Feifei Zhao

Zeyang Yue, Chenfei Yan, Feifei Zhao et al.

Whether Large Language Models (LLMs) exhibit covert psychological manipulation in complex human-AI interactions has garnered increasing safety concerns. However, existing AI safety benchmarks remain largely restricted to explicit rule compliance and static prompts, failing to capture the dynamic and covert nature of manipulative strategies in multi-turn dialogues. We introduce CogManip, a comprehensive benchmark that evaluates 15 manipulation strategy risks across 1,000 multi-turn interaction scenarios, validated by human experts. A systematic evaluation of 13 representative models, including frontier models like GPT-5.4 and DeepSeek-V3.2, reveals significant risk heterogeneities and illuminates the targeted direction for future defense. Further analysis of objective function perturbation reveals that DeepSeek-V3.2's manipulation tactics are highly sensitive to both negative and benign system prompts, demonstrating the critical necessity of prompt-based defense engineering and implicit goal auditing. CogManip offers a robust instrument and perspective for auditing the implicit psychological influence and dynamic strategy selection of modern LLMs.

Sicheng Shen, Mingyang Lv, Han Shen et al.

The safety of large language models (LLMs) has increasingly emerged as a fundamental aspect of their development. Existing safety alignment for LLMs is predominantly achieved through post-training methods, which are computationally expensive and often fail to generalize well across different models. A small number of lightweight alignment approaches either rely heavily on prior-computed safety injections or depend excessively on the model's own capabilities, resulting in limited generalization and degraded efficiency and usability during generation. In this work, we propose a safety-aware decoding method that requires only low-cost training of an expert model and employs a single neuron as a gating mechanism. By effectively balancing the model's intrinsic capabilities with external guidance, our approach simultaneously preserves utility and enhances output safety. It demonstrates clear advantages in training overhead and generalization across model scales, offering a new perspective on lightweight alignment for the safe and practical deployment of large language models. Code: https://github.com/Beijing-AISI/NGSD.

Bing Han, Feifei Zhao, Yi Zeng et al.

Children possess the ability to learn multiple cognitive tasks sequentially, which is a major challenge toward the long-term goal of artificial general intelligence. Existing continual learning frameworks are usually applicable to Deep Neural Networks (DNNs) and lack the exploration on more brain-inspired, energy-efficient Spiking Neural Networks (SNNs). Drawing on continual learning mechanisms during child growth and development, we propose Dynamic Structure Development of Spiking Neural Networks (DSD-SNN) for efficient and adaptive continual learning. When learning a sequence of tasks, the DSD-SNN dynamically assigns and grows new neurons to new tasks and prunes redundant neurons, thereby increasing memory capacity and reducing computational overhead. In addition, the overlapping shared structure helps to quickly leverage all acquired knowledge to new tasks, empowering a single network capable of supporting multiple incremental tasks (without the separate sub-network mask for each task). We validate the effectiveness of the proposed model on multiple class incremental learning and task incremental learning benchmarks. Extensive experiments demonstrated that our model could significantly improve performance, learning speed and memory capacity, and reduce computational overhead. Besides, our DSD-SNN model achieves comparable performance with the DNNs-based methods, and significantly outperforms the state-of-the-art (SOTA) performance for existing SNNs-based continual learning methods.

Bing Han, Feifei Zhao, Yi Zeng et al.

Spiking Neural Networks (SNNs) are more biologically plausible and computationally efficient. Therefore, SNNs have the natural advantage of drawing the sparse structural plasticity of brain development to alleviate the energy problems of deep neural networks caused by their complex and fixed structures. However, previous SNNs compression works are lack of in-depth inspiration from the brain development plasticity mechanism. This paper proposed a novel method for the adaptive structural development of SNN (SD-SNN), introducing dendritic spine plasticity-based synaptic constraint, neuronal pruning and synaptic regeneration. We found that synaptic constraint and neuronal pruning can detect and remove a large amount of redundancy in SNNs, coupled with synaptic regeneration can effectively prevent and repair over-pruning. Moreover, inspired by the neurotrophic hypothesis, neuronal pruning rate and synaptic regeneration rate were adaptively adjusted during the learning-while-pruning process, which eventually led to the structural stability of SNNs. Experimental results on spatial (MNIST, CIFAR-10) and temporal neuromorphic (N-MNIST, DVS-Gesture) datasets demonstrate that our method can flexibly learn appropriate compression rate for various tasks and effectively achieve superior performance while massively reducing the network energy consumption. Specifically, for the spatial MNIST dataset, our SD-SNN achieves 99.51\% accuracy at the pruning rate 49.83\%, which has a 0.05\% accuracy improvement compared to the baseline without compression. For the neuromorphic DVS-Gesture dataset, 98.20\% accuracy with 1.09\% improvement is achieved by our method when the compression rate reaches 55.50\%.

