Embedding-as-a-Service (EaaS) is an effective and convenient deployment solution for addressing various NLP tasks. Nevertheless, recent research has shown that EaaS is vulnerable to model extraction attacks, which could lead to significant economic losses for model providers. For copyright protection, existing methods inject watermark embeddings into text embeddings and use them to detect copyright infringement. However, current watermarking methods often resist only a subset of attacks and fail to provide comprehensive protection. To this end, we present the region-triggered semantic watermarking framework called RegionMarker, which defines trigger regions within a low-dimensional space and injects watermarks into text embeddings associated with these regions. By utilizing a secret dimensionality reduction matrix to project onto this subspace and randomly selecting trigger regions, RegionMarker makes it difficult for watermark removal attacks to evade detection. Furthermore, by embedding watermarks across the entire trigger region and using the text embedding as the watermark, RegionMarker is resilient to both paraphrasing and dimension-perturbation attacks. Extensive experiments on various datasets show that RegionMarker is effective in resisting different attack methods, thereby protecting the copyright of EaaS.

Zifeng Cheng, Zhonghui Wang, Yuchen Fu, Zhiwei Jiang, Yafeng Yin, Cong Wang, Qing Gu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Yuchen Fu, Zifeng Cheng, Zhiwei Jiang, Zhonghui Wang, Yafeng Yin, Zhengliang Li, Qing Gu. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Large language models (LLMs) have achieved remarkable performance across many generation tasks. Nevertheless, effectively aligning them with desired behaviors remains a significant challenge. Activation steering is an effective and cost-efficient approach that directly modifies the activations of LLMs during the inference stage, aligning their responses with the desired behaviors and avoiding the high cost of fine-tuning. Existing methods typically indiscriminately intervene to all generations or rely solely on the question to determine intervention, which limits the accurate assessment of the intervention strength. To this end, we propose the **F**lexible **A**ctivation **S**teering with **B**acktracking (**FASB**) framework, which dynamically determines both the necessity and strength of intervention by tracking the internal states of the LLMs during generation, considering both the question and the generated content. Since intervening after detecting a deviation from the desired behavior is often too late, we further propose the backtracking mechanism to correct the deviated tokens and steer the LLMs toward the desired behavior. Extensive experiments on the TruthfulQA dataset and six multiple-choice datasets demonstrate that our method outperforms baselines. Our code will be released at https://github.com/gjw185/FASB.

Recent advancements in Automatic Prompt Optimization (APO) for text-to-image generation have streamlined user input while ensuring high-quality image output. However, most APO methods are trained assuming a fixed text-to-image model, which is impractical given the emergence of new models. To address this, we propose a novel task, model-generalized automatic prompt optimization (MGAPO), which trains APO methods on a set of known models to enable generalization to unseen models during testing. MGAPO presents significant challenges. First, we experimentally confirm the suboptimal performance of existing APO methods on unseen models. We then introduce a two-stage prompt optimization method, AP-Adapter. In the first stage, a large language model is used to rewrite the prompts. In the second stage, we propose a novel method to construct an enhanced representation space by leveraging inter-model differences. This space captures the characteristics of multiple domain models, storing them as domain prototypes. These prototypes serve as anchors to adjust prompt representations, enabling generalization to unseen models. The optimized prompt representations are subsequently used to generate conditional representations for controllable image generation. We curate a multi-modal, multi-model dataset that includes multiple diffusion models and their corresponding text-image data, and conduct experiments under a model generalization setting. The experimental results demonstrate the AP-Adapter's ability to enable the automatic prompts to generalize well to previously unseen diffusion models, generating high-quality images.

Behavioral decision systems may suffer from poor performance due to the failure in capturing the vibrations of environmental information. To better capture such vibrations and then make more accurate predictions, a parallel deep neural network based on dual fusions including feature and decision is proposed, called DFFD-Net. DFFD-NET is composed of two parts, the feature fusion network and the driving data network. The feature fusion model adopts two different operations, deconvolution and linear weighting, to fuse local features and global features, respectively. Deconvolution is applied between the convolutional layers, while linear weighting is operated among the outputs of SPP and LSTM. To further improve the accuracy of the prediction, the decisions generated from both networks are further weighed to get the final decision. Experimentally, DFFD-NET is implemented in the benchmarks BDDV and TORCS, and the results show that the final performance is benefited from both feature fusion and decision fusion. From the comparison, DFFD-NET can get state-of-the-art results on both perplexity and precision by only using the images captured from the front-facing camera as well as a few sensing data.

Recent ground-breaking efforts such as CompCert have made a convincing case that mechanized verification of the compiler correctness for realistic C programs is both viable and practical. Unfortunately, existing verified compilers can only handle whole programs---this severely limits their applicability and prevents the linking of verified C programs with verified external libraries. In this paper, we present a novel compositional semantics for reasoning about open modules and for supporting verified separate compilation and linking. More specifically, we replace external function calls with explicit events in the behavioral semantics. We then develop a verified linking operator that makes lazy substitutions on (potentially reacting) behaviors by replacing each external function call event with a behavior simulating the requested function. Finally, we show how our new semantics can be applied to build a refinement infrastructure that supports both vertical composition and horizontal composition.