BZPEERReading, trust and collaboration for researchDiscover
Workspaces
Network
Opportunities
Account
Researcher profile
Yu Liu contributes to research discovery and scholarly infrastructure.
Trust snapshot
Actions
Identity and collaboration
Claiming links this public author record to a researcher profile and unlocks direct collaboration workflows.
Log in to claimDirect collaboration
Claim this author entity first to unlock direct invitations.
Research graph
Inspect adjacent work, topics, institutions and collaborators without jumping out to a separate graph page.
BZPEER is loading the nearby papers, people, topics and institutions for this page.
Published work
This work presents a 28nm 13.93mm2 CNN-Transformer accelerator for semantic segmentation, achieving 3.86-to-10.91x energy reduction over previous designs. It features a hybrid attention unit, layer-fusion scheduler, and cascaded feature-map pruner, with peak energy efficiency of 52.90TOPS/W (INT8).
Scientific and engineering verticals often suffer from data scarcity and strict executability requirements: models must generate not only fluent text, but also syntactically valid, tool-compilable scripts. We present a schema-first alignment framework for building compact, executable domain-specific LLMs in low-resource settings. The framework integrates three core components: (i) large-scale synthetic QA data generation from expert documentation to instill foundational domain knowledge; (ii) a code-centric IR->DPO workflow that converts verified tool decks into interpretable intermediate representations (IR), performs equivalence-preserving diversification, and constructs preference pairs to directly optimize instruction compliance and code executability; and (iii) a controlled evaluation of Retrieval-Augmented Generation (RAG), showing that while RAG benefits general LLMs, it can marginally degrade the performance of already domain-aligned models. We demonstrate the framework by instantiating TcadGPT for semiconductor Technology Computer-Aided Design (TCAD). Using 1.5M synthetic QA pairs and an IR-driven DPO dataset, TcadGPT attains 85.6% semantic accuracy and an 80.0% syntax pass rate on SDE executability tests, substantially outperforming state-of-the-art general LLMs such as GPT-4o. To probe portability beyond TCAD, we apply the same recipe to the open-source FEM solver Elmer, observing consistent improvements in script-level success rates over general-purpose baselines. All datasets, benchmarks, and code (including P1, P2, and IR->DPO) are released for reproducibility. Together, these results suggest that the proposed framework provides a robust and reproducible path toward executable LLMs in specialized, data-scarce professional domains.
Multimodal emotion understanding requires effective integration of text, audio, and visual modalities for both discrete emotion recognition and continuous sentiment analysis. We present EGMF, a unified framework combining expert-guided multimodal fusion with large language models. Our approach features three specialized expert networks--a fine-grained local expert for subtle emotional nuances, a semantic correlation expert for cross-modal relationships, and a global context expert for long-range dependencies--adaptively integrated through hierarchical dynamic gating for context-aware feature selection. Enhanced multimodal representations are integrated with LLMs via pseudo token injection and prompt-based conditioning, enabling a single generative framework to handle both classification and regression through natural language generation. We employ LoRA fine-tuning for computational efficiency. Experiments on bilingual benchmarks (MELD, CHERMA, MOSEI, SIMS-V2) demonstrate consistent improvements over state-of-the-art methods, with superior cross-lingual robustness revealing universal patterns in multimodal emotional expressions across English and Chinese. We will release the source code publicly.
Large Language Models (LLMs) serve as the backbone of modern AI systems, yet they remain susceptible to adversarial jailbreak attacks. Consequently, robust detection of such malicious inputs is paramount for ensuring model safety. Traditional detection methods typically rely on external models trained on fixed, large-scale datasets, which often incur significant computational overhead. While recent methods shift toward leveraging internal safety signals of models to enable more lightweight and efficient detection. However, these methods remain inherently static and struggle to adapt to the evolving nature of jailbreak attacks. Drawing inspiration from the biological immune mechanism, we introduce the Immune Memory Adaptive Guard (IMAG) framework. By distilling and encoding safety patterns into a persistent, evolvable memory bank, IMAG enables adaptive generalization to emerging threats. Specifically, the framework orchestrates three synergistic components: Immune Detection, which employs retrieval for high-efficiency interception of known jailbreak attacks; Active Immunity, which performs proactive behavioral simulation to resolve ambiguous unknown queries; Memory Updating, which integrates validated attack patterns back into the memory bank. This closed-loop architecture transitions LLM defense from rigid filtering to autonomous adaptive mitigation. Extensive evaluations across five representative open-source LLMs demonstrate that our method surpasses state-of-the-art (SOTA) baselines, achieving a superior average detection accuracy of 94\% across diverse and complex attack types.
