Stable-RAG: Mitigating Retrieval-Permutation-Induced Hallucinations in Retrieval-Augmented Generation

Retrieval-Augmented Generation (RAG) has become a key paradigm for reducing factual hallucinations in large language models (LLMs), yet little is known about how the order of retrieved documents affects model behavior. We empirically show that under Top-5 retrieval with the gold document included, LLM answers vary substantially across permutations of the retrieved set, even when the gold document is fixed in the first position. This reveals a previously underexplored sensitivity to retrieval permutations. Although robust RAG methods primarily focus on enhancing LLM robustness to low-quality retrieval and mitigating positional bias to distribute attention fairly over long contexts, neither approach directly addresses permutation sensitivity. In this paper, we propose Stable-RAG, which exploits permutation sensitivity estimation to mitigate permutation-induced hallucinations. Stable-RAG runs the generator under multiple retrieval orders, clusters hidden states, and decodes from a cluster-center representation that captures the dominant reasoning pattern. It then uses these reasoning results to align hallucinated outputs toward the correct answer, encouraging the model to produce consistent and accurate predictions across document permutations. Experiments on three QA datasets show that Stable-RAG significantly improves answer accuracy, reasoning consistency and robust generalization across datasets, retrievers, and input lengths compared with baselines.

Parameter Importance-Driven Continual Learning for Foundation Models

Domain-specific post-training often causes catastrophic forgetting, making foundation models lose their general reasoning ability and limiting their adaptability to dynamic real-world environments. Preserving general capabilities while acquiring downstream domain knowledge is a central challenge for large language and multimodal models. Traditional continual learning methods, such as regularization, replay and architectural isolation, suffer from poor downstream performance, reliance on inaccessible historical data, or additional parameter overhead. While recent parameter-efficient tuning (PET) methods can alleviate forgetting, their effectiveness strongly depends on the choice of parameters and update strategies. In this paper, we introduce PIECE, a Parameter Importance Estimation-based Continual Enhancement method that preserves general ability while efficiently learning domain knowledge without accessing prior training data or increasing model parameters. PIECE selectively updates only 0.1% of core parameters most relevant to new tasks, guided by two importance estimators: PIECE-F based on Fisher Information, and PIECE-S based on a second-order normalization that combines gradient and curvature information. Experiments across three language models and two multimodal models show that PIECE maintains general capabilities and achieves state-of-the-art continual learning performance across diverse downstream tasks. Our results highlight a practical path to scalable, domain-adaptive foundation models without catastrophic forgetting.

Unlocking Dynamic Inter-Client Spatial Dependencies: A Federated Spatio-Temporal Graph Learning Method for Traffic Flow Forecasting

Spatio-temporal graphs are powerful tools for modeling complex dependencies in traffic time series. However, the distributed nature of real-world traffic data across multiple stakeholders poses significant challenges in modeling and reconstructing inter-client spatial dependencies while adhering to data locality constraints. Existing methods primarily address static dependencies, overlooking their dynamic nature and resulting in suboptimal performance. In response, we propose Federated Spatio-Temporal Graph with Dynamic Inter-Client Dependencies (FedSTGD), a framework designed to model and reconstruct dynamic inter-client spatial dependencies in federated learning. FedSTGD incorporates a federated nonlinear computation decomposition module to approximate complex graph operations. This is complemented by a graph node embedding augmentation module, which alleviates performance degradation arising from the decomposition. These modules are coordinated through a client-server collective learning protocol, which decomposes dynamic inter-client spatial dependency learning tasks into lightweight, parallelizable subtasks. Extensive experiments on four real-world datasets demonstrate that FedSTGD achieves superior performance over state-of-the-art baselines in terms of RMSE, MAE, and MAPE, approaching that of centralized baselines. Ablation studies confirm the contribution of each module in addressing dynamic inter-client spatial dependencies, while sensitivity analysis highlights the robustness of FedSTGD to variations in hyperparameters.

Detecting Stealthy Backdoor Samples based on Intra-class Distance for Large Language Models

Fine-tuning LLMs with datasets containing stealthy backdoors from publishers poses security risks to downstream applications. Mainstream detection methods either identify poisoned samples by analyzing the prediction probability of poisoned classification models or rely on the rewriting model to eliminate the stealthy triggers. However, the former cannot be applied to generation tasks, while the latter may degrade generation performance and introduce new triggers. Therefore, efficiently eliminating stealthy poisoned samples for LLMs remains an urgent problem. We observe that after applying TF-IDF clustering to the sample response, there are notable differences in the intra-class distances between clean and poisoned samples. Poisoned samples tend to cluster closely because of their specific malicious outputs, whereas clean samples are more scattered due to their more varied responses. Thus, in this paper, we propose a stealthy backdoor sample detection method based on Reference-Filtration and Tfidf-Clustering mechanisms (RFTC). Specifically, we first compare the sample response with the reference model's outputs and consider the sample suspicious if there's a significant discrepancy. And then we perform TF-IDF clustering on these suspicious samples to identify the true poisoned samples based on the intra-class distance. Experiments on two machine translation datasets and one QA dataset demonstrate that RFTC outperforms baselines in backdoor detection and model performance. Further analysis of different reference models also confirms the effectiveness of our Reference-Filtration.

