Real-Time Privacy Risk Measurement with Privacy Tokens for Gradient Leakage

The widespread deployment of deep learning models in privacy-sensitive domains has amplified concerns regarding privacy risks, particularly those stemming from gradient leakage during training. Current privacy assessments primarily rely on post-training attack simulations. However, these methods are inherently reactive, unable to encompass all potential attack scenarios, and often based on idealized adversarial assumptions. These limitations underscore the need for proactive approaches to privacy risk assessment during the training process. To address this gap, we propose the concept of privacy tokens, which are derived directly from private gradients during training. Privacy tokens encapsulate gradient features and, when combined with data features, offer valuable insights into the extent of private information leakage from training data, enabling real-time measurement of privacy risks without relying on adversarial attack simulations. Additionally, we employ Mutual Information (MI) as a robust metric to quantify the relationship between training data and gradients, providing precise and continuous assessments of privacy leakage throughout the training process. Extensive experiments validate our framework, demonstrating the effectiveness of privacy tokens and MI in identifying and quantifying privacy risks. This proactive approach marks a significant advancement in privacy monitoring, promoting the safer deployment of deep learning models in sensitive applications.

Privacy Token: Surprised to Find Out What You Accidentally Revealed

The widespread deployment of deep learning models in privacy-sensitive domains has amplified concerns regarding privacy risks, particularly those stemming from gradient leakage during training. Current privacy assessments primarily rely on post-training attack simulations. However, these methods are inherently reactive, unable to encompass all potential attack scenarios, and often based on idealized adversarial assumptions. These limitations underscore the need for proactive approaches to privacy risk assessment during the training process. To address this gap, we propose the concept of privacy tokens, which are derived directly from private gradients during training. Privacy tokens encapsulate gradient features and, when combined with data features, offer valuable insights into the extent of private information leakage from training data, enabling real-time measurement of privacy risks without relying on adversarial attack simulations. Additionally, we employ Mutual Information (MI) as a robust metric to quantify the relationship between training data and gradients, providing precise and continuous assessments of privacy leakage throughout the training process. Extensive experiments validate our framework, demonstrating the effectiveness of privacy tokens and MI in identifying and quantifying privacy risks. This proactive approach marks a significant advancement in privacy monitoring, promoting the safer deployment of deep learning models in sensitive applications.

Can LLMs Maintain Fundamental Abilities under KV Cache Compression?

This paper investigates an under-explored challenge in large language models (LLMs): the impact of KV cache compression methods on LLMs' fundamental capabilities. While existing methods achieve impressive compression ratios on long-context benchmarks, their effects on core model capabilities remain understudied. We present a comprehensive empirical study evaluating prominent KV cache compression methods across diverse tasks, spanning world knowledge, commonsense reasoning, arithmetic reasoning, code generation, safety, and long-context understanding and generation.Our analysis reveals that KV cache compression methods exhibit task-specific performance degradation. Arithmetic reasoning tasks prove particularly sensitive to aggressive compression, with different methods showing performance drops of $17.4\%$-$43.3\%$. Notably, the DeepSeek R1 Distill model exhibits more robust compression tolerance compared to instruction-tuned models, showing only $9.67\%$-$25.53\%$ performance degradation. Based on our analysis of attention patterns and cross-task compression performance, we propose ShotKV, a novel compression approach that distinctly handles prefill and decoding phases while maintaining shot-level semantic coherence. Empirical results show that ShotKV achieves $9\%$-$18\%$ performance improvements on long-context generation tasks under aggressive compression ratios.

