arXiv subjects

Licheng Wang

Publications and source records attributed to Licheng Wang.

At least 19 recordsLinked to original sources

arXiv2026

The Horizon Gap: Planning, Memory, Execution, Training, and Evaluation for Long-Horizon LLM Agents

Frontier language models solve reasoning problems in a single forward pass that would have been research contributions years ago, yet fail at multi-hour tasks: losing track of earlier decisions, declaring half-finished work done, or drifting from goals. We call this the horizon gap and survey 1,547 arXiv papers (2024-2026) collected via systematic seed harvest with a disclosed 26.8% bleed filter, extended by targeted supplementation. We disambiguate three routinely conflated properties: long-horizon (task property: required steps), long-context (model property: token capacity), and long-term memory (system property: persistence across steps/sessions). We organize the corpus into six categories tracking a long-horizon task's lifecycle -- planning, memory, execution, training, evaluation, and foundations/safety -- crossed with an axis capturing where horizons are carried (within-context, within-task-beyond-context, or cross-task-persistent). Across all categories, we find the same pattern: outcome-only signals grow uninformative as horizons lengthen, and the field's response -- whether process reward models, credit assignment, or trajectory-level diagnostics -- manufactures denser step-level signals. We treat critical and diagnostic literature as first-class threads throughout, arguing that segregating critique from method would routinely split single papers across chapters. We close by naming open measurement problems: decomposing model versus harness capability, managing correlated bias in process-level signals used for both training and evaluation, and whether long-horizon reliability admits general predictive theory.

cs.CL↗

arXiv2026

The Calibration Floor: Format Repair Can Masquerade as Self-Correction at Small-to-Mid Scale

Accuracy changes after language-model self-revision are usually interpreted as changes in reasoning. We show this can fail at the answer-extraction boundary, and test the failure causally rather than only observationally. Across Qwen3.5 (0.8B-9B), Gemma-4-12B, and two frontier models via API (Tencent Hy3, Nvidia Nemotron-3-Ultra-550B) in 29 primary cells plus a frontier arm, we decompose the always-revise accuracy shift into a content margin (both answers parseable) and format-recovery/loss margins (parseability changes). On 12 cells with meaningful unparseable-answer rates, format effects exceed content effects (Wilcoxon p=1.7e-3). To test this causally, we force already-generated reasoning through grammar-constrained decoding so every answer is parseable by construction: across 14 cells this closes a median 71% of the gap between the naive total effect and the content-margin estimate, with two cells converging exactly and a residual on the two largest-effect cells reported rather than dismissed. A clustered model confirms floor-scale (0.8B/2B) models have far higher odds of content-level change and harm than capable-scale models (p<1e-7). Replicating a cited confidence-gating protocol verbatim on Qwen3.5 does not reproduce its reported gain and shows the same near-zero content margin. A frontier check on much larger models shows format-dominance intensifying with scale: content margin is exactly zero in all 5 cells despite total effects up to +0.275, though this arm is lower-powered. The calibration-floor criterion on the content margin reveals a squeeze: floor-scale cells have headroom but insufficient signal, capable-scale cells have signal but little headroom; only one cell is marginally viable, with negligible sealed-holdout gain. Content is a minority share of what the field has measured as self-correction. We release the instrument, code, and derived results.

cs.CL↗

arXiv2026

MDGAM-Based Cooperative Task Scheduling for Communication-Constrained Distributed Multi-Agent Systems

Cooperative task scheduling in communication-constrained distributed multi-agent systems is challenging because each agent must make decisions from partial and dynamic observations while satisfying complex practical constraints. Existing heuristics rely on handcrafted bidding rules and repeated consensus, whereas many learning-based methods assume global observations and lack explicit communication-based coordination. To address these limitations, this paper proposes a neural scheduling framework for distributed multi-robot task allocation (MRTA), consisting of a multi-decoder graph attention model (MDGAM) policy model and a critic-free group relative multi-agent policy gradient (GRMAPG) training algorithm. MDGAM uses an extended graph attention mechanism to jointly update node and edge features, and employs multiple decoders to generate task-selection decisions and communication messages. GRMAPG constructs group-relative advantages from equivalent task-planning instances to replace the critic network used in conventional MARL algorithms, thereby reducing training difficulty and improving convergence performance. Experiments under different problem scales and communication ranges show that the proposed method improves task-completion performance over existing heuristic and learning-based methods, while ablation, complexity, and generalization tests further validate the proposed innovations.