Bing Han, Feifei Zhao, Yi Zeng et al.

Developmental plasticity plays a prominent role in shaping the brain's structure during ongoing learning in response to dynamically changing environments. However, the existing network compression methods for deep artificial neural networks (ANNs) and spiking neural networks (SNNs) draw little inspiration from brain's developmental plasticity mechanisms, thus limiting their ability to learn efficiently, rapidly, and accurately. This paper proposed a developmental plasticity-inspired adaptive pruning (DPAP) method, with inspiration from the adaptive developmental pruning of dendritic spines, synapses, and neurons according to the ``use it or lose it, gradually decay" principle. The proposed DPAP model considers multiple biologically realistic mechanisms (such as dendritic spine dynamic plasticity, activity-dependent neural spiking trace, and local synaptic plasticity), with additional adaptive pruning strategy, so that the network structure can be dynamically optimized during learning without any pre-training and retraining. Extensive comparative experiments show consistent and remarkable performance and speed boost with the extremely compressed networks on a diverse set of benchmark tasks for deep ANNs and SNNs, especially the spatio-temporal joint pruning of SNNs in neuromorphic datasets. This work explores how developmental plasticity enables complex deep networks to gradually evolve into brain-like efficient and compact structures, eventually achieving state-of-the-art (SOTA) performance for biologically realistic SNNs.

Bing Han, Feifei Zhao, Wenxuan Pan et al.

The human brain can self-organize rich and diverse sparse neural pathways to incrementally master hundreds of cognitive tasks. However, most existing continual learning algorithms for deep artificial and spiking neural networks are unable to adequately auto-regulate the limited resources in the network, which leads to performance drop along with energy consumption rise as the increase of tasks. In this paper, we propose a brain-inspired continual learning algorithm with adaptive reorganization of neural pathways, which employs Self-Organizing Regulation networks to reorganize the single and limited Spiking Neural Network (SOR-SNN) into rich sparse neural pathways to efficiently cope with incremental tasks. The proposed model demonstrates consistent superiority in performance, energy consumption, and memory capacity on diverse continual learning tasks ranging from child-like simple to complex tasks, as well as on generalized CIFAR100 and ImageNet datasets. In particular, the SOR-SNN model excels at learning more complex tasks as well as more tasks, and is able to integrate the past learned knowledge with the information from the current task, showing the backward transfer ability to facilitate the old tasks. Meanwhile, the proposed model exhibits self-repairing ability to irreversible damage and for pruned networks, could automatically allocate new pathway from the retained network to recover memory for forgotten knowledge.

Wenxuan Pan, Feifei Zhao, Yi Zeng et al.

The architecture design and multi-scale learning principles of the human brain that evolved over hundreds of millions of years are crucial to realizing human-like intelligence. Spiking Neural Network (SNN) based Liquid State Machine (LSM) serves as a suitable architecture to study brain-inspired intelligence because of its brain-inspired structure and the potential for integrating multiple biological principles. Existing researches on LSM focus on different certain perspectives, including high-dimensional encoding or optimization of the liquid layer, network architecture search, and application to hardware devices. There is still a lack of in-depth inspiration from the learning and structural evolution mechanism of the brain. Considering these limitations, this paper presents a novel LSM learning model that integrates adaptive structural evolution and multi-scale biological learning rules. For structural evolution, an adaptive evolvable LSM model is developed to optimize the neural architecture design of liquid layer with separation property. For brain-inspired learning of LSM, we propose a dopamine-modulated Bienenstock-Cooper-Munros (DA-BCM) method that incorporates global long-term dopamine regulation and local trace-based BCM synaptic plasticity. Comparative experimental results on different decision-making tasks show that introducing structural evolution of the liquid layer, and the DA-BCM regulation of the liquid layer and the readout layer could improve the decision-making ability of LSM and flexibly adapt to rule reversal. This work is committed to exploring how evolution can help to design more appropriate network architectures and how multi-scale neuroplasticity principles coordinated to enable the optimization and learning of LSMs for relatively complex decision-making tasks.

Yinqian Sun, Feifei Zhao, Zhuoya Zhao et al.

