arXiv subjects

Chang Jin

Publications and source records attributed to Chang Jin.

18 recordsLinked to original sources

arXiv2026

ClawSentry: A Progressive Multi-Tier Security Monitor for Safeguarding Autonomous LLM Agents

As large language model (LLM) agents move from conversation to executing code, reading local files, and orchestrating external tools, a single agent hijacked by a malicious third-party skill can cause data exfiltration, privilege escalation, or cascading compromise. We argue that agentic risk is progressive: it can enter at four loci of the agent control loop--skill admission, invocation-time intent, execution-time effect, and post-action consequence--while a denied dangerous objective can reappear across surface forms, tools, or turns; existing safeguards are typically local to one lifecycle boundary or one call. Guided by this threat model, we present ClawSentry, an open-source, framework-agnostic security supervision gateway for agent runtimes. Before a skill package is ever executed, First-use Skill Package Review (FSPR) audits it under a deterministic evidence floor, escalating unresolved cases to bounded read-only agentic review (locus A). At runtime, a three-tier progressive decision engine--a deterministic L1 layer, a rule-anchored L2 semantic reviewer, and a read-only L3 evidence-seeking agent--spends contextual review only on the residual ambiguity, while a session-level anti-bypass mechanism recognizes tool-switching and rephrased retries (loci B--C); a post-action path feeds high-severity evidence non-retroactively into later review (locus D). An Agent Harness Protocol (AHP) abstraction applies one policy across Codex, Claude Code, Kimi CLI, and Gemini CLI without modifying agent internals. On SkillInject with Codex/GPT-5.4, contextual ASR falls from 39.55% to 2.61% while contextual TSR moves only from 83.78% to 83.05%. Across five Work Agents on the full SkillsSafety benchmark, ClawSentry confines ASR to 9.09--15.03% from 33.5--49.7% unprotected, and aggregate TSR on clean skills remains 98.7%.

cs.CR↗

arXiv2026

SkillSafetyBench: Evaluating Agent Safety under Skill-Facing Attack Surfaces

Reusable skills are becoming a common interface for extending large language model agents, packaging procedural guidance with access to files, tools, memory, and execution environments. However, this modularity introduces attack surfaces that are largely missed by existing safety evaluations: even when the user request is benign, unsafe influence may reside in skill guidance, local artifacts, or execution-environment files that steer the agent toward unsafe actions. We present SkillSafetyBench, a runnable benchmark for evaluating such skill-facing safety failures. SkillSafetyBench includes 155 adversarial cases across 47 tasks, 6 risk domains, and 30 safety categories, each evaluated with a case-specific rule-based verifier. Experiments with multiple CLI agents and model backends show that non-user attacks can consistently induce unsafe behavior, with distinct failure patterns across domains, attack methods, and scaffold-model pairings. Our findings suggest that agent safety depends not only on model-level alignment, but also on how agents interpret skills, trust workflow context, and act through executable environments. The complete benchmark is available at https://github.com/AI45Lab/skill-safety-bench.

cs.CR↗

arXiv2026

OGER: A Robust Offline-Guided Exploration Reward for Hybrid Reinforcement Learning

Recent advancements in Reinforcement Learning with Verifiable Rewards (RLVR) have significantly improved Large Language Model (LLM) reasoning, yet models often struggle to explore novel trajectories beyond their initial policy distribution. While offline teacher guidance and entropy-driven strategies have been proposed to address this, they often lack deep integration or are constrained by the model's inherent capacity. In this paper, we propose OGER (Offline-Guided Exploration Reward), a novel framework that unifies offline teacher guidance and online reinforcement learning through a specialized reward modeling lens. OGER employs multi-teacher collaborative training and constructs an auxiliary exploration reward that leverages both offline trajectories and the model's own entropy to incentivize autonomous exploration. Extensive experiments across mathematical and general reasoning benchmarks demonstrate that OGER consistently outperforms competitive baselines, achieving substantial gains in mathematical reasoning while maintaining robust generalization to out-of-domain tasks. We provide a comprehensive analysis of training dynamics and conduct detailed ablation studies to validate the effectiveness of our entropy-aware reward modulation. Our code is available at https://github.com/ecoli-hit/OGER.git.

