Yuping Lin

Yuping Lin, Zitao Li, Yue Xing et al.

Recent studies have identified "retrieval heads" in Large Language Models (LLMs) responsible for extracting information from input contexts. However, prior works largely rely on static statistics aggregated across datasets, identifying heads that perform retrieval on average. This perspective overlooks the fine-grained temporal dynamics of autoregressive generation. In this paper, we investigate retrieval heads from a dynamic perspective. Through extensive analysis, we establish three core claims: (1) Dynamism: Retrieval heads vary dynamically across timesteps; (2) Irreplaceability: Dynamic retrieval heads are specific at each timestep and cannot be effectively replaced by static retrieval heads; and (3) Correlation: The model's hidden state encodes a predictive signal for future retrieval head patterns, indicating an internal planning mechanism. We validate these findings on the Needle-in-a-Haystack task and a multi-hop QA task, and quantify the differences on the utility of dynamic and static retrieval heads in a Dynamic Retrieval-Augmented Generation framework. Our study provides new insights into the internal mechanisms of LLMs.

Yingqian Cui, Jie Ren, Yuping Lin et al.

Text-to-image generative models, especially those based on latent diffusion models (LDMs), have demonstrated outstanding ability in generating high-quality and high-resolution images from textual prompts. With this advancement, various fine-tuning methods have been developed to personalize text-to-image models for specific applications such as artistic style adaptation and human face transfer. However, such advancements have raised copyright concerns, especially when the data are used for personalization without authorization. For example, a malicious user can employ fine-tuning techniques to replicate the style of an artist without consent. In light of this concern, we propose FT-Shield, a watermarking solution tailored for the fine-tuning of text-to-image diffusion models. FT-Shield addresses copyright protection challenges by designing new watermark generation and detection strategies. In particular, it introduces an innovative algorithm for watermark generation. It ensures the seamless transfer of watermarks from training images to generated outputs, facilitating the identification of copyrighted material use. To tackle the variability in fine-tuning methods and their impact on watermark detection, FT-Shield integrates a Mixture of Experts (MoE) approach for watermark detection. Comprehensive experiments validate the effectiveness of our proposed FT-Shield.

Yuping Lin, Jiayuan Ding, Yue Xing et al.

KV cache growth is a major bottleneck for long-context inference in large language models. Existing methods are often dominated by binary eviction or representation approximation, which may underutilize tokens that are not critical for exact retention but are still reconstructable. We present VECTOR, a plug-and-play augmentation for eviction-based pipelines that introduces three-way token routing: retention, approximation, and eviction. VECTOR combines an importance signal from the base scorer with a reconstructability signal from an offline-calibrated regression-based value estimation. By leveraging reconstructability, VECTOR recovers useful value information that would otherwise be irreversibly lost under binary eviction, while preserving key vectors for attention routing stability. Experimental results show that VECTOR improves quality-memory trade-offs under medium-to-high compression, with especially clear gains in stricter budget regimes.

Yuping Lin, Pengfei He, Yue Xing et al.

Supervised fine-tuning (SFT) induces new behaviors in large language models, yet imposes no structural constraint on how these behaviors are distributed within the model. Existing behavior interpretation methods, such as circuit attribution approaches, identify sparse subnetworks correlated with SFT-induced behaviors post-hoc. However, such correlations do not imply *causal necessity*, limiting the ability to selectively control SFT-induced behaviors at inference time. We pursue an alternative by asking: can an SFT-induced behavior be deliberately compressed into a sparse, mechanistically necessary subnetwork, termed a *carrier*, while remaining controllable at inference time without weight modification? We propose (a) **Loss-Constrained Dual Descent (LCDD)**, which constructs such carriers by jointly optimizing routing masks and model weights under an explicit utility budget, and (b) **SFT-Eraser**, a soft prompt optimized via activation matching on extracted carrier channels, to reverse the SFT-induced behavior. Across safety, fixed-response, and style behaviors on multiple model families, LCDD yields sparse carriers that preserve target behaviors while enabling strong reversion when triggered by SFT-Eraser. Ablations further establish that the sparse structure is the key precondition for reversal: the same trigger optimization fails on standard SFT models, confirming that structure rather than trigger design is the operative factor. These results provide direct evidence that the learned carriers are causally necessary for the behaviors, pointing to a new direction for systematically localizing and selectively suppressing SFT-induced behaviors in deployed models.

Zidi Xiong, Yuping Lin, Wenya Xie et al.