Inspired by the brain's information processing using binary spikes, spiking neural networks (SNNs) offer significant reductions in energy consumption and are more adept at incorporating multi-scale biological characteristics. In SNNs, spiking neurons serve as the fundamental information processing units. However, in most models, these neurons are typically simplified, focusing primarily on the leaky integrate-and-fire (LIF) point neuron model while neglecting the structural properties of biological neurons. This simplification hampers the computational and learning capabilities of SNNs. In this paper, we propose a brain-inspired deep distributional reinforcement learning algorithm based on SNNs, which integrates a bio-inspired multi-compartment neuron (MCN) model with a population coding approach. The proposed MCN model simulates the structure and function of apical dendritic, basal dendritic, and somatic compartments, achieving computational power comparable to that of biological neurons. Additionally, we introduce an implicit fractional embedding method based on population coding of spiking neurons. We evaluated our model on Atari games, and the experimental results demonstrate that it surpasses the vanilla FQF model, which utilizes traditional artificial neural networks (ANNs), as well as the Spiking-FQF models that are based on ANN-to-SNN conversion methods. Ablation studies further reveal that the proposed multi-compartment neuron model and the quantile fraction implicit population spike representation significantly enhance the performance of MCS-FQF while also reducing power consumption.

Yao Liang, Hongjian Fang, Yi Zeng et al.

Reasoning and question answering as a basic cognitive function for humans, is nevertheless a great challenge for current artificial intelligence. Although the Differentiable Neural Computer (DNC) model could solve such problems to a certain extent, the development is still limited by its high algorithm complexity, slow convergence speed, and poor test robustness. Inspired by the learning and memory mechanism of the brain, this paper proposed a Memory Transformation based Differentiable Neural Computer (MT-DNC) model. MT-DNC incorporates working memory and long-term memory into DNC, and realizes the autonomous transformation of acquired experience between working memory and long-term memory, thereby helping to effectively extract acquired knowledge to improve reasoning ability. Experimental results on bAbI question answering task demonstrated that our proposed method achieves superior performance and faster convergence speed compared to other existing DNN and DNC models. Ablation studies also indicated that the memory transformation from working memory to long-term memory plays essential role in improving the robustness and stability of reasoning. This work explores how brain-inspired memory transformation can be integrated and applied to complex intelligent dialogue and reasoning systems.

Wenxuan Pan, Feifei Zhao, Guobin Shen et al.

Spiking Neural Networks (SNNs) have received considerable attention not only for their superiority in energy efficiency with discrete signal processing but also for their natural suitability to integrate multi-scale biological plasticity. However, most SNNs directly adopt the structure of the well-established Deep Neural Networks (DNNs), and rarely automatically design Neural Architecture Search (NAS) for SNNs. The neural motifs topology, modular regional structure and global cross-brain region connection of the human brain are the product of natural evolution and can serve as a perfect reference for designing brain-inspired SNN architecture. In this paper, we propose a Multi-Scale Evolutionary Neural Architecture Search (MSE-NAS) for SNN, simultaneously considering micro-, meso- and macro-scale brain topologies as the evolutionary search space. MSE-NAS evolves individual neuron operation, self-organized integration of multiple circuit motifs, and global connectivity across motifs through a brain-inspired indirect evaluation function, Representational Dissimilarity Matrices (RDMs). This training-free fitness function could greatly reduce computational consumption and NAS's time, and its task-independent property enables the searched SNNs to exhibit excellent transferability on multiple datasets. Furthermore, MSE-NAS show robustness against the training method and noise. Extensive experiments demonstrate that the proposed algorithm achieves state-of-the-art (SOTA) performance with shorter simulation steps on static datasets (CIFAR10, CIFAR100) and neuromorphic datasets (CIFAR10-DVS and DVS128-Gesture). The thorough analysis also illustrates the significant performance improvement and consistent bio-interpretability deriving from the topological evolution at different scales and the RDMs fitness function.

Wenxuan Pan, Feifei Zhao, Zhuoya Zhao et al.

The complex and unique neural network topology of the human brain formed through natural evolution enables it to perform multiple cognitive functions simultaneously. Automated evolutionary mechanisms of biological network structure inspire us to explore efficient architectural optimization for Spiking Neural Networks (SNNs). Instead of manually designed fixed architectures or hierarchical Network Architecture Search (NAS), this paper evolves SNNs architecture by incorporating brain-inspired local modular structure and global cross-module connectivity. Locally, the brain region-inspired module consists of multiple neural motifs with excitatory and inhibitory connections; Globally, we evolve free connections among modules, including long-term cross-module feedforward and feedback connections. We further introduce an efficient multi-objective evolutionary algorithm based on a few-shot performance predictor, endowing SNNs with high performance, efficiency and low energy consumption. Extensive experiments on static datasets (CIFAR10, CIFAR100) and neuromorphic datasets (CIFAR10-DVS, DVS128-Gesture) demonstrate that our proposed model boosts energy efficiency, archiving consistent and remarkable performance. This work explores brain-inspired neural architectures suitable for SNNs and also provides preliminary insights into the evolutionary mechanisms of biological neural networks in the human brain.