A Federated Splitting Framework for LLMs: Security, Efficiency, and Adaptability

Private data is typically larger and of higher quality than public data, offering great potential to improve LLM. However, its scattered distribution across data silos and the high computational demands of LLMs limit their deployment in federated environments. To address this, the transformer-based split learning model has emerged, offloading most model parameters to the server while retaining only the embedding and output layers on clients to ensure privacy. However, it still faces significant challenges in security, efficiency, and adaptability: 1) embedding gradients are vulnerable to attacks, leading to reverse engineering of private data; 2) the autoregressive nature of LLMs means that federated split learning can only train and infer sequentially, causing high communication overhead; 3) fixed partition points lack adaptability to downstream tasks. In this paper, we introduce FL-LLaMA, a secure, efficient, and adaptive federated split framework based on LLaMA2. First, we place some input and output blocks on the local client and inject Gaussian noise into forward-pass hidden states, enabling secure end-to-end propagation. Second, we employ client-batch and server-hierarchical strategies to achieve parallel training, along with attention-mask compression and KV cache mechanisms to accelerate inference, reducing communication costs effectively. Third, we allow users to dynamically adjust the partition points for input/output blocks based on specific task requirements and hardware limitations. Experiments on NLU, summarization and conversational QA tasks show that FL-LLaMA maintains performance comparable to centralized LLaMA2, and achieves up to 2x train speedups and 8x inference speedups. Further analysis of privacy attacks and different partition points also demonstrates the effectiveness of FL-LLaMA in security and adaptability.

CodeBC: A More Secure Large Language Model for Smart Contract Code Generation in Blockchain

Large language models (LLMs) excel at generating code from natural language instructions, yet they often lack an understanding of security vulnerabilities. This limitation makes it difficult for LLMs to avoid security risks in generated code, particularly in high-security programming tasks such as smart contract development for blockchain. Researchers have attempted to enhance the vulnerability awareness of these models by training them to differentiate between vulnerable and fixed code snippets. However, this approach relies heavily on manually labeled vulnerability data, which is only available for popular languages like Python and C++. For low-resource languages like Solidity, used in smart contracts, large-scale annotated datasets are scarce and difficult to obtain. To address this challenge, we introduce CodeBC, a code generation model specifically designed for generating secure smart contracts in blockchain. CodeBC employs a three-stage fine-tuning approach based on CodeLlama, distinguishing itself from previous methods by not relying on pairwise vulnerability location annotations. Instead, it leverages vulnerability and security tags to teach the model the differences between vulnerable and secure code. During the inference phase, the model leverages security tags to generate secure and robust code. Experimental results demonstrate that CodeBC outperforms baseline models in terms of BLEU, CodeBLEU, and compilation pass rates, while significantly reducing vulnerability rates. These findings validate the effectiveness and cost-efficiency of our three-stage fine-tuning strategy, making CodeBC a promising solution for generating secure smart contract code.

Learning to Erase Private Knowledge from Multi-Documents for Retrieval-Augmented Large Language Models

Retrieval-Augmented Generation (RAG) is a promising technique for applying LLMs to proprietary domains. However, retrieved documents may contain sensitive knowledge, posing risks of privacy leakage in generative results. Thus, effectively erasing private information from retrieved documents is a key challenge for RAG. Unlike traditional text anonymization, RAG should consider: (1) the inherent multi-document reasoning may face de-anonymization attacks; (2) private knowledge varies by scenarios, so users should be allowed to customize which information to erase; (3) preserving sufficient publicly available knowledge for generation tasks. This paper introduces the privacy erasure task for RAG and proposes Eraser4RAG, a private knowledge eraser which effectively removes user-defined private knowledge from documents while preserving sufficient public knowledge for generation. Specifically, we first construct a global knowledge graph to identify potential knowledge across documents, aiming to defend against de-anonymization attacks. Then we randomly split it into private and public sub-graphs, and fine-tune Flan-T5 to rewrite the retrieved documents excluding private triples. Finally, PPO algorithm optimizes the rewriting model to minimize private triples and maximize public triples retention. Experiments on four QA datasets demonstrate that Eraser4RAG achieves superior erase performance than GPT-4o.

