BZBiao Zhao

Biao Zhao

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Biao Zhao has a total of 53 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Verifiability-Aware Fact Verification: Leveraging Weakly Supervised Task Transfer from Rumor Detection for Reliable Claim Assessment

    by Yang Gao, Weiqiang Jin, Yang Liu, Shixiang Tang, Yanghao Zhou, Biao Zhao

    The rapid growth of online information has brought rumor detection and automated fact verification into increasing focus. Fact-verification systems typically assume that every input claim can beresolved as supported, refuted, or lacking sufficient evidence. Yet real-world claims are often vague, subjective, or unfalsifiable, making conventional factual verification ill-defined or potentially misleading. We argue that claim verifiability is a crucial but overlooked prerequisite for reliable fact verification. Because direct verifiability annotations are scarce, we use rumor-detection corpora as a source of weak supervision and construct high-confidence pseudo-labels with a deterministic cue-and-filter rule that keeps factuality and verifiability conceptually distinct. We introduce Claim Verifiability Prediction based Rumor detection for fact verification (CVPR-FV), an explicit extension of Det2Ver that adds Claim Verifiability Prediction as an auxiliary task and integrates its output with decomposition based verification through a heuristic score-fusion rule, without requiring additional human-labelled training data. Experiments on FEVER, VitaminC, and SciFACT show that verifiability awareness generally improves few-shot and zero-shot fact verification, with the clearest benefits in low-resource and conflict-prone settings. Further analysis shows that low-verifiability claims are more likely to produce conflicting intermediate judgments and that CVPR-FV mitigates these conflicts while improving not-enough-information decisions. These findings highlight the importance of modelling claim verifiability explicitly and open new directions for building more reliable and interpretable fact verification systems. Code is available at GitHub: https://github.com/albert-jin/CVPR-fv/.

  • Open Access

    InterWeave Reasoning: Adaptive Alternation between Latent Exploration and Symbolic Grounding in Large Language Models

    by Weiqiang Jin, Jing Gao, Shixiang Tang, Yang Liu, Jinhu Qi, Wentao Zhang, Junli Wang, Biao Zhao, Guan Yang

    Large language models (LLMs) typically reason in a discrete token space, which provides stability but forces each step to collapse a rich internal belief distribution into a single symbolic output. Recent continuous reasoning methods relax this bottleneck by propagating latent semantic representations, but they often suffer from representation drift, grounding failure, and even non-terminating reasoning trajectories when operating beyond the pre-training distribution. To address this challenge, we introduce InterWeave Reasoning, a framework that enables a frozen LLM to adaptively switch between latent exploration and symbolic grounding during multi-step reasoning. Instead of statically combining the two modes, our method formulates mode selection as a sequential decision problem and learns a lightweight controller with reinforcement learning (RL) algorithms. To further improve training stability, we develop two complementary strategies: Failure-Aware Replay, which emphasizes error-prone trajectories such as representation drift and early collapse, and Exploration-Biased Initialization, which encourages sufficient latent exploration before policy refinement. Extensive experiments demonstrate that InterWeave Reasoning consistently improves both reasoning performance and efficiency. On GSM8K with Llama-3.1-8B, it achieves 86.43, outperforming Soft-Thinking by +3.11. On GPQA, it reaches 71.72, surpassing Soft-Thinking on QwQ-32B by +4.55. In addition, it significantly reduces reasoning cost, shortening the average trajectory length from 3875 to 2127 tokens across datasets on DeepSeek-32B. Notably, these gains are achieved with a lightweight 0.1M-parameter MLP controller, demonstrating strong efficiency without increasing model complexity. These suggest that the proposed InterWeave Reasoning provides an effective and robust paradigm for complex LLM reasoning.

  • Open Access

    A Comprehensive Review of Complex Logical Reasoning in Large Vision-Language Models

    by Weiqiang Jin, Yang Liu, Yang Gao, Shixiang Tang, Yanghao Zhou, Jinhu Qi, Wentao Zhang, Junli Wang, Jing Gao, Yue Ma, et al.

    Large Vision-Language Models (LVLMs) have achieved strong performance in multimodal perception, understanding, and generation, but their ability to perform complex logical reasoning remains insufficiently understood. In particular, it is still unclear whether current LVLMs can reliably conduct explicit logical operations, multi-step inference, abstract relational reasoning, and cross-modal evidence integration. Reasoning abilities such as deductive, inductive, abductive, multi-hop, and causal inference are fundamental to robust decision making, trustworthy interaction, and real-world deployment, yet they have not been systematically examined in the LVLM literature. Existing surveys mainly discuss mathematical reasoning, general multimodal intelligence, or benchmark progress, but they do not provide a unified account of complex logical reasoning in LVLMs, including its definition, reasoning types, modeling paradigms, evaluation protocols, and unresolved limitations. To address this gap, this survey develops a unified analytical framework for complex logical reasoning in LVLMs. This survey provides a structured review of this emerging area. We first formalize complex logical reasoning in multimodal settings and organize the literature into five recurrent reasoning families: deductive, inductive, abductive, multi-hop, and causal reasoning. We then review reasoning-oriented LVLM architectures, including unified, modular, and tool-augmented paradigms, and summarize major reasoning mechanisms such as chain-of-thought, program-based reasoning, self-correction, and interpretability-oriented analysis. We further examine representative benchmarks and evaluation protocols, with particular attention to the mismatch between final-answer accuracy and genuine reasoning validity. Based on empirical evidence from representative LVLMs and datasets, we identify common capability trends, recurring failure modes, and key open challenges. Our analysis shows that current LVLMs still struggle with reasoning faithfulness, long-horizon inference, cross-modal grounding, hallucination control, and process-aware evaluation. Finally, we outline future directions in reasoning-oriented data construction, model design, training strategies, evaluation methodology, and deployment. Overall, this survey offers a unified conceptual framework and technical roadmap for advancing LVLMs from strong perceptual systems toward reliable multimodal reasoning agents.