Memory is a critical component in large language model (LLM)-based agents, enabling them to store and retrieve past executions to improve task performance over time. In this paper, we conduct an empirical study on how memory management choices impact the LLM agents' behavior, especially their long-term performance. Specifically, we focus on two fundamental memory management operations that are widely used by many agent frameworks-memory addition and deletion-to systematically study their impact on the agent behavior. Through our quantitative analysis, we find that LLM agents display an experience-following property: high similarity between a task input and the input in a retrieved memory record often results in highly similar agent outputs. Our analysis further reveals two significant challenges associated with this property: error propagation, where inaccuracies in past experiences compound and degrade future performance, and misaligned experience replay, where some seemingly correct executions can provide limited or even misleading value as experiences. Through controlled experiments, we demonstrate the importance of regulating experience quality within the memory bank and show that future task evaluations can serve as free quality labels for stored memory. Our findings offer insights into the behavioral dynamics of LLM agent memory systems and provide practical guidance for designing memory components that support robust, long-term agent performance.

Shenglai Zeng, Jiankun Zhang, Bingheng Li et al.

Retrieval-Augmented Generation (RAG) systems have shown promise in enhancing the performance of Large Language Models (LLMs). However, these systems face challenges in effectively integrating external knowledge with the LLM's internal knowledge, often leading to issues with misleading or unhelpful information. This work aims to provide a systematic study on knowledge checking in RAG systems. We conduct a comprehensive analysis of LLM representation behaviors and demonstrate the significance of using representations in knowledge checking. Motivated by the findings, we further develop representation-based classifiers for knowledge filtering. We show substantial improvements in RAG performance, even when dealing with noisy knowledge databases. Our study provides new insights into leveraging LLM representations for enhancing the reliability and effectiveness of RAG systems.

Tai-Quan Peng, Kaiqi Yang, Sanguk Lee et al.

As large language models (LLMs) become increasingly embedded in civic, educational, and political information environments, concerns about their potential political bias have grown. Prior research often evaluates such bias through simulated personas or predefined ideological typologies, which may introduce artificial framing effects or overlook how models behave in general use scenarios. This study adopts a persona-free, topic-specific approach to evaluate political behavior in LLMs, reflecting how users typically interact with these systems-without ideological role-play or conditioning. We introduce a two-dimensional framework: one axis captures partisan orientation on highly polarized topics (e.g., abortion, immigration), and the other assesses sociopolitical engagement on less polarized issues (e.g., climate change, foreign policy). Using survey-style prompts drawn from the ANES and Pew Research Center, we analyze responses from 43 LLMs developed in the U.S., Europe, China, and the Middle East. We propose an entropy-weighted bias score to quantify both the direction and consistency of partisan alignment, and identify four behavioral clusters through engagement profiles. Findings show most models lean center-left or left ideologically and vary in their nonpartisan engagement patterns. Model scale and openness are not strong predictors of behavior, suggesting that alignment strategy and institutional context play a more decisive role in shaping political expression.

Yuping Lin, Pengfei He, Han Xu et al.

Large language models (LLMs) are susceptible to a type of attack known as jailbreaking, which misleads LLMs to output harmful contents. Although there are diverse jailbreak attack strategies, there is no unified understanding on why some methods succeed and others fail. This paper explores the behavior of harmful and harmless prompts in the LLM's representation space to investigate the intrinsic properties of successful jailbreak attacks. We hypothesize that successful attacks share some similar properties: They are effective in moving the representation of the harmful prompt towards the direction to the harmless prompts. We leverage hidden representations into the objective of existing jailbreak attacks to move the attacks along the acceptance direction, and conduct experiments to validate the above hypothesis using the proposed objective. We hope this study provides new insights into understanding how LLMs understand harmfulness information.

Yuping Lin, Kasra Ahmadi K. A., Hui Jiang

In this paper, we study the adversarial attack and defence problem in deep learning from the perspective of Fourier analysis. We first explicitly compute the Fourier transform of deep ReLU neural networks and show that there exist decaying but non-zero high frequency components in the Fourier spectrum of neural networks. We demonstrate that the vulnerability of neural networks towards adversarial samples can be attributed to these insignificant but non-zero high frequency components. Based on this analysis, we propose to use a simple post-averaging technique to smooth out these high frequency components to improve the robustness of neural networks against adversarial attacks. Experimental results on the ImageNet dataset have shown that our proposed method is universally effective to defend many existing adversarial attacking methods proposed in the literature, including FGSM, PGD, DeepFool and C&W attacks. Our post-averaging method is simple since it does not require any re-training, and meanwhile it can successfully defend over 95% of the adversarial samples generated by these methods without introducing any significant performance degradation (less than 1%) on the original clean images.