arXiv subjects

Jindong Li

Publications and source records attributed to Jindong Li.

At least 37 records · Page 2Linked to original sources

arXiv2026

HeLa-Mem: Hebbian Learning and Associative Memory for LLM Agents

Long-term memory is a critical challenge for Large Language Model agents, as fixed context windows cannot preserve coherence across extended interactions. Existing memory systems represent conversation history as unstructured embedding vectors, retrieving information through semantic similarity. This paradigm fails to capture the associative structure of human memory, wherein related experiences progressively strengthen interconnections through repeated co-activation. Inspired by cognitive neuroscience, we identify three mechanisms central to biological memory: association, consolidation, and spreading activation, which remain largely absent in current research. To bridge this gap, we propose HeLa-Mem, a bio-inspired memory architecture that models memory as a dynamic graph with Hebbian learning dynamics. HeLa-Mem employs a dual-level organization: (1) an episodic memory graph that evolves through co-activation patterns, and (2) a semantic memory store populated via Hebbian Distillation, wherein a Reflective Agent identifies densely connected memory hubs and distills them into structured, reusable semantic knowledge. This dual-path design leverages both semantic similarity and learned associations, mirroring the episodic-semantic distinction in human cognition. Experiments on LoCoMo demonstrate superior performance across four question categories while using significantly fewer context tokens. Code is available on GitHub: https://github.com/ReinerBRO/HeLa-Mem

cs.CL↗

arXiv2026

ForesightSafety Bench: A Frontier Risk Evaluation and Governance Framework towards Safe AI

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/.

cs.AI↗

arXiv2026

FireFly-S: Exploiting Dual-Side Sparsity for Spiking Neural Networks Acceleration with Reconfigurable Spatial Architecture

Spiking Neural Networks (SNNs), with brain-inspired structure using discrete spikes instead of continuous activations, are gaining attention for their efficient processing on neuromorphic chips. While current SNN hardware accelerators often prioritize temporal spike sparsity, exploiting sparse synaptic weights offers significant untapped potential for even greater efficiency. To address this, we propose FireFly-S, a Sparse extension of the FireFly series. This co-optimized software-hardware design focuses on leveraging dual-side sparsity for acceleration. On the software side, we propose a algorithmic optimization framework that combines gradient rewiring for pruning and modified Learned Step Size Quantization (LSQ) for SNNs, achieving a weight sparsity exceeding 85\% and enabling efficient 4-bit quantization with negligible accuracy loss. On the hardware side, we present an efficient dual-side sparsity detector employing a Bitmap-based sparse decoding logic to pinpoint the positions of non-zero weights and input spikes. The logic allows for direct bypassing of redundant computations, thereby enhancing computational efficiency. Different from the overlay architecture adopted by previous FireFly series, we adopt a parametric spatial architecture with inter-layer pipelining that can fully exploit the fine-grained programmability and reconfigurability of Field-Programmable Gate Arrays (FPGAs), enabling fast deployment for various models. A spatial-temporal dataflow is also proposed to support such inter-layer pipelining and avoid long-term temporal dependencies. In experiments conducted on the MNIST, DVS-Gesture and CIFAR-10 datasets, the FireFly-S model achieves 85--95\% sparsity with 4-bit quantization and the hardware accelerator effectively leverages the dual-side sparsity, delivering performance metrics of 10,047~FPS/W on MNIST, 3,683~FPS/W on DVS-Gesture, and 2,327~FPS/W on CIFAR-10.

cs.AR↗

arXiv2026

Beyond Retraining: Training-Free Unknown Class Filtering for Source-Free Open Set Domain Adaptation of Vision-Language Models

Vision-language models (VLMs) have gained widespread attention for their strong zero-shot capabilities across numerous downstream tasks. However, these models assume that each test image's class label is drawn from a predefined label set and lack a reliable mechanism to reject samples from emerging unknown classes when only unlabeled data are available. To address this gap, open-set domain adaptation methods retrain models to push potential unknowns away from known clusters. Yet, some unknown samples remain stably anchored to specific known classes in the VLM feature space due to semantic relevance, which is termed as Semantic Affinity Anchoring (SAA). Forcibly repelling these samples unavoidably distorts the native geometry of VLMs and degrades performance. Meanwhile, existing score-based unknown detectors use simplistic thresholds and suffer from threshold sensitivity, resulting in sub-optimal performance. To address aforementioned issues, we propose VLM-OpenXpert, which comprises two training-free, plug-and-play inference modules. SUFF performs SVD on high-confidence unknowns to extract a low-rank "unknown subspace". Each sample's projection onto this subspace is weighted and softly removed from its feature, suppressing unknown components while preserving semantics. BGAT corrects score skewness via a Box-Cox transform, then fits a bimodal Gaussian mixture to adaptively estimate the optimal threshold balancing known-class recognition and unknown-class rejection. Experiments on 9 benchmarks and three backbones (CLIP, SigLIP, ALIGN) under source-free OSDA settings show that our training-free pipeline matches or outperforms retraining-heavy state-of-the-art methods, establishing a powerful lightweight inference calibration paradigm for open-set VLM deployment.

