Alignment plays a crucial role in Large Language Models (LLMs) in aligning with human preferences on a specific task/domain. Traditional alignment methods suffer from catastrophic forgetting, where models lose previously learned values when adapting to new preferences or domains. We introduce LifeAlign, a novel framework for lifelong alignment that enables LLMs to maintain consistent human preference alignment across sequential learning tasks without forgetting previously learned values. Our approach consists of two key innovations. First, we propose a focalized preference optimization strategy that aligns LLMs with new preferences while preventing the erosion of alignment acquired from previous tasks. Second, we develop a short-to-long memory consolidation mechanism that merges denoised short-term preference representations into stable long-term memory using intrinsic dimensionality reduction, enabling efficient storage and retrieval of alignment patterns across diverse domains. We evaluate LifeAlign across multiple sequential alignment tasks spanning different domains and preference types. Experimental results demonstrate that our method achieves superior performance in maintaining both preference alignment quality and knowledge retention compared to existing lifelong learning approaches.

Euphemism identification deciphers the true meaning of euphemisms, such as linking "weed" (euphemism) to "marijuana" (target keyword) in illicit texts, aiding content moderation and combating underground markets. While existing methods are primarily text-based, the rise of social media highlights the need for multimodal analysis, incorporating text, images, and audio. However, the lack of multimodal datasets for euphemisms limits further research. To address this, we regard euphemisms and their corresponding target keywords as keywords and first introduce a keyword-oriented multimodal corpus of euphemisms (KOM-Euph), involving three datasets (Drug, Weapon, and Sexuality), including text, images, and speech. We further propose a keyword-oriented multimodal euphemism identification method (KOM-EI), which uses cross-modal feature alignment and dynamic fusion modules to explicitly utilize the visual and audio features of the keywords for efficient euphemism identification. Extensive experiments demonstrate that KOM-EI outperforms state-of-the-art models and large language models, and show the importance of our multimodal datasets1.

Automatic math correction aims to check students’ solutions to mathematical problems via artificial intelligence technologies. Most existing studies focus on judging the final answer at the problem level, while they ignore detailed feedback on each step in a math problem-solving process, which requires abilities of semantic understanding and reasoning. In this paper, we propose a reinforcement learning (RL)-based method to boost large language model (LLM) for step-level automatic math correction, named StepAMC. Particularly, we convert the step-level automatic math correction within the text classification task into an RL problem to enhance the reasoning capabilities of LLMs. Then, we design a space-constrained policy network to improve the stability of RL. Then, we introduce a fine-grained reward network to convert the binary human feedback into a continuous value. We conduct extensive experiments over two benchmark datasets and the results show that our model outperforms the eleven strong baselines.

Euphemisms are commonly used on social media and darknet marketplaces to evade platform regulations by masking their true meanings with innocent ones. For instance, “weed” is used instead of “marijuana” for illicit transactions. Thus, euphemism identification, i.e., mapping a given euphemism (“weed”) to its specific target word (“marijuana”), is essential for improving content moderation and combating underground markets. Existing methods employ self-supervised schemes to automatically construct labeled training datasets for euphemism identification. However, they overlook the text-text domain gap caused by the discrepancy between the constructed training data and the test data, leading to performance deterioration. In this paper, we present the text-text domain gap and explain how it forms in terms of the data distribution and the cone effect. Moreover, to bridge this gap, we introduce a feature alignment network (FA-Net), which can both align the in-domain and cross-domain features, thus mitigating the domain gap from training data to test data and improving the performance of the base models for euphemism identification. We apply this FA-Net to the base models, obtaining markedly better results, and creating a state-of-the-art model which beats the large language models.

Euphemisms are widely used on social media and darknet markets to evade supervision. For instance, "ice" serves as a euphemism for the target keyword "methamphetamine" in illicit transactions. Thus, euphemism identification which aims to map the euphemism to its secret meaning (target keyword) is a crucial task in ensuring social network security. However, this task poses significant challenges, including resource limitations due to the unavailable of annotated datasets and linguistic challenges arising from subtle differences in meaning between target keywords. Existing methods employed self-supervised schemes to automatically construct labeled training data, addressing the resource limitations. Yet, these methods rely on static embedding methods that fail to distinguish between target keywords with similar meanings. In addition, we observe that different euphemisms in similar contexts confuse the identification results. To overcome these obstacles, we propose a feature fusion and individualization (FFI) method for euphemism identification. First, we reformulate the task as a cloze task, making it more feasible. Next, we develop a feature fusion module to capture both dynamic global and static local features, enhancing discrimination between different euphemisms in similar contexts. Additionally, we employ a feature individualization module to ensure each target keyword has a unique feature representation by projecting features into their orthogonal space. As a result, FFI can effectively identify similar euphemisms that refer to target keywords with similar meanings. Experimental results demonstrate that our method outperforms state-of-the-art methods and large language models, providing robust support for its effectiveness.

