Recent advancements in Large Language Models have increasingly demonstrated their potential for event reasoning. However, LLMs still struggle with this task due to inadequate modeling of event structures. Although introducing schema knowledge has been shown to improve event reasoning performance, existing methods rely on predefined schema library, compromising their scalability and lightweight deployment. To address these challenges, we propose SGER, a plug-and-play Schema-Guided Event Reasoning framework. In the schema extraction stage, the model maps event descriptions with diverse surface forms to potential semantic structure representations, achieving an abstract transformation from instances to schemas. The schema prediction stage captures the potential associations between historical event schemas to make forward-looking inferences about possible future event schemas. In the event reasoning stage, we integrate historical events and predicted schemas into prompts to guide LLMs in generating specific, contextually consistent predicted events. Experimental evaluations demonstrate that our framework significantly improves event reasoning performance of LLMs.

Wenqing Hou, Hongkui Tu, Ye Wang, Yue Zhang, Yuying Liu, Dong Zhu, Liqun Gao, Bin Zhou. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

When fact-checking methods based on large language models (LLMs) use external evidence to validate claims, knowledge conflicts often arise. These conflicts typically stem from inconsistencies between the external evidence and LLMs' internal pre-existing knowledge. Such an inconsistency could lead LLMs to draw incorrect answers when validating claims, especially when they are overly confident in their internal incorrect knowledge. Previous works on LLM-based fact-checking have overlooked this issue. This paper, for the first time, proposes a framework (namely KnowFC) to navigate this issue. Our key insight is dividing and adaptively utilizing the knowledge that LLMs know and do not know, thereby avoiding conflicts while enhancing the correctness and efficiency of fact-checking. Specifically, in KnowFC, we propose an adaptive retrieval method, where we train an LLM using a reinforcement learning algorithm coupled with the Dunning-Kruger effect-inspired reward mechanism to identify its knowledge boundaries through confidence calibration, thereby realizing adaptive evidence retrieval. Besides, we propose a reliable and debiased fact verification method, where we organize and construct reasoning graphs using retrieved evidence to verify claims, followed by a causal intervention method using causal mediation analysis to mitigate internal knowledge interference. Experimental results on both FEVEROUS and AVeriTeC datasets show that our method outperforms baseline methods in terms of accuracy and F1 score, while also improving fact-checking efficiency.

Penetration testing (pentesting) assesses cybersecurity through controlled, authorized attacks, but traditional manual methods demand considerable human and time resources. Reinforcement learning (RL), with its agent-environment interaction paradigm, offers a promising approach for autonomous pentesting. Despite remarkable advancements in this field, there is a lack of comprehensive reviews and perspectives on RL-based autonomous pentesting. To address this gap, this paper presents a systematic review of RL-based autonomous pentesting research. We outline the key challenges faced when applying RL in autonomous pentesting and categorize the existing literature into two main areas: attack path planning and autonomous pentesting frameworks, based on the research objectives and hypotheses. Additionally, we offer an in-depth analysis of the latest advancements and limitations in this field, while proposing a perspective on future research directions in the field of RL-based autonomous pentesting. We hope that our work will provide valuable insights for researchers, contributing to the advancement of autonomous pentesting and its practical application in the complex and diverse scenarios of the real world.

Large language models (LLMs) demonstrate significant generative capabilities but often face ethical alignment and robustness challenges. Conventional alignment methods rely on extensive human-annotated data and require retraining, leading to high computational costs and resource demands. Therefore, we propose a novel approach, Decoding Probability Correction (DPC), that aligns frozen LLMs without additional training or annotated data. DPC dynamically adjusts the probability distribution during inference, ensuring the generated content aligns with human values in real-time. Additionally, DPC incorporates a discriminator-based backtracking mechanism, further enhancing content safety by re-evaluating and refining generation choices. Experimental results on datasets such as the HH and AdvBench show that DPC significantly reduces harmful outputs while maintaining high levels of informativeness and helpfulness. The proposed method offers a cost-effective and efficient solution for enhancing the ethical alignment of LLMs in real-world applications.