cs.CV↗

arXiv2026

FireFly-P: FPGA-Accelerated Spiking Neural Network Plasticity for Robust Adaptive Control

Spiking Neural Networks (SNNs) offer a biologically plausible learning mechanism through synaptic plasticity, enabling unsupervised adaptation without the computational overhead of backpropagation. To harness this capability for robotics, this paper presents FireFly-P, an FPGA-based hardware accelerator that implements a novel plasticity algorithm for real-time adaptive control. By leveraging on-chip plasticity, our architecture enhances the network's generalization, ensuring robust performance in dynamic and unstructured environments. The hardware design achieves an end-to-end latency of just 8~$μ$s for both inference and plasticity updates, enabling rapid adaptation to unseen scenarios. Implemented on a tiny Cmod A7-35T FPGA, FireFly-P consumes only 0.713~W and $\sim$10K~LUTs, making it ideal for power- and resource-constrained embedded robotic platforms. This work demonstrates that hardware-accelerated SNN plasticity is a viable path toward enabling adaptive, low-latency, and energy-efficient control systems.

cs.AR↗

arXiv2025

GLADMamba: Unsupervised Graph-Level Anomaly Detection Powered by Selective State Space Model

Unsupervised graph-level anomaly detection (UGLAD) is a critical and challenging task across various domains, such as social network analysis, anti-cancer drug discovery, and toxic molecule identification. However, existing methods often struggle to capture long-range dependencies efficiently and neglect the spectral information. Recently, selective state space models, particularly Mamba, have demonstrated remarkable advantages in capturing long-range dependencies with linear complexity and a selection mechanism. Motivated by their success across various domains, we propose GLADMamba, a novel framework that adapts the selective state space model into UGLAD field. We design a View-Fused Mamba (VFM) module with a Mamba-Transformer-style architecture to efficiently fuse information from different graph views with a selective state mechanism. We also design a Spectrum-Guided Mamba (SGM) module with a Mamba-Transformer-style architecture to leverage the Rayleigh quotient to guide the embedding refinement process, considering the spectral information for UGLAD. GLADMamba can dynamically focus on anomaly-related information while discarding irrelevant information for anomaly detection. To the best of our knowledge, this is the first work to introduce Mamba and explicit spectral information to UGLAD. Extensive experiments on 12 real-world datasets demonstrate that GLADMamba outperforms existing state-of-the-art methods, achieving superior performance in UGLAD. The code is available at https://github.com/Yali-Fu/GLADMamba.

cs.LG↗

arXiv2025

STEP: A Unified Spiking Transformer Evaluation Platform for Fair and Reproducible Benchmarking

Spiking Transformers have recently emerged as promising architectures for combining the efficiency of spiking neural networks with the representational power of self-attention. However, the lack of standardized implementations, evaluation pipelines, and consistent design choices has hindered fair comparison and principled analysis. In this paper, we introduce STEP a unified benchmark framework for Spiking Transformers that supports a wide range of tasks, including classification, segmentation, and detection across static, event-based, and sequential datasets. STEP provides modular support for diverse components such as spiking neurons, input encodings, surrogate gradients, and multiple backends (e.g., SpikingJelly, BrainCog). Using STEP, we reproduce and evaluate several representative models, and conduct systematic ablation studies on attention design, neuron types, encoding schemes, and temporal modeling capabilities. We also propose a unified analytical model for energy estimation, accounting for spike sparsity, bitwidth, and memory access, and show that quantized ANNs may offer comparable or better energy efficiency. Our results suggest that current Spiking Transformers rely heavily on convolutional frontends and lack strong temporal modeling, underscoring the need for spike-native architectural innovations. The full code is available at: https://github.com/Fancyssc/STEP

cs.NE↗

arXiv2025

HyperbolicRAG: Enhancing Retrieval-Augmented Generation with Hyperbolic Representations