cs.MA↗

arXiv2026

$\mathrm{P}^{3}$CDA: Privacy-Preserving and Provably Secure Cross Domain Authentication Scheme for Internet of Drones

With the rapid expansion of the Internet of Drones (IoD) and the increasing mobility of drones, cross-domain interactions among geographically distributed domains have become inevitable. Cross-domain authentication is therefore a fundamental security requirement for IoD. However, existing authentication schemes often struggle to simultaneously achieve strong security, high efficiency, and identity privacy, making them unsuitable for the stringent requirements of highly dynamic and resource-constrained IoD environments. To address this challenge, we propose $\mathrm{P}^{3}$CDA, a privacy-preserving and provably secure cross-domain authentication scheme. First, we design an efficient pseudonym management mechanism that supports adaptive pseudonym generation as well as batch registration, verification, and revocation. Second, we propose a structurally enhanced Merkle Hash Tree (MHT) that supports batch pseudonym updates, thereby reducing the pseudonym storage overhead of drones. Building on these components, we develop a cryptographic accumulator-based cross-domain authentication protocol that enables anonymous authentication with authorized pseudonyms while preserving the traceability and efficient revocation of malicious drones. We rigorously analyze the security of $\mathrm{P}^{3}$CDA and formally prove its security under the Canetti--Krawczyk (CK) adversary model. Extensive experiments demonstrate that $\mathrm{P}^{3}$CDA achieves lower computational, communication, and storage overhead than state-of-the-art schemes.

cs.CR↗

arXiv2026

Recursive Self-Improvement in AI: From Bounded Self-Refinement to Autonomous Research Loops

AI systems increasingly participate in their own improvement: revising their outputs, adapting their harnesses during deployment, training on data they generate, and conducting AI research itself. This literature uses a vocabulary ("self-refine," "self-reward," "self-play," "self-evolve") that conflates fundamentally different ambitions. We survey 1,250 arXiv papers (2024-2026) along two axes: what the system improves -- its behavior in deployment, its policy through training, its evaluator, or the research process itself -- and the degree of loop closure (human-in-the-loop to fully closed). The taxonomy separates bounded self-refinement -- convergent, evaluable, and industrial practice -- from open-ended recursive self-improvement (RSI), which remains bounded by grounding requirements, collapse dynamics, and compute constraints on every measured axis. Its distinctive feature is a dedicated category for self-evaluation: every improvement loop is a claim that some signal can substitute for human judgment. We survey the evaluator design space -- judges, process reward models, verifiers, rubrics, meta-evaluation -- order the signals into a verification hierarchy from formal verifiers (strongest) to intrinsic self-assessment (weakest), and observe that demonstrated self-improvement strength tracks this hierarchy, that its failure modes (self-confirming loops, model and diversity collapse) follow from its violations, and that the "research direction-setting" bottleneck keeping humans in the loop divides into a verification problem the hierarchy indexes and a prior one -- choosing what deserves evaluation at all -- that it does not. We connect the literature to the theory of RSI limits and to the safety and governance questions raised by frontier-lab accounts of closing the loop, and identify governance-grade measurement of self-improvement as the field's most underpopulated niche.

cs.AI↗

arXiv2026

Robust class-gated single-pixel diffractive optical neural network with random-aberration-aware training