FineFilter: A Fine-grained Noise Filtering Mechanism for Retrieval-Augmented Large Language Models

Retrieved documents containing noise will hinder Retrieval-Augmented Generation (RAG) from detecting answer clues, necessitating noise filtering mechanisms to enhance accuracy. Existing methods use re-ranking or summarization to identify the most relevant sentences, but directly and accurately locating answer clues from these large-scale and complex documents remains challenging. Unlike these document-level operations, we treat noise filtering as a sentence-level MinMax optimization problem: first identifying the potential clues from multiple documents using contextual information, then ranking them by relevance, and finally retaining the least clues through truncation. In this paper, we propose FineFilter, a novel fine-grained noise filtering mechanism for RAG consisting of a clue extractor, a re-ranker, and a truncator. We optimize each module to tackle complex reasoning challenges: (1) Clue extractor firstly uses sentences containing the answer and similar ones as fine-tuned targets, aiming at extracting sufficient potential clues; (2) Re-ranker is trained to prioritize effective clues based on the real feedback from generation module, with clues capable of generating correct answer as positive samples and others as negative; (3) Truncator takes the minimum clues needed to answer the question (truncation point) as fine-tuned targets, and performs truncation on the re-ranked clues to achieve fine-grained noise filtering. Experiments on three QA datasets demonstrate that FineFilter significantly outperforms baselines in terms of performance and inference cost. Further analysis on each module shows the effectiveness of our optimizations for complex reasoning.

AdaComp: Extractive Context Compression with Adaptive Predictor for Retrieval-Augmented Large Language Models

Retrieved documents containing noise will hinder RAG from detecting answer clues and make the inference process slow and expensive. Therefore, context compression is necessary to enhance its accuracy and efficiency. Existing context compression methods use extractive or generative models to retain the most query-relevant sentences or apply the information bottleneck theory to preserve sufficient information. However, these methods may face issues such as over-compression or high computational costs. We observe that the retriever often ranks relevant documents at the top, but the exact number of documents needed to answer the query is uncertain due to the impact of query complexity and retrieval quality: complex queries like multi-hop questions may require retaining more documents than simpler queries, and a low-quality retrieval may need to rely on more documents to generate accurate outputs. Therefore, determining the minimum number of required documents (compression rate) is still a challenge for RAG. In this paper, we introduce AdaComp, a low-cost extractive context compression method that adaptively determines the compression rate based on both query complexity and retrieval quality. Specifically, we first annotate the minimum top-k documents necessary for the RAG system to answer the current query as the compression rate and then construct triplets of the query, retrieved documents, and its compression rate. Then, we use this triplet dataset to train a compression-rate predictor. Experiments on three QA datasets and one conversational Muiti-doc QA dataset show that AdaComp significantly reduces inference costs while maintaining performance nearly identical to uncompressed models, achieving a balance between efficiency and performance.

HSF: Defending against Jailbreak Attacks with Hidden State Filtering

With the growing deployment of LLMs in daily applications like chatbots and content generation, efforts to ensure outputs align with human values and avoid harmful content have intensified. However, increasingly sophisticated jailbreak attacks threaten this alignment, aiming to induce unsafe outputs. Current defense efforts either focus on prompt rewriting or detection, which are limited in effectiveness due to the various design of jailbreak prompts, or on output control and detection, which are computationally expensive as they require LLM inference. Therefore, designing a pre-inference defense method that resists diverse jailbreak prompts is crucial for preventing LLM jailbreak attacks. We observe that jailbreak attacks, safe queries, and harmful queries exhibit different clustering patterns within the LLM's hidden state representation space. This suggests that by leveraging the LLM's hidden state representational capabilities, we can analyze the LLM's forthcoming behavior and proactively intervene for defense. In this paper, we propose a jailbreak attack defense strategy based on a Hidden State Filter (HSF), a lossless architectural defense mechanism that enables the model to preemptively identify and reject adversarial inputs before the inference process begins. We activate its defensive potential through an additional plugin module, effectively framing the defense task as a classification problem. Experimental results on two benchmark datasets, utilizing three different LLMs, show that HSF significantly enhances resilience against six cutting-edge jailbreak attacks. It significantly reduces the success rate of jailbreak attacks while minimally impacting responses to benign user queries, with negligible inference overhead, and outperforming defense baselines.Our code and data are available at https://anonymous.4open.science/r/Hidden-State-Filtering-8652/