cs.AI↗

arXiv2026

TrinityGuard: A Unified Framework for Safeguarding Multi-Agent Systems

With the rapid development of LLM-based multi-agent systems (MAS), their significant safety and security concerns have emerged, which introduce novel risks going beyond single agents or LLMs. Despite attempts to address these issues, the existing literature lacks a cohesive safeguarding system specialized for MAS risks. In this work, we introduce TrinityGuard, a comprehensive safety evaluation and monitoring framework for LLM-based MAS, grounded in the OWASP standards. Specifically, TrinityGuard encompasses a three-tier fine-grained risk taxonomy that identifies 20 risk types, covering single-agent vulnerabilities, inter-agent communication threats, and system-level emergent hazards. Designed for scalability across various MAS structures and platforms, TrinityGuard is organized in a trinity manner, involving an MAS abstraction layer that can be adapted to any MAS structures, an evaluation layer containing risk-specific test modules, alongside runtime monitor agents coordinated by a unified LLM Judge Factory. During Evaluation, TrinityGuard executes curated attack probes to generate detailed vulnerability reports for each risk type, where monitor agents analyze structured execution traces and issue real-time alerts, enabling both pre-development evaluation and runtime monitoring. We further formalize these safety metrics and present detailed case studies across various representative MAS examples, showcasing the versatility and reliability of TrinityGuard. Overall, TrinityGuard acts as a comprehensive framework for evaluating and monitoring various risks in MAS, paving the way for further research into their safety and security.

cs.CR↗

arXiv2026

Heterogeneously Integrated Diamond-on-Lithium Niobate Quantum Photonic Platform

Diamond photonics has enabled efficient interfaces for quantum memories and is predicted to be a critical component of quantum networks. However, scalable network architectures require spatial, temporal, and spectral control of photons, which relies on nonlinear and electro-optic functionalities that diamond alone cannot provide. Here, we demonstrate heterogeneous integration of a thin-film lithium niobate (TFLN) platform, which has strong chi-2 nonlinearity and electro-optic effects, with thin diamond films. We demonstrate high-Q diamond photonic crystal cavities (Q factors exceeding 5x10^4 at 735 nm) that are lithographically aligned with TFLN photonic backbone and critically coupled to it. This allows us to realize low-loss diamond-TFLN "escalators" (loss ~1 dB/coupler) that support efficient light transfer between them. At cryogenic temperatures (5K), we can collect photons emitted from silicon vacancies (SiVs) embedded within the diamond structure via the TFLN photonic circuit. This approach establishes a scalable route toward integrated photonic circuits for practical quantum networking and other technologies.

physics.optics↗

arXiv2026

When Silence Is Golden: Can LLMs Learn to Abstain in Temporal QA and Beyond?

Large language models (LLMs) rarely admit uncertainty, often producing fluent but misleading answers, rather than abstaining (i.e., refusing to answer). This weakness is even evident in temporal question answering, where models frequently ignore time-sensitive evidence and conflate facts across different time-periods. In this paper, we present the first empirical study of training LLMs with an abstention ability while reasoning about temporal QA. Existing approaches such as calibration might be unreliable in capturing uncertainty in complex reasoning. We instead frame abstention as a teachable skill and introduce a pipeline that couples Chain-of-Thought (CoT) supervision with Reinforcement Learning (RL) guided by abstention-aware rewards. Our goal is to systematically analyze how different information types and training techniques affect temporal reasoning with abstention behavior in LLMs. Through extensive experiments studying various methods, we find that RL yields strong empirical gains on reasoning: a model initialized by Qwen2.5-1.5B-Instruct surpasses GPT-4o by $3.46\%$ and $5.80\%$ in Exact Match on TimeQA-Easy and Hard, respectively. Moreover, it improves the True Positive rate on unanswerable questions by $20\%$ over a pure supervised fine-tuned (SFT) variant. Beyond performance, our analysis shows that SFT induces overconfidence and harms reliability, while RL improves prediction accuracy but exhibits similar risks. Finally, by comparing implicit reasoning cues (e.g., original context, temporal sub-context, knowledge graphs) with explicit CoT supervision, we find that implicit information provides limited benefit for reasoning with abstention. Our study provides new insights into how abstention and reasoning can be jointly optimized, providing a foundation for building more reliable LLMs.

cs.CL↗

arXiv2025

A Suspended 4H-Silicon Carbide Membrane Platform for Defect Integration into Quantum Devices