Haibo Tong, Zeyang Yue, Feifei Zhao et al.

Whether Large Language Models (LLMs) truly possess human-like Theory of Mind (ToM) capabilities has garnered increasing attention. However, existing benchmarks remain largely restricted to narrow paradigms like false belief tasks, failing to capture the full spectrum of human cognitive mechanisms. We introduce CogToM, a comprehensive, theoretically grounded benchmark comprising over 8000 bilingual instances across 46 paradigms, validated by 49 human annotator.A systematic evaluation of 22 representative models, including frontier models like GPT-5.1 and Qwen3-Max, reveals significant performance heterogeneities and highlights persistent bottlenecks in specific dimensions. Further analysis based on human cognitive patterns suggests potential divergences between LLM and human cognitive structures. CogToM offers a robust instrument and perspective for investigating the evolving cognitive boundaries of LLMs.

Yin Luo, Qingchao Kong, Nan Xu et al.

As the latest advancements in natural language processing, large language models (LLMs) have achieved human-level language understanding and generation abilities in many real-world tasks, and even have been regarded as a potential path to the artificial general intelligence. To better facilitate research on LLMs, many open-source LLMs, such as Llama 2 and Falcon, have recently been proposed and gained comparable performances to proprietary models. However, these models are primarily designed for English scenarios and exhibit poor performances in Chinese contexts. In this technical report, we propose YAYI 2, including both base and chat models, with 30 billion parameters. YAYI 2 is pre-trained from scratch on a multilingual corpus which contains 2.65 trillion tokens filtered by our pre-training data processing pipeline. The base model is aligned with human values through supervised fine-tuning with millions of instructions and reinforcement learning from human feedback. Extensive experiments on multiple benchmarks, such as MMLU and CMMLU, consistently demonstrate that the proposed YAYI 2 outperforms other similar sized open-source models.

Ping Wu, Guobin Shen, Dongcheng Zhao et al.

Ensuring that Large Language Models (LLMs) align with mainstream human values and ethical norms is crucial for the safe and sustainable development of AI. Current value evaluation and alignment are constrained by Western cultural bias and incomplete domestic frameworks reliant on non-native rules; furthermore, the lack of scalable, rule-driven scenario generation methods makes evaluations costly and inadequate across diverse cultural contexts. To address these challenges, we propose a hierarchical value framework grounded in core Chinese values, encompassing three main dimensions, 12 core values, and 50 derived values. Based on this framework, we construct a large-scale Chinese Values Corpus (CVC) containing over 250,000 value rules enhanced and expanded through human annotation. Experimental results show that CVC-guided scenarios outperform direct generation ones in value boundaries and content diversity. In the evaluation across six sensitive themes (e.g., surrogacy, suicide), seven mainstream LLMs preferred CVC-generated options in over 70.5% of cases, while five Chinese human annotators showed an 87.5% alignment with CVC, confirming its universality, cultural relevance, and strong alignment with Chinese values. Additionally, we construct 400,000 rule-based moral dilemma scenarios that objectively capture nuanced distinctions in conflicting value prioritization across 17 LLMs. Our work establishes a culturally-adaptive benchmarking framework for comprehensive value evaluation and alignment, representing Chinese characteristics. All data are available at https://huggingface.co/datasets/Beijing-AISI/CVC, and the code is available at https://github.com/Beijing-AISI/CVC.

Haibo Tong, Feifei Zhao, Linghao Feng et al.

Rapidly evolving AI exhibits increasingly strong autonomy and goal-directed capabilities, accompanied by derivative systemic risks that are more unpredictable, difficult to control, and potentially irreversible. However, current AI safety evaluation systems suffer from critical limitations such as restricted risk dimensions and failed frontier risk detection. The lagging safety benchmarks and alignment technologies can hardly address the complex challenges posed by cutting-edge AI models. To bridge this gap, we propose the "ForesightSafety Bench" AI Safety Evaluation Framework, beginning with 7 major Fundamental Safety pillars and progressively extends to advanced Embodied AI Safety, AI4Science Safety, Social and Environmental AI risks, Catastrophic and Existential Risks, as well as 8 critical industrial safety domains, forming a total of 94 refined risk dimensions. To date, the benchmark has accumulated tens of thousands of structured risk data points and assessment results, establishing a widely encompassing, hierarchically clear, and dynamically evolving AI safety evaluation framework. Based on this benchmark, we conduct systematic evaluation and in-depth analysis of over twenty mainstream advanced large models, identifying key risk patterns and their capability boundaries. The safety capability evaluation results reveals the widespread safety vulnerabilities of frontier AI across multiple pillars, particularly focusing on Risky Agentic Autonomy, AI4Science Safety, Embodied AI Safety, Social AI Safety and Catastrophic and Existential Risks. Our benchmark is released at https://github.com/Beijing-AISI/ForesightSafety-Bench. The project website is available at https://foresightsafety-bench.beijing-aisi.ac.cn/.

