyond the capability of traditional optimization techniques. To address this issue, we propose LLM-OptiRA, the first framework that leverages large language models (LLMs) to automatically detect and transform non-convex components into solvable forms, enabling fully automated resolution of non-convex resource allocation problems in wireless communication systems. LLM-OptiRA not only simplifies problem-solving by reducing reliance on expert knowledge, but also integrates error correction and feasibility validation mechanisms to ensure robustness. Experimental results show that LLM-OptiRA achieves an execution rate of 96% and a success rate of 80% on GPT-4, significantly outperforming baseline approaches in complex optimization tasks across diverse scenarios.

cs.CL↗

arXiv2025

ASMA-Tune: Unlocking LLMs' Assembly Code Comprehension via Structural-Semantic Instruction Tuning

Assembly code analysis and comprehension play critical roles in applications like reverse engineering, yet they face substantial challenges due to low information density and a lack of explicit syntactic structures. While traditional masked language modeling (MLM) approaches do not explicitly focus on natural language interaction, emerging decoder-focused large language models (LLMs) demonstrate partial success in binary analysis yet remain underexplored for holistic comprehension. We present Assembly Augmented Tuning, an end-to-end structural-semantic instruction tuning framework that synergizes encoder architecture with decoder-based LLMs through a projector module, where the assembly encoder extracts hardware-level structural features, the projector bridges representations with the semantic space, and the instruction-tuned LLM preserves natural language capabilities. Experimental results demonstrate three key advantages: (1) State-of-the-art performance in assembly comprehension with +39.7% Recall@1 and +17.8% MRR improvements over GPT-4-Turbo, (2) Consistent enhancements across base models (24.6-107.4% Recall@1 and 15.2-106.3% MRR on Qwen2.5-Coder, Deepseek-Coder and CodeLlama variants), and (3) Superior instruction-following capabilities (41.5%-118% improvements) with controlled code generation degradation (-8.9% to -35% across architectures).

cs.SE↗

arXiv2025

Cosmological distance forecasts for the CSST Galaxy Survey using BAO peaks

The measurement of cosmological distances using baryon acoustic oscillations (BAO) is crucial for studying the universe's expansion. The Chinese Space Station Telescope (CSST) galaxy redshift survey, with its vast volume and sky coverage, provides an opportunity to address key challenges in cosmology. However, redshift uncertainties in galaxy surveys can degrade both angular and radial distance estimates. In this study, we forecast the precision of BAO distance measurements using mock CSST galaxy samples, applying a two-point correlation function (2PCF) wedge approach to mitigate redshift errors. We simulate redshift uncertainties of $\sigma_0 = 0.003$ and $\sigma_0 = 0.006$, representative of expected CSST errors, and examine their effects on the BAO peak and distance scaling factors, $\alpha_\perp$ and $\alpha_\parallel$, across redshift bins within $0.0 < z \leqslant 1.0$. The wedge 2PCF method proves more effective in detecting the BAO peak compared to the monopole 2PCF, particularly for $\sigma_0 = 0.006$. Constraints on the BAO peaks show that $\alpha_\perp$ is well constrained around 1.0, regardless of $\sigma_0$, with precision between 1% and 3% across redshift bins. In contrast, $\alpha_\parallel$ measurements are more sensitive to increases in $\sigma_0$. For $\sigma_0 = 0.003$, the results remain close to the fiducial value, with uncertainties ranging between 4% and 9%; for $\sigma_0 = 0.006$, significant deviations from the fiducial value are observed. We also study the ability to measure parameters $(\Omega_m, H_0r_\mathrm{d})$ using distance measurements, proving robust constraints as a cosmological probe under CSST-like redshift uncertainties.

astro-ph.CO↗

arXiv2024

Movable-Antenna Aided Secure Transmission for RIS-ISAC Systems

Integrated sensing and communication (ISAC) systems have the issue of secrecy leakage when using the ISAC waveforms for sensing, thus posing a potential risk for eavesdropping. To address this problem, we propose to employ movable antennas (MAs) and reconfigurable intelligent surface (RIS) to enhance the physical layer security (PLS) performance of ISAC systems, where an eavesdropping target potentially wiretaps the signals transmitted by the base station (BS). To evaluate the synergistic performance gain provided by MAs and RIS, we formulate an optimization problem for maximizing the sum-rate of the users by jointly optimizing the transmit/receive beamformers of the BS, the reflection coefficients of the RIS, and the positions of MAs at communication users, subject to a minimum communication rate requirement for each user, a minimum radar sensing requirement, and a maximum secrecy leakage to the eavesdropping target. To solve this non-convex problem with highly coupled variables, a two-layer penalty-based algorithm is developed by updating the penalty parameter in the outer-layer iterations to achieve a trade-off between the optimality and feasibility of the solution. In the inner-layer iterations, the auxiliary variables are first obtained with semi-closed-form solutions using Lagrange duality. Then, the receive beamformer filter at the BS is optimized by solving a Rayleigh-quotient subproblem. Subsequently, the transmit beamformer matrix is obtained by solving a convex subproblem. Finally, the majorization-minimization (MM) algorithm is employed to optimize the RIS reflection coefficients and the positions of MAs. Extensive simulation results validate the considerable benefits of the proposed MAs-aided RIS-ISAC systems in enhancing security performance compared to traditional fixed position antenna (FPA)-based systems.