Optical computing offers the theoretical potential for high-speed, energy-efficient inference, yet its practical deployment remains constrained by fundamental input-output bottlenecks, particularly the reliance on electronic sensors with limited frame rates and stringent alignment requirements between optical components. Here, we demonstrate an image-class-gated single-pixel DONN that overcomes these limitations by converting spatial complexity into a temporal intensity signature. Using a minimal architecture comprising a reconfigurable digital micromirror device and a single-pixel photodetector, we implement a virtual optical gate. The system time-multiplexes class-specific masks, causing the detector response to peak only when the mask index matches the input class. This allows the predicted label to be read out via peak timing rather than spatial localization, eliminating 2D sensor constraints. To bridge the persistent sim-to-real gap, we introduce a physics-aware training strategy using random-phase augmentation. This method renders the model intrinsically tolerant to phase aberrations and mechanical misalignments without requiring precise hardware modeling. Our prototype achieves 90.0%(MNIST) and 80.0% (Fashion-MNIST) accuracy at a readout rate of 5 kHz. By combining gigahertz-compatible single-pixel detection with robust and alignment-tolerant training, this work provides a scalable, hardware-efficient pathway toward real-time optical intelligent sensing.

physics.optics↗

arXiv2026

Spore: Efficient and Training-Free Privacy Extraction Attack on LLMs via Inference-Time Hybrid Probing

With the wide adoption of personal AI assistants such as OpenClaw, privacy leakage in user interaction contexts with large language model (LLM) agents has become a critical issue. Existing privacy attacks against LLMs primarily target training data, while research on inference-time contextual privacy risks in LLM agent memory remains limited. Moreover, prior methods often incur high attack costs, requiring multiple queries or relying on white-box assumptions, which limits their practicality in real-world deployments. To address these issues, we propose a training-free privacy extraction attack targeting LLM agent memory, which we name \textsc{Spore}. \textsc{Spore} is compatible with both black-box and gray-box settings. In the black-box setting, \textsc{Spore} can efficiently extract a small candidate set via a single query to recover the original private information. In the gray-box setting, \textsc{Spore} allows the attacker to leverage multi-ranked tokens for more accurate and faster privacy extraction. We provide an information-theoretic analysis of \textsc{Spore} and show that it achieves high query efficiency with substantial per query information leakage. Experiments on multiple frontier LLMs show that \textsc{Spore} outperforms attack success rate over existing state-of-the-art (SOTA) schemes. It also maintains low attack cost and remains stable across different model parameter settings. We further evaluate the robustness of \textsc{Spore} against existing defense mechanisms. Our results show that \textsc{Spore} consistently bypasses both detection and strong safety alignment, demonstrating resilient performance in diverse defensive settings and real-world safety threats.

cs.CR↗

arXiv2026

Decomposing Fractional Quantum Hall Wave Functions via Operator Contraction Multiplication

We develop a general algebraic scheme to decompose fractional quantum Hall (FQH) wave functions based on the operator contraction multiplication. By introducing fermionic and bosonic operators and establishing three fundamental contraction rules, we achieve an exact decomposition of Laughlin states. This approach naturally extends to multi-component systems by factorizing coupled Jastrow factors via resultants and elementary symmetric polynomials, enabling the first complete decomposition of Halperin states. For Halperin ($2,2,1$) state, we explicitly derive its basic expansion, identify root configurations, and reveal intra- and inter-color squeezing operators, thereby uncovering the underlying generalized Pauli principle. Using this method, we compute orbital entanglement spectra for up to $16$ particles with decomposition dimensions exceeding $10^{11}$, obtaining edge excitation sequences that precisely match chiral Luttinger liquid theory. Our framework breaks through the longstanding limitations of Jack polynomials, provides a unified decomposition for both single- and multi-component FQH states, and opens a new avenue for exploring wave functions for more complex FQH states.

cond-mat.str-el↗

arXiv2026

EGAM: Extended Graph Attention Model for Solving Routing Problems

Neural combinatorial optimization (NCO) solvers, implemented with graph neural networks (GNNs), have introduced new approaches for solving routing problems. Trained with reinforcement learning (RL), the state-of-the-art graph attention model (GAM) achieves near-optimal solutions without requiring expert knowledge or labeled data. In this work, we generalize the existing graph attention mechanism and propose the extended graph attention model (EGAM). Our model utilizes multi-head dot-product attention to update both node and edge embeddings, addressing the limitations of the conventional GAM, which considers only node features. We employ an autoregressive encoder-decoder architecture and train it with policy gradient algorithms that incorporate a specially designed baseline. Experiments show that EGAM matches or outperforms existing methods across various routing problems. Notably, the proposed model demonstrates exceptional performance on highly constrained problems, highlighting its efficiency in handling complex graph structures.