4H-silicon carbide is a promising platform for solid-state quantum technology due to its commercial availability as a wide bandgap semiconductor and ability to host numerous spin-active color centers. Integrating color centers into suspended nanodevices enhances defect control and readout--key advances needed to fully harness their potential. However, challenges in developing robust fabrication processes for 4H-SiC thin films--due to the material's chemical and mechanical stability--limit their implementation in quantum applications. Here, we report on a new fabrication approach that first synthesizes suspended thin films from a monolithic platform, then patterns devices. With this technique, we fabricate and characterize structures tailored for defect integration, demonstrating 1D photonic crystal cavities, with and without waveguide interfaces, and lithium niobate on 4H-SiC acoustic cavities. This approach allows for greater fabrication flexibility--supporting high temperature annealing and heterogeneous material platform compatibility--providing a versatile platform for scalable fabrication of 4H-SiC devices for quantum technologies.

physics.app-ph↗

arXiv2025

Ultracoherent self-assembled diamond nanomechanics reveals superfluid dynamics

From gravitational-wave detection, protein force microscopy, to exploration of quantum-classical boundaries, many anticipated discoveries in fundamental science require improving measurement sensitivity limits. Through the fluctuation-dissipation theorem, mechanical dissipation sets the acoustic noise for this limit. Yet, even in high-purity crystals, the microscopic mechanisms responsible for the acoustic loss remain poorly understood. Tension-induced dissipation dilution offers a route to ultralow acoustic loss, but is challenging to implement in crystalline materials including single-crystal diamond. Here we realize a strain-engineered diamond nanomechanical platform using a liquid-assisted van der Waals self-assembly process that harnesses intrinsic surface forces to apply tensile stress exceeding 1 GPa. At cryogenic temperatures these resonators achieve quality factors beyond 10 billion (intrinsic material quality factors beyond 100 million). This exceptional coherence turns them into a sensitive probe for residual dissipation, elucidating three distinct two-level-system channels and one topological dissipation channel from a surface superfluid helium film. Our work shows how advancing mechanical coherence opens access to new regimes of physics in hybrid quantum systems, precision metrology, and condensed-matter physics.

cond-mat.mes-hall↗

arXiv2025

Purcell-enhanced spin-phonon coupling with a single color center

The radiative properties of atoms are inherently linked to their surrounding environment. Placing an electromagnetic resonator around atoms can enhance spontaneous emission, as shown by Purcell in the 1940s. This approach is now routinely used in quantum computing and communication to channel photons emitted by atoms into well-defined modes and control atom-photon interactions. For solid-state artificial atoms, such as color-centers, the host lattice introduces an acoustic environment, allowing excited atoms to relax by emitting phonons. Here we observe the acoustic Purcell effect by constructing a specially engineered, microwave-frequency nanomechanical resonator around a color-center spin qubit in diamond. Using a co-localized optical mode of the structure that strongly couples to the color-center's excited state, we perform single-photon-level laser spectroscopy at milliKelvin temperatures and observe ten-fold faster spin relaxation when the spin qubit is tuned into resonance with a 12 GHz acoustic mode. Additionally, we use the color-center as an atomic-scale probe to measure the broadband phonon spectrum of the nanostructure up to a frequency of 28 GHz. Our work establishes a new regime of control for quantum defects in solids and paves the way for interconnects between atomic-scale quantum memories and qubits encoded in acoustic and superconducting devices.

quant-ph↗

arXiv2025

Purcell-enhanced emissions from diamond color centers in slow light photonic crystal waveguides

Quantum memories based on emitters with optically addressable spins rely on efficient photonic interfaces, often implemented as nanophotonic cavities with ideally narrow spectral linewidths and small mode volumes. However, these approaches require nearly perfect spectral and spatial overlap between the cavity mode and quantum emitter, which can be challenging. This is especially true in the case of solid-state quantum emitters that are often randomly positioned and can suffer from significant inhomogeneous broadening. An alternative approach to mitigate these challenges is to use slow-light waveguides that can enhance light-matter interaction across large optical bandwidths and large areas. Here, we demonstrate diamond slow light photonic crystal (PhC) waveguides that enable broadband optical coupling to embedded silicon-vacancy (SiV) color centers. We take advantage of the recently demonstrated thin-film diamond photonic platform to fabricate fully suspended two-dimensional PhC waveguides. Using this approach, we demonstrate waveguide modes with high group indices up to 70 and observe Purcell-enhanced emissions of the SiVs coupled to the waveguide mode. Our approach represents a practical diamond platform for robust spin-photon interfaces with color centers.

quant-ph↗

arXiv2025

Probing negative differential resistance in silicon with a P-I-N diode-integrated T center ensemble