With the increasing number of vulnerabilities exposed on the Internet, autonomous penetration testing (pentesting) has emerged as a promising research area. Reinforcement learning (RL) is a natural fit for studying this topic. However, two key challenges limit the applicability of RL-based autonomous pentesting in real-world scenarios: the training environment dilemma—training agents in simulated environments is sample-efficient while ensuring that their realism remains challenging; poor generalization ability—agents’ policies often perform poorly when transferred to unseen scenarios, with even slight changes potentially causing a significant generalization gap. To address both challenges, we propose a generalizable autonomous pentesting framework termed GAP, which aims to achieve efficient policy training in realistic environments and train generalizable agents capable of drawing inferences about other cases from one instance. GAP introduces a real-to-sim-to-real pipeline that enables end-to-end policy learning in unknown real environments while constructing realistic simulations and improves agents’ generalization ability by leveraging domain randomization and meta-RL learning. We are among the first to apply domain randomization in autonomous pentesting and propose a large language model-powered domain randomization method for synthetic environment generation. We further apply meta-RL to improve agents’ generalization ability in unseen environments by leveraging synthetic environments. Combining the two methods effectively bridges the generalization gap and improves agents’ policy adaptation performance. Simulations are conducted on various vulnerable virtual machines, with results showing that GAP can enable policy learning in various realistic environments, achieve zero-shot policy transfer in similar environments, and achieve rapid policy adaptation in dissimilar environments.

Penetration testing (pentesting) assesses cybersecurity through simulated attacks, while the conventional manual-based method is costly, time-consuming, and personnel-constrained. Reinforcement learning (RL) provides an agent-environment interaction learning paradigm, making it a promising way for autonomous pentesting. However, agents’ scalability in large action spaces and policy transferability across scenarios limit the applicability of RL-based autonomous pentesting. To address these challenges, we present a novel autonomous pentesting framework based on reinforcement learning (namely APRIL) to train agents that are scalable and transferable in large action spaces. In APRIL, we construct realistic, bounded, host-level state space via embedding techniques to avoid the complexities of dealing with unbounded network-level information. We employ semantic correlations between pentesting actions as prior knowledge to represent discrete action space into a continuous and semantically meaningful embedding space. Agents are then trained to reason over actions within the action embedding space, where two key methods are applied: an upper-confidence bound-based action refinement method to encourage efficient exploration, and a distance-aware loss to improve learning efficiency and generalization performance. We conduct experiments in simulated scenarios constructed based on virtualized vulnerable environments. The results demonstrate APRIL's scalability in large action spaces and its ability to facilitate policy transfer across diverse scenarios.

Counterspeech is an effective way to combat online hate speech. Considering the multifaceted nature of online hate speech, counterspeech with varying intents (e.g., denouncing or empathy) has significant potential to mitigate hate speech effectively. Recently, controlled approaches based on large language models (LLMs) have been explored to generate intent-specific counterspeech. Due to the lack of attention to intent-specific information by LLMs during the decoding process, those methods cater more to the semantic information rather than matching with the desired intents. Further, there are still limitations in quantitatively evaluating the effectiveness of counterspeech with different intents in mitigating hate speech. In this paper, to address the above issues, we propose DART, an LLMs-based DuAl-discRiminaTor guided framework for counterspeech generation. We employ an intent-aware discriminator and hate-mitigating discriminator to jointly guide the decoding preferences of LLMs, which facilitates the model towards generating counterspeech catering to specific intent and hate mitigation. We apply a maximum-margin relative objective for training discriminators. This objective leverages the distance between counterspeech aligned with the desired target (such as specific intent or effectiveness in hate mitigation) and undesired as an effective learning signal. Extensive experiments show that DART achieves excellent performances in matching the desired intent and mitigating hate.