Machine unlearning in the domain of large language models (LLMs) has attracted great attention recently, which aims to effectively eliminate undesirable behaviors from LLMs without full retraining from scratch. In this paper, we explore the Gradient Ascent (GA) approach in LLM unlearning, which is a proactive way to decrease the prediction probability of the model on the target data in order to remove their influence. We analyze two challenges that render the process impractical: gradient explosion and catastrophic forgetting. To address these issues, we propose Multi-Objective Large Language Model Unlearning (MOLLM) algorithm. We first formulate LLM unlearning as a multi-objective optimization problem, in which the cross-entropy loss is modified to the unlearning version to overcome the gradient explosion issue. A common descent update direction is then calculated, which enables the model to forget the target data while preserving the utility of the LLM. Our empirical results verify that MoLLM outperforms the SOTA GA-based LLM unlearning methods in terms of unlearning effect and model utility preservation. The source code is available at https://github.com/zibinpan/MOLLM.

Weakly-Supervised Video Anomaly Detection (WS-VAD) primarily struggles to robustly model normal temporal dynamics and sensitively detect sparse anomalies using only video-level labels. Current methods often inadequately represent normal patterns or are too computationally costly due to complex architectures, hindering practical use. We introduce DSIB-Net, a lightweight, dual-stream Information Bottleneck (IB) based framework. Its core Dual-Output Information Bottleneck Encoder (DOIBE), driven by a Temporal Information Bottleneck Loss (TIBL), operationalizes IB. DOIBE produces a predictive representation (zpred) for dynamic modeling and a compression control vector (zcomp) for redundancy compression, with TIBL assigning them distinct functional roles. Since IB compression in DOIBE discards fine-grained appearance details, a parallel Instantaneous Mode Monitoring Path (IMP) provides crucial compensation by extracting these neglected frame-level features. An Adaptive Feature Arbitration (AFA) module then intelligently fuses these streams, dynamically weighting zpred based on IMP’s appearance cues to balance robustness and sensitivity. Experiments show DSIB-Net delivers highly competitive detection performance with minimal computational overhead, offering an efficient and practical WS-VAD solution.The code and dataset for the model are open-source and available at https://github.com/modadundun/DSIB-Net.

With extensive pretrained knowledge and high-level general capabilities, large language models (LLMs) emerge as a promising avenue to augment reinforcement learning (RL) in aspects, such as multitask learning, sample efficiency, and high-level task planning. In this survey, we provide a comprehensive review of the existing literature in LLM-enhanced RL and summarize its characteristics compared with conventional RL methods, aiming to clarify the research scope and directions for future studies. Utilizing the classical agent-environment interaction paradigm, we propose a structured taxonomy to systematically categorize LLMs' functionalities in RL, including four roles: information processor, reward designer, decision-maker, and generator. For each role, we summarize the methodologies, analyze the specific RL challenges that are mitigated and provide insights into future directions. Finally, the comparative analysis of each role, potential applications, prospective opportunities, and challenges of the LLM-enhanced RL are discussed. By proposing this taxonomy, we aim to provide a framework for researchers to effectively leverage LLMs in the RL field, potentially accelerating RL applications in complex applications, such as robotics, autonomous driving, and energy systems.

Recently, remarkable progress has been made on automated problem solving through societies of agents based on large language models (LLMs). Previous LLM-based multi-agent systems can already solve simple dialogue tasks. More complex tasks, however, face challenges through logic inconsistencies due to cascading hallucinations caused by naively chaining LLMs. Here we introduce MetaGPT, an innovative meta-programming framework incorporating efficient human workflows into LLM-based multi-agent collaborations. MetaGPT encodes Standardized Operating Procedures (SOPs) into prompt sequences for more streamlined workflows, thus allowing agents with human-like domain expertise to verify intermediate results and reduce errors. MetaGPT utilizes an assembly line paradigm to assign diverse roles to various agents, efficiently breaking down complex tasks into subtasks involving many agents working together. On collaborative software engineering benchmarks, MetaGPT generates more coherent solutions than previous chat-based multi-agent systems.