Dense retrievers are a core component of Retrieval-Augmented Generation (RAG) systems, yet their growing adoption introduces critical security risks in the Al supply chain. Existing retrieval backdoor attacks primarily rely on coarse-grained parameter updates, which often degrade benign retrieval behavior and require substantial computational cost, limiting their stealth and practicality across deployment settings. In this paper, we propose LOKI (Locating Optimal Knock-In), a framework that reformulates stealthy retrieval backdoor injection as a neuron-level local editing problem. Following a Locate-then-Edit paradigm, LOKI first identifies pivotal neurons that are highly sensitive to triggers yet decoupled from benign retrieval semantics via a multi-dimensional diagnostic analysis. And then LOKI implants backdoor logic through constrained local editing that balances trigger-induced alignment with semantic manifold preservation. Extensive experiments show that LOKI achieves near 100% attack success while modifying fewer than 1% of model parameters, with only limited impact on clean retrieval performance.

Previous Sign Language Translation (SLT) methods achieve superior performance by relying on gloss annotations. However, labeling high-quality glosses is a labor-intensive task, which limits the further development of SLT. Although some approaches work towards gloss-free SLT through jointly training the visual encoder and translation network, these efforts still suffer from poor performance and inefficient use of the powerful Large Language Model (LLM). Most seriously, we find that directly introducing LLM into SLT will lead to insufficient learning of visual representations as LLM dominates the learning curve. To address these problems, we propose Factorized Learning assisted with Large Language Model (FLa-LLM) for gloss-free SLT. Concretely, we factorize the training process into two stages. In the visual initialing stage, we employ a lightweight translation model after the visual encoder to pre-train the visual encoder. In the LLM fine-tuning stage, we freeze the acquired knowledge in the visual encoder and integrate it with a pre-trained LLM to inspire the LLM’s translation potential. This factorized training strategy proves to be highly effective as evidenced by significant improvements achieved across three SLT datasets which are all conducted under the gloss-free setting.

Sequential decision-making refers to algorithms that take into account the dynamics of the environment, where early decisions affect subsequent decisions. With large language models (LLMs) demonstrating powerful capabilities between tasks, we can’t help but ask: Can Current LLMs Effectively Make Sequential Decisions? In order to answer this question, we propose the UNO Arena based on the card game UNO to evaluate the sequential decision-making capability of LLMs and explain in detail why we choose UNO. In UNO Arena, We evaluate the sequential decision-making capability of LLMs dynamically with novel metrics based Monte Carlo methods. We set up random players, DQN-based reinforcement learning players, and LLM players (e.g. GPT-4, Gemini-pro) for comparison testing. Furthermore, in order to improve the sequential decision-making capability of LLMs, we propose the TUTRI player, which can involves having LLMs reflect their own actions with the summary of game history and the game strategy. Numerous experiments demonstrate that the TUTRI player achieves a notable breakthrough in the performance of sequential decision-making compared to the vanilla LLM player.