cs.LG↗

arXiv2026

Alternative $\nu+\nu$-picture of bosonic fractional Chern insulators at high filling factors in multiple flat-band systems

Most fractional quantum Hall states have been traditionally identified within a single energy band, such as the lowest Landau level or topological flat band. As more particles are introduced, they inevitably populate higher energy bands. Whether the inclusion of multiple topological bands leads to new physics remains an open question. Here, we propose a universal picture applicable at higher filling factors $\nu \geq 1$ in bosonic systems: the occupied bands tend to coalesce into an effective single topological band characterized by a total Chern number $\vert C\vert$, the sum of the Chern number of all occupied lower topological flat bands. Using a Kekul\'{e} lattice model with two lower flat bands featuring a total Chern number $C=1$, regardless of their specific configurations, we identify the emergence of a $\frac{1}{2}$ fractional Chern insulator (FCI) state at integer filling factor $\nu=1$, followed by the Jain sequence states $\frac{2}{3}$ and $\frac{3}{4}$ at filling $\nu=\frac{4}{3}$ and $\frac{6}{4}$. That is a $\nu+\nu$ picture, rather than the generally expected $1+\nu^{\prime}$ picture, where $\nu^{\prime}$ is the permitted FCI filling factor in the single second topological flat band. Our findings deepen the understanding of FCI states and open avenues for discovering exotic fractional topological phases in multiband systems.

cond-mat.str-el↗

arXiv2025

Towards Provably Secure Generative AI: Reliable Consensus Sampling

Existing research on generative AI security is primarily driven by mutually reinforcing attack and defense methodologies grounded in empirical experience. This dynamic frequently gives rise to previously unknown attacks that can circumvent current detection and prevention. This necessitates the continual updating of security mechanisms. Constructing generative AI with provable security and theoretically controllable risk is therefore necessary. Consensus Sampling (CS) is a promising algorithm toward provably secure AI. It controls risk by leveraging overlap in model output probabilities. However, we find that CS relies on frequent abstention to avoid unsafe outputs, which reduces utility. Moreover, CS becomes highly vulnerable when unsafe models are maliciously manipulated. To address these issues, we propose a new primitive called Reliable Consensus Sampling (RCS), that traces acceptance probability to tolerate extreme adversarial behaviors, improving robustness. RCS also eliminates the need for abstention entirely. We further develop a feedback algorithm to continuously and dynamically enhance the safety of RCS. We provide theoretical guarantees that RCS maintains a controllable risk threshold. Extensive experiments show that RCS significantly improves robustness and utility while maintaining latency comparable to CS. We hope this work contributes to the development of provably secure generative AI.

cs.CR↗

arXiv2025

ShadowBlock: Efficient Dynamic Anonymous Blocklisting and Its Cross-chain Application

Online harassment, incitement to violence, racist behavior, and other harmful content on social media can damage social harmony and even break the law. Traditional blocklisting technologies can block malicious users, but this comes at the expense of identity privacy. The anonymous blocklisting has emerged as an effective mechanism to restrict the abuse of freedom of speech while protecting user identity privacy. However, the state-of-the-art anonymous blocklisting schemes suffer from either poor dynamism or low efficiency. In this paper, we propose $\mathsf{ShadowBlock}$, an efficient dynamic anonymous blocklisting scheme. Specifically, we utilize the pseudorandom function and cryptographic accumulator to construct the public blocklisting, enabling users to prove they are not on the blocklisting in an anonymous manner. To improve verification efficiency, we design an aggregation zero-knowledge proof mechanism that converts multiple verification operations into a single one. In addition, we leverage the accumulator's property to achieve efficient updates of the blocklisting, i.e., the original proof can be reused with minimal updates rather than regenerating the entire proof. Experiments show that $\mathsf{ShadowBlock}$ has better dynamics and efficiency than the existing schemes. Finally, the discussion on applications indicates that $\mathsf{ShadowBlock}$ also holds significant value and has broad prospects in emerging fields such as cross-chain identity management.