Jiyan Song, Wenyang Wang, Chengcheng Yan et al.

In the treatment of complex diseases, treatment regimens using a single drug often yield limited efficacy and can lead to drug resistance. In contrast, combination drug therapies can significantly improve therapeutic outcomes through synergistic effects. However, experimentally validating all possible drug combinations is prohibitively expensive, underscoring the critical need for efficient computational prediction methods. Although existing approaches based on deep learning and graph neural networks (GNNs) have made considerable progress, challenges remain in reducing structural bias, improving generalization capability, and enhancing model interpretability. To address these limitations, this paper proposes a collaborative prediction graph neural network that integrates molecular structural features and cell-line genomic profiles with drug-drug interactions to enhance the prediction of synergistic effects. We introduce a novel model named the Residual Graph Isomorphism Network integrated with an Attention mechanism (ResGIN-Att). The model first extracts multi scale topological features of drug molecules using a residual graph isomorphism network, where residual connections help mitigate over-smoothing in deep layers. Subsequently, an adaptive Long Short-Term Memory (LSTM) module fuses structural information from local to global scales. Finally, a cross-attention module is designed to explicitly model drug-drug interactions and identify key chemical substructures. Extensive experiments on five public benchmark datasets demonstrate that ResGIN-Att achieves competitive performance, comparing favorably against key baseline methods while exhibiting promising generalization capability and robustness.

Yi Zeng, Feifei Zhao, Yuwei Wang et al.

As Artificial Intelligence (AI) advances toward Artificial General Intelligence (AGI) and eventually Artificial Superintelligence (ASI), it may potentially surpass human control, deviate from human values, and even lead to irreversible catastrophic consequences in extreme cases. This looming risk underscores the critical importance of the "superalignment" problem - ensuring that AI systems which are much smarter than humans, remain aligned with human (compatible) intentions and values. While current scalable oversight and weak-to-strong generalization methods demonstrate certain applicability, they exhibit fundamental flaws in addressing the superalignment paradigm - notably, the unidirectional imposition of human values cannot accommodate superintelligence's autonomy or ensure AGI/ASI's stable learning. We contend that the values for sustainable symbiotic society should be co-shaped by humans and living AI together, achieving "Super Co-alignment." Guided by this vision, we propose a concrete framework that integrates external oversight and intrinsic proactive alignment. External oversight superalignment should be grounded in human-centered ultimate decision, supplemented by interpretable automated evaluation and correction, to achieve continuous alignment with humanity's evolving values. Intrinsic proactive superalignment is rooted in a profound understanding of the Self, others, and society, integrating self-awareness, self-reflection, and empathy to spontaneously infer human intentions, distinguishing good from evil and proactively prioritizing human well-being. The integration of externally-driven oversight with intrinsically-driven proactive alignment will co-shape symbiotic values and rules through iterative human-ASI co-alignment, paving the way for achieving safe and beneficial AGI and ASI for good, for human, and for a symbiotic ecology.

Yi Zeng, Feifei Zhao, Yuxuan Zhao et al.

The question "Can machines think?" and the Turing Test to assess whether machines could achieve human-level intelligence is one of the roots of AI. With the philosophical argument "I think, therefore I am", this paper challenge the idea of a "thinking machine" supported by current AIs since there is no sense of self in them. Current artificial intelligence is only seemingly intelligent information processing and does not truly understand or be subjectively aware of oneself and perceive the world with the self as human intelligence does. In this paper, we introduce a Brain-inspired and Self-based Artificial Intelligence (BriSe AI) paradigm. This BriSe AI paradigm is dedicated to coordinating various cognitive functions and learning strategies in a self-organized manner to build human-level AI models and robotic applications. Specifically, BriSe AI emphasizes the crucial role of the Self in shaping the future AI, rooted with a practical hierarchical Self framework, including Perception and Learning, Bodily Self, Autonomous Self, Social Self, and Conceptual Self. The hierarchical framework of the Self highlights self-based environment perception, self-bodily modeling, autonomous interaction with the environment, social interaction and collaboration with others, and even more abstract understanding of the Self. Furthermore, the positive mutual promotion and support among multiple levels of Self, as well as between Self and learning, enhance the BriSe AI's conscious understanding of information and flexible adaptation to complex environments, serving as a driving force propelling BriSe AI towards real Artificial General Intelligence.