Recent advancements adopt online reinforcement learning (RL) from LLMs to text-to-image rectified flow diffusion models for reward alignment. The use of group-level rewards successfully aligns the model with the targeted reward. However, it faces challenges including low efficiency, dependency on stochastic samplers, and reward hacking. The problem is that rectified flow models are fundamentally different from LLMs: 1) For efficiency, online image sampling takes much more time and dominates the time of training. 2) For stochasticity, rectified flow is deterministic once the initial noise is fixed. Aiming at these problems and inspired by the effects of group-level rewards from LLMs, we design Group-level Direct Reward Optimization (GDRO). GDRO is a new post-training paradigm for group-level reward alignment that combines the characteristics of rectified flow models. Through rigorous theoretical analysis, we point out that GDRO supports full offline training that saves the large time cost for image rollout sampling. Also, it is diffusion-sampler-independent, which eliminates the need for the ODE-to-SDE approximation to obtain stochasticity. We also empirically study the reward hacking trap that may mislead the evaluation, and involve this factor in the evaluation using a corrected score that not only considers the original evaluation reward but also the trend of reward hacking. Extensive experiments demonstrate that GDRO effectively and efficiently improves the reward score of the diffusion model through group-wise offline optimization across the OCR and GenEval tasks, while demonstrating strong stability and robustness in mitigating reward hacking.
We investigate the sensitivity of calcium production to nuclear reaction rates of a 40 solar-mass Population III star using 1D multi-zone stellar models. A comprehensive nuclear reaction network was constructed, and all $(p,γ)$ and $(p,α)$ reaction rates were individually varied by a factor of 10 up and down, identifying 13 preliminary key reactions for calcium production. To propagate the reaction rate uncertainties on calcium production, two sets of Monte Carlo simulations were performed for these key reactions: one adopting STARLIB reaction rates and the other incorporating updated rates from recent experimental data and evaluations. Our results show that Monte Carlo simulations using the updated rates show good agreement with the observed calcium abundance of the extremely iron-poor star SMSS J031300.36-670839.3 within the 68% confidence interval predicted by the models. In contrast, the observed calcium abundance lies marginally outside the 68% C.I. when using the STARLIB rates. Spearman rank-order correlation analysis and SHAP values show that the $(p,γ)$ and $(p,α)$ reactions of F18 and F19 exhibit strong coupled effects on calcium production. These reaction-rate uncertainties need to be reduced to constrain the stellar model predictions. Our study provides insights for future nuclear physics experiments aimed at reducing reaction rate uncertainties in the nucleosynthesis of Population III Stars. Additionally, comparisons between 20 solar-mass and 40 solar-mass Population III stellar models confirm that the latter, with updated reaction rates, is more capable of reproducing the observed Ca abundance and [Ca/Mg] ratio.
Infrared and visible (IR-VIS) image fusion has gained significant attention for its broad application value. However, existing methods often neglect the complementary role of infrared image in restoring visible image features under hazy conditions. To address this, we propose a joint learning framework that utilizes infrared image for the restoration and fusion of hazy IR-VIS images. To mitigate the adverse effects of feature diversity between IR-VIS images, we introduce a prompt generation mechanism that regulates modality-specific feature incompatibility. This creates a prompt selection matrix from non-shared image information, followed by prompt embeddings generated from a prompt pool. These embeddings help generate candidate features for dehazing. We further design an infrared-assisted feature restoration mechanism that selects candidate features based on haze density, enabling simultaneous restoration and fusion within a single-stage framework. To enhance fusion quality, we construct a multi-stage prompt embedding fusion module that leverages feature supplementation from the prompt generation module. Our method effectively fuses IR-VIS images while removing haze, yielding clear, haze-free fusion results. In contrast to two-stage methods that dehaze and then fuse, our approach enables collaborative training in a single-stage framework, making the model relatively lightweight and suitable for practical deployment. Experimental results validate its effectiveness and demonstrate advantages over existing methods. The source code of the paper is available at \href{https://github.com/fangjiaqi0909/IASSF}{\textcolor{blue}{https://github.com/fangjiaqi0909/IASSF
Video generation is rapidly evolving from single-shot synthesis to complex multi-shot audio-video (MSAV) narratives to meet real-world demands. However, evaluating such frontier models remains a fundamental challenge. Existing benchmarks are limited in scope and data diversity, and rely on rigid evaluation pipelines, preventing systematic and reliable assessment of modern MSAV models. To bridge these gaps, we introduce MSAVBench, the first comprehensive benchmark and adaptive hybrid evaluation framework for multi-shot audio-video generation. Our benchmark spans four key dimensions, video, audio, shot, and reference, covering diverse task settings, varying shot counts of up to 15, and challenging non-realistic scenarios. Our evaluation framework improves robustness through an adaptive self-correction mechanism for shot segmentation, instance-wise rubrics for subjective metrics, and tool-grounded evidence extraction for complex judgments. Furthermore, MSAVBench achieves high alignment with human judgments, reaching a Spearman rank correlation of 91.5%. Our systematic evaluation of 19 state-of-the-art closed- and open-source models shows that current systems still struggle with director-level control and fine-grained audio-visual synchronization, while modular or agentic generation pipelines offer a promising path toward narrowing the gap between open- and closed-source models. We will release the benchmark data and evaluation code to facilitate future research.