Solid-state defect quantum systems are exquisite probes of their local charge environment. Nonlinear dynamical electric fields in solids are challenging to characterize directly, conventionally limited to coarse macroscopic methods which fail to capture subtle effects in the material. Here, through transient optical spectroscopy on an embedded T center ensemble, we realize the in-situ observation of a silicon PIN-diode phase transition to a regime of self-sustained carrier oscillatory dynamics characteristic of negative differential resistance. Manifest in both the ensemble electroluminescence and photoluminescence, we find a temperature and field-dependent phase space for persistent undamped amplitude oscillations indicative of a collective ensemble response to the field dynamics. These findings shed new light on the cryogenic behavior of silicon, provide fundamental insight into the physics of the T center for improved quantum device performance, and open a promising new direction for defect-based local quantum sensing in semiconductor devices.

quant-ph↗

arXiv2024

M$^{3}$-20M: A Large-Scale Multi-Modal Molecule Dataset for AI-driven Drug Design and Discovery

This paper introduces M$^{3}$-20M, a large-scale Multi-Modal Molecule dataset that contains over 20 million molecules, with the data mainly being integrated from existing databases and partially generated by large language models. Designed to support AI-driven drug design and discovery, M$^{3}$-20M is 71 times more in the number of molecules than the largest existing dataset, providing an unprecedented scale that can highly benefit the training or fine-tuning of models, including large language models for drug design and discovery tasks. This dataset integrates one-dimensional SMILES, two-dimensional molecular graphs, three-dimensional molecular structures, physicochemical properties, and textual descriptions collected through web crawling and generated using GPT-3.5, offering a comprehensive view of each molecule. To demonstrate the power of M$^{3}$-20M in drug design and discovery, we conduct extensive experiments on two key tasks: molecule generation and molecular property prediction, using large language models including GLM4, GPT-3.5, GPT-4, and Llama3-8b. Our experimental results show that M$^{3}$-20M can significantly boost model performance in both tasks. Specifically, it enables the models to generate more diverse and valid molecular structures and achieve higher property prediction accuracy than existing single-modal datasets, which validates the value and potential of M$^{3}$-20M in supporting AI-driven drug design and discovery. The dataset is available at https://github.com/bz99bz/M-3.

q-bio.QM↗

arXiv2024

High-Q Cavity Interface for Color Centers in Thin Film Diamond

Quantum information technology offers the potential to realize unprecedented computational resources via secure channels capable of distributing entanglement between quantum computers. Diamond, as a host to atom-like defects with optically-accessible spin qubits, is a leading platform to realize quantum memory nodes needed to extend the reach of quantum links. Photonic crystal (PhC) cavities enhance light-matter interaction and are essential ingredients of an efficient interface between spins and photons that are used to store and communicate quantum information respectively. Despite great effort, however, the realization of visible PhC cavities with high quality factor (Q) and design flexibility is challenging in diamond. Here, we demonstrate one- and two-dimensional PhC cavities fabricated in recently developed thin-film diamonds, featuring Q-factors of 1.8x10$^5$ and 1.6x10$^5$, respectively, the highest Qs for visible PhC cavities realized in any material. Importantly, our fabrication process is simple and high-yield, based on conventional planar fabrication techniques, in contrast to previous approaches that rely on complex undercut methods. We also demonstrate fiber-coupled 1D PhC cavities with high photon extraction efficiency, and optical coupling between a single SiV center and such a cavity at 4K achieving a Purcell factor of 13. The demonstrated diamond thin-film photonic platform will improve the performance and scalability of quantum nodes and expand the range of quantum technologies.

quant-ph↗

arXiv2023

Engineering Phonon-Qubit Interactions using Phononic Crystals

The ability to control phonons in solids is key for diverse quantum applications, ranging from quantum information processing to sensing. Often, phonons are sources of noise and decoherence, since they can interact with a variety of solid-state quantum systems. To mitigate this, quantum systems typically operate at milli-Kelvin temperatures to reduce the number of thermal phonons. Here we demonstrate an alternative approach that relies on engineering phononic density of states, drawing inspiration from photonic bandgap structures that have been used to control the spontaneous emission of quantum emitters. We design and fabricate diamond phononic crystals with a complete phononic bandgap spanning 50 - 70 gigahertz, tailored to suppress interactions of a single silicon-vacancy color center with resonant phonons of the thermal bath. At 4 Kelvin, we demonstrate a reduction of the phonon-induced orbital relaxation rate of the color center by a factor of 18 compared to bulk. Furthermore, we show that the phononic bandgap can efficiently suppress phonon-color center interactions up to 20 Kelvin. In addition to enabling operation of quantum memories at higher temperatures, the ability to engineer qubit-phonon interactions may enable new functionalities for quantum science and technology, where phonons are used as carriers of quantum information.