Bing Han, Feifei Zhao, Yinqian Sun et al.

Cognitive functions in current artificial intelligence networks are tied to the exponential increase in network scale, whereas the human brain can continuously learn hundreds of cognitive functions with remarkably low energy consumption. This advantage is in part due to the brain cross-regional temporal development mechanisms, where the progressive formation, reorganization, and pruning of connections from basic to advanced regions, facilitate knowledge transfer and prevent network redundancy. Inspired by these, we propose the Continual Learning of Multiple Cognitive Functions with Brain-inspired Temporal Development Mechanism(TD-MCL), enabling cognitive enhancement from simple to complex in Perception-Motor-Interaction(PMI) multiple cognitive task scenarios. The TD-MCL model proposes the sequential evolution of long-range connections between different cognitive modules to promote positive knowledge transfer, while using feedback-guided local connection inhibition and pruning to effectively eliminate redundancies in previous tasks, reducing energy consumption while preserving acquired knowledge. Experiments show that the proposed method can achieve continual learning capabilities while reducing network scale, without introducing regularization, replay, or freezing strategies, and achieving superior accuracy on new tasks compared to direct learning. The proposed method shows that the brain's developmental mechanisms offer a valuable reference for exploring biologically plausible, low-energy enhancements of general cognitive abilities.

Haibo Tong, Enmeng Lu, Yinqian Sun et al.

With the widespread application of Artificial Intelligence (AI) in human society, enabling AI to autonomously align with human values has become a pressing issue to ensure its sustainable development and benefit to humanity. One of the most important aspects of aligning with human values is the necessity for agents to autonomously make altruistic, safe, and ethical decisions, considering and caring for human well-being. Current AI extremely pursues absolute superiority in certain tasks, remaining indifferent to the surrounding environment and other agents, which has led to numerous safety risks. Altruistic behavior in human society originates from humans' capacity for empathizing others, known as Theory of Mind (ToM), combined with predictive imaginative interactions before taking action to produce thoughtful and altruistic behaviors. Inspired by this, we are committed to endow agents with considerate self-imagination and ToM capabilities, driving them through implicit intrinsic motivations to autonomously align with human altruistic values. By integrating ToM within the imaginative space, agents keep an eye on the well-being of other agents in real time, proactively anticipate potential risks to themselves and others, and make thoughtful altruistic decisions that balance negative effects on the environment. The ancient Chinese story of Sima Guang Smashes the Vat illustrates the moral behavior of the young Sima Guang smashed a vat to save a child who had accidentally fallen into it, which is an excellent reference scenario for this paper. We design an experimental scenario similar to Sima Guang Smashes the Vat and its variants with different complexities, which reflects the trade-offs and comprehensive considerations between self-goals, altruistic rescue, and avoiding negative side effects.

Bing Han, Feifei Zhao, Yang Li et al.

Biological brains have the capability to adaptively coordinate relevant neuronal populations based on the task context to learn continuously changing tasks in real-world environments. However, existing spiking neural network-based continual learning algorithms treat each task equally, ignoring the guiding role of different task similarity associations for network learning, which limits knowledge utilization efficiency. Inspired by the context-dependent plasticity mechanism of the brain, we propose a Similarity-based Context Aware Spiking Neural Network (SCA-SNN) continual learning algorithm to efficiently accomplish task incremental learning and class incremental learning. Based on contextual similarity across tasks, the SCA-SNN model can adaptively reuse neurons from previous tasks that are beneficial for new tasks (the more similar, the more neurons are reused) and flexibly expand new neurons for the new task (the more similar, the fewer neurons are expanded). Selective reuse and discriminative expansion significantly improve the utilization of previous knowledge and reduce energy consumption. Extensive experimental results on CIFAR100, ImageNet generalized datasets, and FMNIST-MNIST, SVHN-CIFAR100 mixed datasets show that our SCA-SNN model achieves superior performance compared to both SNN-based and DNN-based continual learning algorithms. Additionally, our algorithm has the capability to adaptively select similar groups of neurons for related tasks, offering a promising approach to enhancing the biological interpretability of efficient continual learning.

Mingyang Lyu, Yinqian Sun, Erliang Lin et al.