Retrieval-augmented generation (RAG) enables large language models (LLMs) to access external knowledge, helping mitigate hallucinations and enhance domain-specific expertise. Graph-based RAG enhances structural reasoning by introducing explicit relational organization that enables information propagation across semantically connected text units. However, these methods typically rely on Euclidean embeddings that capture semantic similarity but lack a geometric notion of hierarchical depth, limiting their ability to represent abstraction relationships inherent in complex knowledge graphs. To capture both fine-grained semantics and global hierarchy, we propose HyperbolicRAG, a retrieval framework that integrates hyperbolic geometry into graph-based RAG. HyperbolicRAG introduces three key designs: (1) a depth-aware representation learner that embeds nodes within a shared Poincare manifold to align semantic similarity with hierarchical containment, (2) an unsupervised contrastive regularization that enforces geometric consistency across abstraction levels, and (3) a mutual-ranking fusion mechanism that jointly exploits retrieval signals from Euclidean and hyperbolic spaces, emphasizing cross-space agreement during inference. Extensive experiments across multiple QA benchmarks demonstrate that HyperbolicRAG outperforms competitive baselines, including both standard RAG and graph-augmented baselines.

cs.CL↗

arXiv2025

Hummingbird: A Smaller and Faster Large Language Model Accelerator on Embedded FPGA

Deploying large language models (LLMs) on embedded devices remains a significant research challenge due to the high computational and memory demands of LLMs and the limited hardware resources available in such environments. While embedded FPGAs have demonstrated performance and energy efficiency in traditional deep neural networks, their potential for LLM inference remains largely unexplored. Recent efforts to deploy LLMs on FPGAs have primarily relied on large, expensive cloud-grade hardware and have only shown promising results on relatively small LLMs, limiting their real-world applicability. In this work, we present Hummingbird, a novel FPGA accelerator designed specifically for LLM inference on embedded FPGAs. Hummingbird is smaller, targeting embedded FPGAs such as the KV260 and ZCU104 with 67% LUT, 39% DSP, and 42% power savings over existing research. Hummingbird is stronger, targeting LLaMA3-8B and supporting longer contexts, overcoming the typical 4GB memory constraint of embedded FPGAs through offloading strategies. Finally, Hummingbird is faste, achieving 4.8 tokens/s and 8.6 tokens/s for LLaMA3-8B on the KV260 and ZCU104 respectively, with 93-94% model bandwidth utilization, outperforming the prior 4.9 token/s for LLaMA2-7B with 84% bandwidth utilization baseline. We further demonstrate the viability of industrial applications by deploying Hummingbird on a cost-optimized Spartan UltraScale FPGA, paving the way for affordable LLM solutions at the edge.

cs.AR↗

arXiv2025

Safety Instincts: LLMs Learn to Trust Their Internal Compass for Self-Defense

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.

cs.AI↗

arXiv2025

Stratify or Die: Rethinking Data Splits in Image Segmentation

Random splitting of datasets in image segmentation often leads to unrepresentative test sets, resulting in biased evaluations and poor model generalization. While stratified sampling has proven effective for addressing label distribution imbalance in classification tasks, extending these ideas to segmentation remains challenging due to the multi-label structure and class imbalance typically present in such data. Building on existing stratification concepts, we introduce Iterative Pixel Stratification (IPS), a straightforward, label-aware sampling method tailored for segmentation tasks. Additionally, we present Wasserstein-Driven Evolutionary Stratification (WDES), a novel genetic algorithm designed to minimize the Wasserstein distance, thereby optimizing the similarity of label distributions across dataset splits. We prove that WDES is globally optimal given enough generations. Using newly proposed statistical heterogeneity metrics, we evaluate both methods against random sampling and find that WDES consistently produces more representative splits. Applying WDES across diverse segmentation tasks, including street scenes, medical imaging, and satellite imagery, leads to lower performance variance and improved model evaluation. Our results also highlight the particular value of WDES in handling small, imbalanced, and low-diversity datasets, where conventional splitting strategies are most prone to bias.