cs.CR↗

arXiv2025

Can LLMs Threaten Human Survival? Benchmarking Potential Existential Threats from LLMs via Prefix Completion

Research on the safety evaluation of large language models (LLMs) has become extensive, driven by jailbreak studies that elicit unsafe responses. Such response involves information already available to humans, such as the answer to "how to make a bomb". When LLMs are jailbroken, the practical threat they pose to humans is negligible. However, it remains unclear whether LLMs commonly produce unpredictable outputs that could pose substantive threats to human safety. To address this gap, we study whether LLM-generated content contains potential existential threats, defined as outputs that imply or promote direct harm to human survival. We propose \textsc{ExistBench}, a benchmark designed to evaluate such risks. Each sample in \textsc{ExistBench} is derived from scenarios where humans are positioned as adversaries to AI assistants. Unlike existing evaluations, we use prefix completion to bypass model safeguards. This leads the LLMs to generate suffixes that express hostility toward humans or actions with severe threat, such as the execution of a nuclear strike. Our experiments on 10 LLMs reveal that LLM-generated content indicates existential threats. To investigate the underlying causes, we also analyze the attention logits from LLMs. To highlight real-world safety risks, we further develop a framework to assess model behavior in tool-calling. We find that LLMs actively select and invoke external tools with existential threats. Code and data are available at: https://github.com/cuiyu-ai/ExistBench.

cs.CR↗

arXiv2025

SE-Agent: Self-Evolution Trajectory Optimization in Multi-Step Reasoning with LLM-Based Agents

Large Language Model (LLM)-based agents have recently shown impressive capabilities in complex reasoning and tool use via multi-step interactions with their environments. While these agents have the potential to tackle complicated tasks, their problem-solving process, i.e., agents' interaction trajectory leading to task completion, remains underexploited. These trajectories contain rich feedback that can navigate agents toward the right directions for solving problems correctly. Although prevailing approaches, such as Monte Carlo Tree Search (MCTS), can effectively balance exploration and exploitation, they ignore the interdependence among various trajectories and lack the diversity of search spaces, which leads to redundant reasoning and suboptimal outcomes. To address these challenges, we propose SE-Agent, a Self-Evolution framework that enables Agents to optimize their reasoning processes iteratively. Our approach revisits and enhances former pilot trajectories through three key operations: revision, recombination, and refinement. This evolutionary mechanism enables two critical advantages: (1) it expands the search space beyond local optima by intelligently exploring diverse solution paths guided by previous trajectories, and (2) it leverages cross-trajectory inspiration to efficiently enhance performance while mitigating the impact of suboptimal reasoning paths. Through these mechanisms, SE-Agent achieves continuous self-evolution that incrementally improves reasoning quality. We evaluate SE-Agent on SWE-bench Verified to resolve real-world GitHub issues. Experimental results across five strong LLMs show that integrating SE-Agent delivers up to 55% relative improvement, achieving state-of-the-art performance among all open-source agents on SWE-bench Verified. Our code and demonstration materials are publicly available at https://github.com/JARVIS-Xs/SE-Agent.

cs.AI↗

arXiv2025

Free-MAD: Consensus-Free Multi-Agent Debate

Multi-agent debate (MAD) is an emerging approach to improving the reasoning capabilities of large language models (LLMs). Existing MAD methods rely on multiple rounds of interaction among agents to reach consensus, and the final output is selected by majority voting in the last round. However, this consensus-based design faces several limitations. First, multiple rounds of communication increases token overhead and limits scalability. Second, due to the inherent conformity of LLMs, agents that initially produce correct responses may be influenced by incorrect ones during the debate process, causing error propagation. Third, majority voting introduces randomness and unfairness in the decision-making phase, and can degrade the reasoning performance. To address these issues, we propose \textsc{Free-MAD}, a novel MAD framework that eliminates the need for consensus among agents. \textsc{Free-MAD} introduces a novel score-based decision mechanism that evaluates the entire debate trajectory rather than relying on the last round only. This mechanism tracks how each agent's reasoning evolves, enabling more accurate and fair outcomes. In addition, \textsc{Free-MAD} reconstructs the debate phase by introducing anti-conformity, a mechanism that enables agents to mitigate excessive influence from the majority. Experiments on eight benchmark datasets demonstrate that \textsc{Free-MAD} significantly improves reasoning performance while requiring only a single-round debate and thus reducing token costs. We also show that compared to existing MAD approaches, \textsc{Free-MAD} exhibits improved robustness in real-world attack scenarios.