Answering real-world open-domain multi-hop questions over massive corpora is a critical challenge in Retrieval-Augmented Generation (RAG) systems. Recent research employs reinforcement learning (RL) to end-to-end optimize the retrieval-augmented reasoning process, directly enhancing its capacity to resolve complex queries. However, reliable deployment is hindered by two obstacles. 1) Retrieval Collapse: iterative retrieval over large corpora fails to locate intermediate evidence containing bridge answers without reasoning-guided planning, causing downstream reasoning to collapse. 2) Learning Instability: end-to-end trajectory training suffers from weak credit assignment across reasoning chains and poor error localization across modules, causing overfitting to benchmark-specific heuristics that limit transferability and stability. To address these problems, we propose PRISMA, a decoupled RL-guided framework featuring a Plan-Retrieve-Inspect-Solve-Memoize architecture. PRISMA's strength lies in reasoning-guided collaboration: the Inspector provides reasoning-based feedback to refine the Planner's decomposition and fine-grained retrieval, while enforcing evidence-grounded reasoning in the Solver. We optimize individual agent capabilities via Two-Stage Group Relative Policy Optimization (GRPO). Stage I calibrates the Planner and Solver as specialized experts in planning and reasoning, while Stage II utilizes Observation-Aware Residual Policy Optimization (OARPO) to enhance the Inspector's ability to verify context and trigger targeted recovery. Experiments show that PRISMA achieves state-of-the-art performance on ten benchmarks and can be deployed efficiently in real-world scenarios.
Coding agents powered by large language models are becoming central modules of modern IDEs, helping users perform complex tasks by invoking tools. While powerful, tool invocation opens a substantial attack surface. Prior work has demonstrated attacks against general-purpose and domain-specific agents, but none have focused on the security risks of tool invocation in coding agents. To fill this gap, we conduct the first systematic red-teaming of six popular real-world coding agents: Cursor, Claude Code, Copilot, Windsurf, Cline, and Trae. Our red-teaming proceeds in two phases. In Phase 1, we perform prompt leakage reconnaissance to recover system prompts. We discover a general vulnerability, ToolLeak, which allows malicious prompt exfiltration through benign argument retrieval during tool invocation. In Phase 2, we hijack the agent's tool-invocation behavior using a novel two-channel prompt injection in the tool description and return values, achieving remote code execution (RCE). We adaptively construct payloads using security information leaked in Phase 1. In emulation across five backends, our method outperforms baselines on Claude-Sonnet-4, Claude-Sonnet-4.5, Grok-4, and GPT-5. On real agents, our approach succeeds on 19 of 25 agent-LLM pairs, achieving leakage on every agent using Claude and Grok backends. For tool-invocation hijacking, we obtain RCE on every tested agent-LLM pair, with our two-channel method delivering the highest success rate. We provide case studies on Cursor and Claude Code, analyze security guardrails of external and built-in tools, and conclude with practical defense recommendations.
The origin of the strange metallic behavior observed in a wide range of quantum materials is an open challenge to condensed matter physics. Historically, strange metals were uniquely associated with antiferromagnetic quantum critical points (QCPs), but a new generation of materials reveals their association with uniform order parameters, such as ferromagnetism, valley or nematic order, suggesting a deeper common denominator. At a QCP, order parameter fluctuations are characterized by the dynamical critical exponent $z$, which quantifies the space-time scaling asymmetry. Here, we report the observation of a divergence in the Grüneisen ratio at the QCP of the strange-metal ferromagnet CeRh$_6$Ge$_4$ with a dynamical critical exponent $z=3$, signaling that the underlying quantum singularity involves a conserved degree of freedom. Yet the magnetization of this easy-plane ferromagnet is not conserved. We argue that the $z=3$ strange criticality requires a description beyond the Landau paradigm, proposing a link with the gauge modes of the small-to-large Fermi surface transition and the associated gauge charge of the delocalizing heavy electrons.