cs.IT↗

arXiv2024

Bridging Context Gaps: Leveraging Coreference Resolution for Long Contextual Understanding

Large language models (LLMs) have shown remarkable capabilities in natural language processing; however, they still face difficulties when tasked with understanding lengthy contexts and executing effective question answering. These challenges often arise due to the complexity and ambiguity present in longer texts. To enhance the performance of LLMs in such scenarios, we introduce the Long Question Coreference Adaptation (LQCA) method. This innovative framework focuses on coreference resolution tailored to long contexts, allowing the model to identify and manage references effectively. The LQCA method encompasses four key steps: resolving coreferences within sub-documents, computing the distances between mentions, defining a representative mention for coreference, and answering questions through mention replacement. By processing information systematically, the framework provides easier-to-handle partitions for LLMs, promoting better understanding. Experimental evaluations on a range of LLMs and datasets have yielded positive results, with a notable improvements on OpenAI-o1-mini and GPT-4o models, highlighting the effectiveness of leveraging coreference resolution to bridge context gaps in question answering. Our code is public at https://github.com/OceannTwT/LQCA.

cs.CL↗

arXiv2024

DP-MemArc: Differential Privacy Transfer Learning for Memory Efficient Language Models

Large language models have repeatedly shown outstanding performance across diverse applications. However, deploying these models can inadvertently risk user privacy. The significant memory demands during training pose a major challenge in terms of resource consumption. This substantial size places a heavy load on memory resources, raising considerable practical concerns. In this paper, we introduce DP-MemArc, a novel training framework aimed at reducing the memory costs of large language models while emphasizing the protection of user data privacy. DP-MemArc incorporates side network or reversible network designs to support a variety of differential privacy memory-efficient fine-tuning schemes. Our approach not only achieves about 2.5 times in memory optimization but also ensures robust privacy protection, keeping user data secure and confidential. Extensive experiments have demonstrated that DP-MemArc effectively provides differential privacy-efficient fine-tuning across different task scenarios.

cs.CR↗

arXiv2024

Tool-Planner: Task Planning with Clusters across Multiple Tools

Large language models (LLMs) have demonstrated exceptional reasoning capabilities, enabling them to solve various complex problems. Recently, this ability has been applied to the paradigm of tool learning. Tool learning involves providing examples of tool usage and their corresponding functions, allowing LLMs to formulate plans and demonstrate the process of invoking and executing each tool. LLMs can address tasks that they cannot complete independently, thereby enhancing their potential across different tasks. However, this approach faces two key challenges. First, redundant error correction leads to unstable planning and long execution time. Additionally, designing a correct plan among multiple tools is also a challenge in tool learning. To address these issues, we propose Tool-Planner, a task-processing framework based on toolkits. Tool-Planner groups tools based on the API functions with the same function into a toolkit and allows LLMs to implement planning across the various toolkits. When a tool error occurs, the language model can reselect and adjust tools based on the toolkit. Experiments show that our approach demonstrates a high pass and win rate across different datasets and optimizes the planning scheme for tool learning in models such as GPT-4 and Claude 3, showcasing the potential of our method. Our code is public at https://github.com/OceannTwT/Tool-Planner

cs.AI↗

arXiv2024

CPA-Enhancer: Chain-of-Thought Prompted Adaptive Enhancer for Object Detection under Unknown Degradations

Object detection methods under known single degradations have been extensively investigated. However, existing approaches require prior knowledge of the degradation type and train a separate model for each, limiting their practical applications in unpredictable environments. To address this challenge, we propose a chain-of-thought (CoT) prompted adaptive enhancer, CPA-Enhancer, for object detection under unknown degradations. Specifically, CPA-Enhancer progressively adapts its enhancement strategy under the step-by-step guidance of CoT prompts, that encode degradation-related information. To the best of our knowledge, it's the first work that exploits CoT prompting for object detection tasks. Overall, CPA-Enhancer is a plug-and-play enhancement model that can be integrated into any generic detectors to achieve substantial gains on degraded images, without knowing the degradation type priorly. Experimental results demonstrate that CPA-Enhancer not only sets the new state of the art for object detection but also boosts the performance of other downstream vision tasks under unknown degradations.

cs.CV↗