cs.CV↗

arXiv2025

Implicit Reasoning in Large Language Models: A Comprehensive Survey

Large Language Models (LLMs) have demonstrated strong generalization across a wide range of tasks. Reasoning with LLMs is central to solving multi-step problems and complex decision-making. To support efficient reasoning, recent studies have shifted attention from explicit chain-of-thought prompting toward implicit reasoning, where reasoning occurs silently via latent structures without emitting intermediate textual steps. Implicit reasoning brings advantages such as lower generation cost, faster inference, and better alignment with internal computation. Although prior surveys have discussed latent representations in the context of reasoning, a dedicated and mechanism-level examination of how reasoning unfolds internally within LLMs remains absent. This survey fills that gap by introducing a taxonomy centered on execution paradigms, shifting the focus from representational forms to computational strategies. We organize existing methods into three execution paradigms based on \textbf{\textit{how and where internal computation unfolds}}: latent optimization, signal-guided control, and layer-recurrent execution. We also review structural, behavioral and representation-based evidence that supports the presence of implicit reasoning in LLMs. We further provide a structured overview of the evaluation metrics and benchmarks used in existing works to assess the effectiveness and reliability of implicit reasoning. We maintain a continuously updated project at: https://github.com/digailab/awesome-llm-implicit-reasoning.

cs.CL↗

arXiv2025

PandaGuard: Systematic Evaluation of LLM Safety against Jailbreaking Attacks

Large language models (LLMs) have achieved remarkable capabilities but remain vulnerable to adversarial prompts known as jailbreaks, which can bypass safety alignment and elicit harmful outputs. Despite growing efforts in LLM safety research, existing evaluations are often fragmented, focused on isolated attack or defense techniques, and lack systematic, reproducible analysis. In this work, we introduce PandaGuard, a unified and modular framework that models LLM jailbreak safety as a multi-agent system comprising attackers, defenders, and judges. Our framework implements 19 attack methods and 12 defense mechanisms, along with multiple judgment strategies, all within a flexible plugin architecture supporting diverse LLM interfaces, multiple interaction modes, and configuration-driven experimentation that enhances reproducibility and practical deployment. Built on this framework, we develop PandaBench, a comprehensive benchmark that evaluates the interactions between these attack/defense methods across 49 LLMs and various judgment approaches, requiring over 3 billion tokens to execute. Our extensive evaluation reveals key insights into model vulnerabilities, defense cost-performance trade-offs, and judge consistency. We find that no single defense is optimal across all dimensions and that judge disagreement introduces nontrivial variance in safety assessments. We release the code, configurations, and evaluation results to support transparent and reproducible research in LLM safety.

cs.CR↗

arXiv2025

FireFly-T: High-Throughput Sparsity Exploitation for Spiking Transformer Acceleration with Dual-Engine Overlay Architecture

Spiking transformers are emerging as a promising architecture that combines the energy efficiency of Spiking Neural Networks (SNNs) with the powerful attention mechanisms of transformers. However, existing hardware accelerators lack support for spiking attention, exhibit limited throughput in exploiting fine-grained sparsity, and struggle with scalable parallelism in sparse computation. To address these, we propose FireFly-T, a dual-engine overlay architecture that integrates a sparse engine for activation sparsity and a binary engine for spiking attention. In the sparse engine, we propose a highthroughput sparse decoder that exploits fine-grained sparsity by concurrently extracting multiple non-zero spikes. To complement this, we introduce a scalable load balancing mechanism with weight dispatch and out-of-order execution, eliminating bank conflicts to support scalable multidimensional parallelism. In the binary engine, we leverage the byte-level write capability of SRAMs to efficiently manipulate the 3D dataflows required for spiking attention with minimal resource overhead. We also optimize the core AND-PopCount operation in spiking attention through a LUT6-based implementation, improving timing closure and reducing LUT utilization on Xilinx FPGAs. As an overlay architecture, FireFly-T further incorporates an orchestrator that dynamically manipulates input dataflows with flexible adaptation for diverse network topologies, while ensuring efficient resource utilization and maintaining high throughput. Experimental results demonstrate that our accelerator achieves $1.39\times$ and $2.40\times$ higher energy efficiency, as well as $4.21\times$ and $7.10\times$ greater DSP efficiency, compared to FireFly v2 and the transformer-enabled SpikeTA, respectively. These results highlight its potential as an efficient hardware platform for spiking transformer.

cs.AR↗

arXiv2025

Pushing up to the Limit of Memory Bandwidth and Capacity Utilization for Efficient LLM Decoding on Embedded FPGA

The extremely high computational and storage demands of large language models have excluded most edge devices, which were widely used for efficient machine learning, from being viable options. A typical edge device usually only has 4GB of memory capacity and a bandwidth of less than 20GB/s, while a large language model quantized to 4-bit precision with 7B parameters already requires 3.5GB of capacity, and its decoding process is purely bandwidth-bound. In this paper, we aim to explore these limits by proposing a hardware accelerator for large language model (LLM) inference on the Zynq-based KV260 platform, equipped with 4GB of 64-bit 2400Mbps DDR4 memory. We successfully deploy a LLaMA2-7B model, achieving a decoding speed of around 5 token/s, utilizing 93.3% of the memory capacity and reaching 85% decoding speed of the theoretical memory bandwidth limit. To fully reserve the memory capacity for model weights and key-value cache, we develop the system in a bare-metal environment without an operating system. To fully reserve the bandwidth for model weight transfers, we implement a customized dataflow with an operator fusion pipeline and propose a data arrangement format that can maximize the data transaction efficiency. This research marks the first attempt to deploy a 7B level LLM on a standalone embedded field programmable gate array (FPGA) device. It provides key insights into efficient LLM inference on embedded FPGA devices and provides guidelines for future architecture design.