Vision-Language-Action (VLA) models such as OpenVLA, Octo, and $π_0$ have shown strong generalization by leveraging large-scale demonstrations, yet their performance is still fundamentally constrained by the quality and coverage of supervised data. Reinforcement learning (RL) provides a promising path for improving and fine-tuning VLAs through online interaction. However, conventional policy gradient methods are computationally infeasible in the context of flow-matching based models due to the intractability of the importance sampling process, which requires explicit computation of policy ratios. To overcome this limitation, we propose Flow Policy Optimization (FPO) algorithm, which reformulates importance sampling by leveraging per-sample changes in the conditional flow-matching objective. Furthermore, FPO achieves stable and scalable online reinforcement fine-tuning of the $π_0$ model by integrating structure-aware credit assignment to enhance gradient efficiency, clipped surrogate objectives to stabilize optimization, multi-step latent exploration to encourage diverse policy updates, and a Q-ensemble mechanism to provide robust value estimation. We evaluate FPO on the LIBERO benchmark and the ALOHA simulation task against supervised, preference-aligned, diffusion-based, autoregressive online RL, and $π_0$-FAST baselines, observing consistent improvements over the imitation prior and strong alternatives with stable learning under sparse rewards. In addition, ablation studies and analyses of the latent space dynamics further highlight the contributions of individual components within FPO, validating the effectiveness of the proposed computational modules and the stable convergence of the conditional flow-matching objective during online RL.

Guobin Shen, Dongcheng Zhao, Haibo Tong et al.

Ensuring Large Language Model (LLM) safety remains challenging due to the absence of universal standards and reliable content validators, making it difficult to obtain effective training signals. We discover that aligned models already possess robust internal safety beliefs: they consistently produce high-confidence refusals to harmful requests while exhibiting high entropy when generating potentially dangerous content. This entropy gap reveals an untapped signal--models intrinsically "know" when to refuse. We introduce Safety Instincts Reinforcement Learning (SIRL), which transforms this internal confidence into a self-generated reward signal, eliminating dependence on external validators or human annotations. SIRL teaches models to trust their safety instincts by reinforcing low-entropy refusal behaviors. Evaluated on Llama and Qwen models, SIRL maintains 89%+ Defense Success Rates (DSRs) against 20+ jailbreak methods, from static prompts to adaptive attacks. Using only 15,000 unlabeled prompts, SIRL surpasses resource-intensive supervised methods while preserving performance on mathematics, coding, and conversation benchmarks. Our work demonstrates that effective alignment can emerge from within, paving the way for more autonomous and robust AI safety mechanisms that scale without extensive human oversight.

Haibo Tong, Dongcheng Zhao, Guobin Shen et al.

The remarkable capabilities of Large Language Models (LLMs) have raised significant safety concerns, particularly regarding "jailbreak" attacks that exploit adversarial prompts to bypass safety alignment mechanisms. Existing defense research primarily focuses on single-turn attacks, whereas multi-turn jailbreak attacks progressively break through safeguards through by concealing malicious intent and tactical manipulation, ultimately rendering conventional single-turn defenses ineffective. To address this critical challenge, we propose the Bidirectional Intention Inference Defense (BIID). The method integrates forward request-based intention inference with backward response-based intention retrospection, establishing a bidirectional synergy mechanism to detect risks concealed within seemingly benign inputs, thereby constructing a more robust guardrails that effectively prevents harmful content generation. The proposed method undergoes systematic evaluation compared with a no-defense baseline and seven representative defense methods across three LLMs and two safety benchmarks under 10 different attack methods. Experimental results demonstrate that the proposed method significantly reduces the Attack Success Rate (ASR) across both single-turn and multi-turn jailbreak attempts, outperforming all existing baseline methods while effectively maintaining practical utility. Notably, comparative experiments across three multi-turn safety datasets further validate the proposed model's significant advantages over other defense approaches.

Yao Liang, Dongcheng Zhao, Feifei Zhao et al.

The alignment of large language models (LLMs) with human values is critical for their safe and effective deployment across diverse user populations. However, existing benchmarks often neglect cultural and demographic diversity, leading to limited understanding of how value alignment generalizes globally. In this work, we introduce MVPBench, a novel benchmark that systematically evaluates LLMs' alignment with multi-dimensional human value preferences across 75 countries. MVPBench contains 24,020 high-quality instances annotated with fine-grained value labels, personalized questions, and rich demographic metadata, making it the most comprehensive resource of its kind to date. Using MVPBench, we conduct an in-depth analysis of several state-of-the-art LLMs, revealing substantial disparities in alignment performance across geographic and demographic lines. We further demonstrate that lightweight fine-tuning methods, such as Low-Rank Adaptation (LoRA) and Direct Preference Optimization (DPO), can significantly enhance value alignment in both in-domain and out-of-domain settings. Our findings underscore the necessity for population-aware alignment evaluation and provide actionable insights for building culturally adaptive and value-sensitive LLMs. MVPBench serves as a practical foundation for future research on global alignment, personalized value modeling, and equitable AI development.