  • Self‐Adaptive LLM Instructions Optimization for Aspect‐Based Sentiment Analysis by Incorporating Emotion‐Oriented In‐Contexts

    by Weiqiang Jin, Junli Wang, Yang Gao, Bohang Shi, Ningwei Wang, Biao Zhao, Guang Yang

    Aspect‐based Sentiment Analysis (ABSA) is a vital NLP task that identifies sentiment towards specific entities or aspect terms within a text. Recently, large language models (LLMs) have shown impressive capabilities in semantic comprehension and logical inference. However, LLM hallucinations pose challenges in accurately determining sentiment polarity for aspect terms, leading to performance issues. Moreover, current ABSA methods often fail to fully leverage the vast prior knowledge embedded within LLMs, resulting in suboptimal classification outcomes for specific aspects. Inspired by these challenges, we propose the BYD‐OBS‐ABSA framework—‘Beyond Simple Observations, Embracing Comprehensive Contextual Insights’ for ABSA tasks. This framework leverages unique in‐context constraints, backgrounds, and analogical reasoning to address LLM hallucinations and uses self‐adaptive bootstrap instructions optimization to enhance LLM predictions. BYD‐OBS‐ABSA integrates various in‐context augmentation strategies, including emotion‐oriented backgrounds, constraints, and analogical reasoning. BYD‐OBS‐ABSA further improves initial LLM instructions through adaptive iterative optimization using a random search bootstrap algorithm, maximizing the benefits of LLM prompting. Extensive zero/few‐shot experiments with GPT‐3.5‐turbo across six public datasets validate the effectiveness and robustness of our framework, even surpassing human judgment in certain scenarios.


  • Open Access

    Cross-Domain Knowledge Transfer in Large Models: Transforming Knowledge Management Dynamics

    by Weiqiang Jin, Biao Zhao

    In today’s rapidly evolving technological landscape, large models—particularly large language models—are reshaping the dynamics of knowledge management. In this chapter, we explore how these advanced models facilitate cross-domain knowledge transfer—a critical capability in an era marked by increasingly interconnected global challenges and data overload—enabling the integration of information from diverse research fields. By leveraging the capabilities of large language models (LLMs), organizations can break down silos, foster innovation, and enhance decision-making processes. We systematically examine the mechanisms through which cross-domain knowledge transfer occurs in large language models, discuss the challenges such as data privacy and model bias, and highlight the implications for businesses operating in transformative environments. Case studies and current research provide insights into optimizing knowledge management strategies to harness the full potential of artificial intelligence in a multidisciplinary context. Overall, we hope that this chapter provides a foundational framework for scholars studying human-artificial intelligence (AI) collaboration and offers practitioners criteria to evaluate LLM adoption risks. We believe the findings in this chapter are critical for organizations seeking to balance agility with ethical accountability in the upcoming artificial general intelligence (AGI) era.


  • Representation-driven sampling and adaptive policy resetting for improving multi-Agent reinforcement learning

    by Weiqiang Jin, Xingwu Tian, Ningwei Wang, Baohai Wu, Bohang Shi, Biao Zhao, Guang Yang


  • Open Access

    A Comprehensive Survey on Multi-Agent Cooperative Decision-Making: Scenarios, Approaches, Challenges and Perspectives

    by Weiqiang Jin, Hongyang Du, Biao Zhao, Xingwu Tian, Bohang Shi, Guang Yang

    With the rapid development of artificial intelligence, intelligent decision-making techniques have gradually surpassed human levels in various human-machine competitions, especially in complex multi-agent cooperative task scenarios. Multi-agent cooperative decision-making involves multiple agents working together to complete established tasks and achieve specific objectives. These techniques are widely applicable in real-world scenarios such as autonomous driving, drone navigation, disaster rescue, and simulated military confrontations. This paper begins with a comprehensive survey of the leading simulation environments and platforms used for multi-agent cooperative decision-making. Specifically, we provide an in-depth analysis for these simulation environments from various perspectives, including task formats, reward allocation, and the underlying technologies employed. Subsequently, we provide a comprehensive overview of the mainstream intelligent decision-making approaches, algorithms and models for multi-agent systems (MAS). Theseapproaches can be broadly categorized into five types: rule-based (primarily fuzzy logic), game theory-based, evolutionary algorithms-based, deep multi-agent reinforcement learning (MARL)-based, and large language models(LLMs)reasoning-based. Given the significant advantages of MARL andLLMs-baseddecision-making methods over the traditional rule, game theory, and evolutionary algorithms, this paper focuses on these multi-agent methods utilizing MARL and LLMs-based techniques. We provide an in-depth discussion of these approaches, highlighting their methodology taxonomies, advantages, and drawbacks. Further, several prominent research directions in the future and potential challenges of multi-agent cooperative decision-making are also detailed.