cs.AR↗

arXiv2025

$SpikePack$: Enhanced Information Flow in Spiking Neural Networks with High Hardware Compatibility

Spiking Neural Networks (SNNs) hold promise for energy-efficient, biologically inspired computing. We identify substantial informatio loss during spike transmission, linked to temporal dependencies in traditional Leaky Integrate-and-Fire (LIF) neuron-a key factor potentially limiting SNN performance. Existing SNN architectures also underutilize modern GPUs, constrained by single-bit spike storage and isolated weight-spike operations that restrict computational efficiency. We introduce ${SpikePack}$, a neuron model designed to reduce transmission loss while preserving essential features like membrane potential reset and leaky integration. ${SpikePack}$ achieves constant $\mathcal{O}(1)$ time and space complexity, enabling efficient parallel processing on GPUs and also supporting serial inference on existing SNN hardware accelerators. Compatible with standard Artificial Neural Network (ANN) architectures, ${SpikePack}$ facilitates near-lossless ANN-to-SNN conversion across various networks. Experimental results on tasks such as image classification, detection, and segmentation show ${SpikePack}$ achieves significant gains in accuracy and efficiency for both directly trained and converted SNNs over state-of-the-art models. Tests on FPGA-based platforms further confirm cross-platform flexibility, delivering high performance and enhanced sparsity. By enhancing information flow and rethinking SNN-ANN integration, ${SpikePack}$ advances efficient SNN deployment across diverse hardware platforms.

cs.NE↗

arXiv2025

Cogito, ergo sum: A Neurobiologically-Inspired Cognition-Memory-Growth System for Code Generation

Large language models based Multi Agent Systems (MAS) have demonstrated promising performance for enhancing the efficiency and accuracy of code generation tasks. However,most existing methods follow a conventional sequence of planning, coding, and debugging,which contradicts the growth-driven nature of human learning process. Additionally,the frequent information interaction between multiple agents inevitably involves high computational costs. In this paper,we propose Cogito,a neurobiologically inspired multi-agent framework to enhance the problem-solving capabilities in code generation tasks with lower cost. Specifically,Cogito adopts a reverse sequence: it first undergoes debugging, then coding,and finally planning. This approach mimics human learning and development,where knowledge is acquired progressively. Accordingly,a hippocampus-like memory module with different functions is designed to work with the pipeline to provide quick retrieval in similar tasks. Through this growth-based learning model,Cogito accumulates knowledge and cognitive skills at each stage,ultimately forming a Super Role an all capable agent to perform the code generation task. Extensive experiments against representative baselines demonstrate the superior performance and efficiency of Cogito. The code is publicly available at https://anonymous.4open.science/r/Cogito-0083.

cs.SE↗

arXiv2024

Data-Efficient CLIP-Powered Dual-Branch Networks for Source-Free Unsupervised Domain Adaptation

Source-free Unsupervised Domain Adaptation (SF-UDA) aims to transfer a model's performance from a labeled source domain to an unlabeled target domain without direct access to source samples, addressing critical data privacy concerns. However, most existing SF-UDA approaches assume the availability of abundant source domain samples, which is often impractical due to the high cost of data annotation. To address the dual challenges of limited source data and privacy concerns, we introduce a data-efficient, CLIP-powered dual-branch network (CDBN). This architecture consists of a cross-domain feature transfer branch and a target-specific feature learning branch, leveraging high-confidence target domain samples to transfer text features of source domain categories while learning target-specific soft prompts. By fusing the outputs of both branches, our approach not only effectively transfers source domain category semantic information to the target domain but also reduces the negative impacts of noise and domain gaps during target training. Furthermore, we propose an unsupervised optimization strategy driven by accurate classification and diversity, preserving the classification capability learned from the source domain while generating more confident and diverse predictions in the target domain. CDBN achieves near state-of-the-art performance with far fewer source domain samples than existing methods across 31 transfer tasks on seven datasets.

cs.CV↗