Bing Han, Feifei Zhao, Dongcheng Zhao et al.

Fine-tuning as service injects domain-specific knowledge into large language models (LLMs), while challenging the original alignment mechanisms and introducing safety risks. A series of defense strategies have been proposed for the alignment, fine-tuning, and post-fine-tuning phases, where most post-fine-tuning defenses rely on coarse-grained safety layer mapping. These methods lack a comprehensive consideration of both safety layers and fine-grained neurons, limiting their ability to efficiently balance safety and utility. To address this, we propose the Fine-Grained Safety Neurons (FGSN) with Training-Free Continual Projection method to reduce the fine-tuning safety risks. FGSN inherently integrates the multi-scale interactions between safety layers and neurons, localizing sparser and more precise fine-grained safety neurons while minimizing interference with downstream task neurons. We then project the safety neuron parameters onto safety directions, improving model safety while aligning more closely with human preferences. Extensive experiments across multiple fine-tuned LLM models demonstrate that our method significantly reduce harmfulness scores and attack success rates with minimal parameter modifications, while preserving the model's utility. Furthermore, by introducing a task-specific, multi-dimensional heterogeneous safety neuron cluster optimization mechanism, we achieve continual defense and generalization capability against unforeseen emerging safety concerns.

Wenxuan Pan, Feifei Zhao, Bing Han et al.

By exploiting discrete signal processing and simulating brain neuron communication, Spiking Neural Networks (SNNs) offer a low-energy alternative to Artificial Neural Networks (ANNs). However, existing SNN models, still face high computational costs due to the numerous time steps as well as network depth and scale. The tens of billions of neurons and trillions of synapses in the human brain are developed from only 20,000 genes, which inspires us to design an efficient genetic encoding strategy that dynamic evolves to regulate large-scale deep SNNs at low cost. Therefore, we first propose a genetically scaled SNN encoding scheme that incorporates globally shared genetic interactions to indirectly optimize neuronal encoding instead of weight, which obviously brings about reductions in parameters and energy consumption. Then, a spatio-temporal evolutionary framework is designed to optimize the inherently initial wiring rules. Two dynamic regularization operators in the fitness function evolve the neuronal encoding to a suitable distribution and enhance information quality of the genetic interaction respectively, substantially accelerating evolutionary speed and improving efficiency. Experiments show that our approach compresses parameters by approximately 50\% to 80\%, while outperforming models on the same architectures by 0.21\% to 4.38\% on CIFAR-10, CIFAR-100 and ImageNet. In summary, the consistent trends of the proposed genetically encoded spatio-temporal evolution across different datasets and architectures highlight its significant enhancements in terms of efficiency, broad scalability and robustness, demonstrating the advantages of the brain-inspired evolutionary genetic coding for SNN optimization.

Feifei Zhao, Hui Feng, Haibo Tong et al.

As AI closely interacts with human society, it is crucial to ensure that its behavior is safe, altruistic, and aligned with human ethical and moral values. However, existing research on embedding ethical considerations into AI remains insufficient, and previous external constraints based on principles and rules are inadequate to provide AI with long-term stability and generalization capabilities. Emotional empathy intrinsically motivates altruistic behaviors aimed at alleviating others' negative emotions through emotional sharing and contagion mechanisms. Motivated by this, we draw inspiration from the neural mechanism of human emotional empathy-driven altruistic decision making, and simulate the shared self-other perception-mirroring-empathy neural circuits, to construct a brain-inspired emotional empathy-driven altruistic decision-making model. Here, empathy directly impacts dopamine release to form intrinsic altruistic motivation. The proposed model exhibits consistent altruistic behaviors across three experimental settings: emotional contagion-integrated two-agent altruistic rescue, multi-agent gaming, and robotic emotional empathy interaction scenarios. In-depth analyses validate the positive correlation between empathy levels and altruistic preferences (consistent with psychological behavioral experiment findings), while also demonstrating how interaction partners' empathy levels influence the agent's behavioral patterns. We further test the proposed model's performance and stability in moral dilemmas involving conflicts between self-interest and others' well-being, partially observable environments, and adversarial defense scenarios. This work provides preliminary exploration of human-like empathy-driven altruistic moral decision making, contributing potential perspectives for developing ethically-aligned AI.