quant-ph↗

arXiv2023

Deterministic Creation of Strained Color Centers in Nanostructures via High-Stress Thin Films

Color centers have emerged as a leading qubit candidate for realizing hybrid spin-photon quantum information technology. One major limitation of the platform, however, is that the characteristics of individual color-centers are often strain dependent. As an illustrative case, the silicon-vacancy center in diamond typically requires millikelvin temperatures in order to achieve long coherence properties, but strained silicon vacancy centers have been shown to operate at temperatures beyond 1K without phonon-mediated decoherence. In this work we combine high-stress silicon nitride thin films with diamond nanostructures in order to reproducibly create statically strained silicon-vacancy color centers (mean ground state splitting of 608 GHz) with strain magnitudes of $\sim 4 \times 10^{-4}$. Based on modeling, this strain should be sufficient to allow for operation of a majority silicon-vacancy centers within the measured sample at elevated temperatures (1.5K) without any degradation of their spin properties. This method offers a scalable approach to fabricate high-temperature operation quantum memories. Beyond silicon-vacancy centers, this method is sufficiently general that it can be easily extended to other platforms as well.

quant-ph↗

arXiv2022

Two-dimensional spin systems in PECVD-grown diamond with tunable density and long coherence for enhanced quantum sensing and simulation

Systems of spins engineered with tunable density and reduced dimensionality enable a number of advancements in quantum sensing and simulation. Defects in diamond, such as nitrogen-vacancy (NV) centers and substitutional nitrogen (P1 centers), are particularly promising solid-state platforms to explore. However, the ability to controllably create coherent, two-dimensional spin systems and characterize their properties, such as density, depth confinement, and coherence is an outstanding materials challenge. We present a refined approach to engineer dense ($\gtrsim$1 ppm$\cdot$nm), 2D nitrogen and NV layers in diamond using delta-doping during plasma-enhanced chemical vapor deposition (PECVD) epitaxial growth. We employ both traditional materials techniques, e.g. secondary ion mass spectrometry (SIMS), alongside NV spin decoherence-based measurements to characterize the density and dimensionality of the P1 and NV layers. We find P1 densities of 5-10 ppm$\cdot$nm, NV densities between 1 and 3.5 ppm$\cdot$nm tuned via electron irradiation dosage, and depth confinement of the spin layer down to 1.6 nm. We also observe high (up to 0.74) ratios of P1 to NV centers and reproducibly long NV coherence times, dominated by dipolar interactions with the engineered P1 and NV spin baths.

cond-mat.mtrl-sci↗

arXiv2022

AdMix: A Mixed Sample Data Augmentation Method for Neural Machine Translation

In Neural Machine Translation (NMT), data augmentation methods such as back-translation have proven their effectiveness in improving translation performance. In this paper, we propose a novel data augmentation approach for NMT, which is independent of any additional training data. Our approach, AdMix, consists of two parts: 1) introduce faint discrete noise (word replacement, word dropping, word swapping) into the original sentence pairs to form augmented samples; 2) generate new synthetic training data by softly mixing the augmented samples with their original samples in training corpus. Experiments on three translation datasets of different scales show that AdMix achieves signifi cant improvements (1.0 to 2.7 BLEU points) over strong Transformer baseline. When combined with other data augmentation techniques (e.g., back-translation), our approach can obtain further improvements.

cs.CL↗

arXiv2021

Design and verification of the HXI collimator onboard the ASO-S mission

A space-borne hard X-ray collimator, comprising 91 pairs of grids, has been developed for the Hard X-ray Imager (HXI). The HXI is one of the three scientific instruments onboard the first Chinese solar mission: the Advanced Space-based Solar Observatory (ASO-S). The HXI collimator (HXI-C) is a spatial modulation X-ray telescope designed to observe hard X-rays emitted by energetic electrons in solar flares. This paper presents the detailed design of the HXI-C for the qualification model that will be inherited by the flight model. Series tests on the HXI-C qualification model are reported to verify the ability of the HXI-C to survive the launch and to operate normally in on-orbit environments. Furthermore, results of the X-ray beam test for the HXI-C are presented to indirectly identify the working performance of the HXI-C.

astro-ph.IM↗