LoRS: Efficient Low-Rank Adaptation for Sparse Large Language Model

Existing low-rank adaptation (LoRA) methods face challenges on sparse large language models (LLMs) due to the inability to maintain sparsity. Recent works introduced methods that maintain sparsity by augmenting LoRA techniques with additional masking mechanisms. Despite these successes, such approaches suffer from an increased memory and computation overhead, which affects efficiency of LoRA methods. In response to this limitation, we introduce LoRS, an innovative method designed to achieve both memory and computation efficiency when fine-tuning sparse LLMs. To mitigate the substantial memory and computation demands associated with preserving sparsity, our approach incorporates strategies of weight recompute and computational graph rearrangement. In addition, we also improve the effectiveness of LoRS through better adapter initialization. These innovations lead to a notable reduction in memory and computation consumption during the fine-tuning phase, all while achieving performance levels that outperform existing LoRA approaches.

Knowledge Sharing-enabled Semantic Rate Maximization for Multi-cell Task-oriented Hybrid Semantic-Bit Communication Networks

In task-oriented semantic communications, the transmitters are designed to deliver task-related semantic information rather than every signal bit to receivers, which alleviates the spectrum pressure by reducing network traffic loads. Effective semantic communications depend on the perfect alignment of shared knowledge between transmitters and receivers, however, the alignment of knowledge cannot always be guaranteed in practice. To tackle this challenge, we propose a novel knowledge sharing-enabled task-oriented hybrid semantic and bit communications mechanism, where a mobile device (MD) can proactively share and upload the task-related mismatched knowledge to associated small base station (SBS). The traditional bit communications can be adopted as an aid to transmit the rest data related to unshared mismatched knowledge to guarantee the effective execution of target tasks. Considering the heterogeneous transceivers in multi-cell networks, target task demands, and channel conditions, an optimization problem is formulated to maximize the generalized effective semantic transmission rate of all MDs by jointly optimizing knowledge sharing, semantic extraction ratio, and SBS association, while satisfying the semantic accuracy requirements and delay tolerances of MD target tasks. The formulated mixed integer nonlinear programming problem is decomposed into multiple subproblems equivalently. An optimum algorithm is proposed and another efficient algorithm is further developed using hierarchical class partitioning and monotonic optimization. Simulation results demonstrate the validity and superior performance of proposed solutions.

Collaborative Knowledge Sharing-empowered Effective Semantic Rate Maximization for Two-tier Semantic-Bit Communication Networks

Effective task-oriented semantic communications relies on perfect knowledge alignment between transmitters and receivers for accurate recovery of task-related semantic information, which can be susceptible to knowledge misalignment and performance degradation in practice. To tackle this issue, continual knowledge updating and sharing are crucial to adapt to evolving task and user related demands, despite the incurred resource overhead and increased latency. In this paper, we propose a novel collaborative knowledge sharing-empowered semantic transmission mechanism in a two-tier edge network, exploiting edge cooperations and bit communications to address KB mismatch. By deriving a generalized effective semantic transmission rate (GESTR) that considers both semantic accuracy and overhead, we formulate a mixed integer nonlinear programming problem to maximize GESTR of all mobile devices by optimizing knowledge sharing decisions, extraction ratios, and BS/subchannel allocations, subject to task accuracy and delay requirements. The joint optimum solution can be obtained by proposed fractional programming based branch and bound algorithm and modified Kuhn-Munkres algorithm efficiently. Simulation results demonstrate the superior performance of proposed solution, especially in low signal-to-noise conditions.

Personalized Clustering via Targeted Representation Learning

Clustering traditionally aims to reveal a natural grouping structure model from unlabeled data. However, this model may not always align with users' preference. In this paper, we propose a personalized clustering method that explicitly performs targeted representation learning by interacting with users via modicum task information (e.g., $\textit{must-link}$ or $\textit{cannot-link}$ pairs) to guide the clustering direction. We query users with the most informative pairs, i.e., those pairs most hard to cluster and those most easy to miscluster, to facilitate the representation learning in terms of the clustering preference. Moreover, by exploiting attention mechanism, the targeted representation is learned and augmented. By leveraging the targeted representation and constrained constrastive loss as well, personalized clustering is obtained. Theoretically, we verify that the risk of personalized clustering is tightly bounded, guaranteeing that active queries to users do mitigate the clustering risk. Experimentally, extensive results show that our method performs well across different clustering tasks and datasets, even with a limited number of queries.