Teaching by demonstration (TbD) is a powerful way to generalize the skills learned from human demonstrations to fulfill complex requirements. The mobile robot can learn new skills from interaction with human being in an intuitive way. In this paper, we propose a human-robot skill transfer system for a mobile robot that is instructed to follow a trajectory demonstrated by a human teacher wearing a motion capturing device while the Kinect sensor is recording the trajectory. With multi-modal sensor fusion, the position and velocity of the human teacher are enhanced for the correction and accuracy. A nonlinear system named a dynamic movement primitive (DMP) is modeled by the trajectories data. The Gaussian mixture model is applied for the appraisal of DMP, so as to model numerous trajectories through the teaching of a demonstration. Further, to achieve the accuracy of the trajectory tracking, a novel nonlinear model predictive control (MPC) approach is proposed for motion control. By comparison with other obstacle avoidance works, the results show improved performances in terms of the computing time and length of the trajectory.

This paper establishes an approximate optimal critic learning algorithm based on single neural network (NN) policy iteration (PI) aiming at solving for continuous-time (CT) 2-player zero-sum games (ZSGs). In fact, we have to face the problem that the errors will disturb the dynamics and in turn identifying dynamics will generate errors. In order to prevent the effect of errors, in this paper, a single NN-based online PI algorithm is developed for the CT system, which is disturbed nonlinear ZSG. With plenty of online data, the Hamilton-Jacobi-Isaacs equation can be solved without complete dynamics. Then by the least-squares method, we can obtain the NN weights. Moreover, in the process of dealing with the undisturbed system, we find the way that obtains NN weights in this paper is equal to the way that obtains the optimal solution by the Gauss-Newton method. Based on the convergence of the Gauss-Newton method, we can efficiently obtain the optimal controller for the undisturbed system by utilizing online data. After getting the controller of the undisturbed system, it is time to take disturbance into consideration, so that we design a robust control pair to overcome the disturbance. In order to demonstrate the effectiveness of this algorithm, we design a set of simulations. The results verify that we can solve the disturbed nonlinear ZSG by this algorithm.

We present a small-step operational semantics for the Python programming language. We present both a core language for Python, suitable for tools and proofs, and a translation process for converting Python source to this core. We have tested the composition of translation and evaluation of the core for conformance with the primary Python implementation, thereby giving confidence in the fidelity of the semantics. We briefly report on the engineering of these components. Finally, we examine subtle aspects of the language, identifying scope as a pervasive concern that even impacts features that might be considered orthogonal.

We suggest applying multi-carrier spread spectrum (MC-SS) techniques to fixed wireless access applications. In these applications, cross-polarization discrimination and directional subscriber antennas can be used to reduce interference in multi-cellular systems. Power control is applied on both uplink and downlink. The suggested system has a high capacity and uses a simple multi-access layer. It is robust against multipath effects and can provide service coverage not only to line-of-sight (LOS) subscribers but also to nearly LOS subscribers. Only a single carrier frequency is used in the entire network.

This paper presents a theoretical analysis of throughput and transmission efficiency for a rapidly deployable data distribution network in UNII-bands. To enhance the network throughput, all radio links are considered to exercise a collision avoidance frequency hopping technique. Such a technique is extremely attractive for efficient utilization of unlicensed bands by a group of uncoordinated radio links. Performance analysis shows that the network can maintain a very high throughput and a low outage even when it is heavily loaded.

In this paper, we propose a multi-access scheme based on OFDM/orthogonal-CDMA/sectored antenna for the last mile LMDS applications. Power control is adopted on both the uplink and downlink. Compared to TDMA/FDMA systems, the proposed configuration has a higher capacity and it offers a simpler multi-access scheme. It is more robust against multipath effects and makes it possible to increase the service coverage. Since only a single carrier frequency is used at 28 GHz, the hardware cost is lower than that for TDMA/FDMA systems.

This investigation has been motivated by a recent increased interest in orthogonal frequency division multiplexing (OFDM) CDMA systems for broadband wireless services. One of the major problems in OFDM systems is nonlinear distortion caused by high-power amplifiers. In this paper, the effect of non-perfect power amplification on OFDM-CDMA systems is examined. Simulation results show that, with an ideally linearized amplifier, the performance of OFDM-CDMA systems can be improved over an AWGN channel. In general, OFDM-CDMA systems perform better than OFDM systems with non-ideal power amplification, by using a simple code word assignment scheme.