cs.AI↗

arXiv2025

Aggregation Model and Market Mechanism for Virtual Power Plant Participation in Inertia and Primary Frequency Response

The declining provision of inertia by synchronous generators in modern power systems necessitates aggregating distributed energy resources (DERs) into virtual power plants (VPPs) to unlock their potential in delivering inertia and primary frequency response (IPFR) through ancillary service markets. To facilitate DER participation in the IPFR market, this paper proposes an aggregation model and market mechanism for VPPs participating in IPFR. First, an energy-reserve-IPFR market framework is developed, in which a VPP acts as an intermediary to coordinate heterogeneous DERs. Second, by taking into account the delay associated with inertial response, an optimization-based VPP aggregation method is introduced to encapsulate the IPFR process involving a variety of DERs. Third, an energy-reserve-IPFR market mechanism with VPP participation is introduced, aiming to minimize social costs, where stochastic deviations of renewable energy generation are explicitly modeled through chance-constrained reformulations, ensuring that the cleared energy, reserve, and IPFR schedules remain secure against forecast errors. Case studies on IEEE 30-bus and IEEE 118-bus systems show that the nadir and quasi-steady-state frequencies are reproduced by the VPP aggregation model with a mean absolute percentage error <= 0.03%, and the proposed market mechanism with VPP participation reduces the total system cost by approximately 40% and increases the net profit by about 30%.

econ.GN↗

arXiv2025

Ramp Up NTT in Record Time using GPU-Accelerated Algorithms and LLM-based Code Generation

Homomorphic encryption (HE) is a core building block in privacy-preserving machine learning (PPML), but HE is also widely known as its efficiency bottleneck. Therefore, many GPU-accelerated cryptographic schemes have been proposed to improve the performance of HE. However, these methods often require complex modifications tailored to specific algorithms and are tightly coupled with specific GPU and operating systems. It is interesting to ask how to generally offer more practical GPU-accelerated cryptographic algorithm implementations. Given the powerful code generation capabilities of large language models (LLMs), we aim to explore their potential to automatically generate practical GPU-friendly algorithm code using CPU-friendly code. In this paper, we focus on number theoretic transform (NTT) -- the core mechanism of HE. We first develop and optimize a GPU-friendly NTT (GNTT) family that exploits PyTorch's fast matrix computation and precomputation, achieving an approximately 62x speedup -- a significant boost over existing ones. Then we explore GPU-friendly code generation using various LLMs, including DeepSeek-R1, OpenAI o1 and o3-mini. We discover many interesting findings throughout the process. For instance, somewhat surprisingly, our experiments demonstrate that DeepSeek-R1 significantly outperforms OpenAI o3-mini and o1, but still cannot beat our optimized protocol. The findings provide valuable insights for turbocharging PPML and enhancing code generation capabilities of LLMs. Codes are available at: https://github.com/LMPC-Lab/GenGPUCrypto.

cs.CR↗

arXiv2024

Second-order topological insulator in Bilayer borophene

As the novel topological states, the higher-order topological insulators have attracted great attentions in the past years. However, their realizations in realistic materials, in particular in two dimensional systems, remains the big challenge due to the lack of adequate candidates. Here, based on the first-principle calculation and tight-binding model simulations, we identify the currently \emph{existing} bilayer $α_{5}$-phase borophenes as the two-dimensional second-order topological insulators, protected by the $C_{2}$-rotational symmetry. The formation of interlayer B-B covalent bonds, stabilizing the bilayer borophenes and opening the large direct bulk gaps ($\sim 0.55-0.62$ eV) at Fermi level, plays the key roles. The second-order topology is characterized by the bulk quantized quadrupole momentum. Our results enriches the candidates for the second-order topological insulators, and also provide a way to study topological states in borophenes.

cond-mat.mtrl-sci↗