RAC3: Retrieval-Augmented Corner Case Comprehension for Autonomous Driving with Vision-Language Models

Understanding and addressing corner cases is essential for ensuring the safety and reliability of autonomous driving systems. Vision-Language Models (VLMs) play a crucial role in enhancing scenario comprehension, yet they face significant challenges, such as hallucination and insufficient real-world grounding, which compromise their performance in critical driving scenarios. In this work, we propose RAC3, a novel framework designed to improve VLMs' ability to handle corner cases effectively. The framework integrates Retrieval-Augmented Generation (RAG) to mitigate hallucination by dynamically incorporating context-specific external knowledge. A cornerstone of RAC3 is its cross-modal alignment fine-tuning, which utilizes contrastive learning to embed image-text pairs into a unified semantic space, enabling robust retrieval of similar scenarios. We evaluate RAC3 through extensive experiments using a curated dataset of corner case scenarios, demonstrating its ability to enhance semantic alignment, improve hallucination mitigation, and achieve superior performance metrics, such as Cosine Similarity and ROUGE-L scores. For example, for the LLaVA-v1.6-34B VLM, the cosine similarity between the generated text and the reference text has increased by 5.22\%. The F1-score in ROUGE-L has increased by 39.91\%, the Precision has increased by 55.80\%, and the Recall has increased by 13.74\%. This work underscores the potential of retrieval-augmented VLMs to advance the robustness and safety of autonomous driving in complex environments.

TSGaussian: Semantic and Depth-Guided Target-Specific Gaussian Splatting from Sparse Views

Recent advances in Gaussian Splatting have significantly advanced the field, achieving both panoptic and interactive segmentation of 3D scenes. However, existing methodologies often overlook the critical need for reconstructing specified targets with complex structures from sparse views. To address this issue, we introduce TSGaussian, a novel framework that combines semantic constraints with depth priors to avoid geometry degradation in challenging novel view synthesis tasks. Our approach prioritizes computational resources on designated targets while minimizing background allocation. Bounding boxes from YOLOv9 serve as prompts for Segment Anything Model to generate 2D mask predictions, ensuring semantic accuracy and cost efficiency. TSGaussian effectively clusters 3D gaussians by introducing a compact identity encoding for each Gaussian ellipsoid and incorporating 3D spatial consistency regularization. Leveraging these modules, we propose a pruning strategy to effectively reduce redundancy in 3D gaussians. Extensive experiments demonstrate that TSGaussian outperforms state-of-the-art methods on three standard datasets and a new challenging dataset we collected, achieving superior results in novel view synthesis of specific objects. Code is available at: https://github.com/leon2000-ai/TSGaussian.

SAM Decoding: Speculative Decoding via Suffix Automaton

Large Language Models (LLMs) have revolutionized natural language processing by unifying tasks into text generation, yet their large parameter sizes and autoregressive nature limit inference speed. SAM-Decoding addresses this by introducing a novel retrieval-based speculative decoding method that uses a suffix automaton for efficient and accurate draft generation. Unlike n-gram matching used by the existing method, SAM-Decoding finds the longest suffix match in generating text and text corpuss, achieving an average time complexity of $O(1)$ per generation step. SAM-Decoding constructs static and dynamic suffix automatons for the text corpus and input prompts, respectively, enabling fast and precise draft generation. Meanwhile, it is designed as an approach that can be combined with existing methods, allowing SAM-Decoding to adaptively select a draft generation strategy based on the matching length, thus increasing the inference speed of the LLM. When combined with Token Recycling, evaluations show SAM-Decoding outperforms existing model-free methods, achieving a speedup of $2.27\times$ over autoregressive decoding on Spec-Bench. When combined with EAGLE2, it reaches a speedup of $2.49\times$, surpassing all current approaches. Our code is available at https://github.com/hyx1999/SAM-Decoding.