Large Language Models (LLMs) suffer from order bias, where their performance is affected by the arrangement order of input elements. This unfairness limits the model’s applications in scenarios such as in-context learning and Retrieval-Augmented Generation (RAG). Recent studies attempt to obtain optimal or suboptimal arrangements based on statistical results or using dataset-based search, but these methods increase inference overhead while leaving the model’s inherent order bias unresolved. Other studies mitigate order sensitivity through supervised fine-tuning using augmented training sets with multiple order variants, but often at the cost of accuracy, trapping the model in consistent yet incorrect hallucinations. In this paper, we propose Dual Group Advantage Optimization (DGAO), which aims to improve model accuracy and order stability simultaneously. DGAO calculates and balances intra-group relative accuracy advantage and inter-group relative stability advantage, rewarding the policy model for generating order-stable and correct outputs while penalizing order-sensitive or incorrect responses. This marks the first time reinforcement learning has been used to mitigate LLMs’ order sensitivity. We also propose two new metrics, Consistency Rate and Overconfidence Rate, to reveal the pseudo-stability of previous methods and guide more comprehensive evaluation. Extensive experiments demonstrate that DGAO achieves superior order fairness while improving performance on RAG, mathematical reasoning, and classification tasks. Our code is available at: https://anonymous.4open.science/r/DGAO-A481/

Graph-related tasks are traditionally addressed with Graph Neural Networks (GNNs) or graph transformers, but their task-specific training limits generalization. Large Language Models (LLMs) offer stronger generalization, yet encoding graphs as one-dimensional text struggles to capture multi-hop dependencies and two-dimensional topology. Vision-Language Models (VLMs) provide an alternative by visualizing graphs, but rendering large graphs in a single image causes clutter, occlusion, and distraction, hindering reasoning. We propose MuSe, a novel multi-stage graph reasoning framework based on VLMs. Instead of processing entire graphs at once, MuSe incrementally samples and visualizes task-relevant subgraphs, enabling progressive reasoning. The framework employs a two-stage training paradigm: supervised fine-tuning to acquire local sampling and reasoning skills, followed by reinforcement learning with GRPO to refine the sampling strategy and control dialog length.To support evaluation, we introduce LGVLQA, a new multimodal dataset with larger and more complex graph structures, addressing the scalability limitations of existing benchmarks. Experiments show that MuSe consistently outperforms leading LLM and VLM baselines, demonstrating improved structural understanding and reasoning ability.

Scene Graph Generation (SGG) structures visual scenes as graphs of objects and their relations. While Multimodal Large Language Models (MLLMs) have advanced end-to-end SGG, current methods are hindered by both a lack of task-specific structured reasoning and the challenges of sparse, long-tailed relation distributions, resulting in incomplete scene graphs characterized by low recall and biased predictions. To address these issues, we introduce SGG-R ³, a structured reasoning framework that integrates task-specific Chain-of-Thought (CoT)-guided supervised fine-tuning (SFT) and reinforcement learning (RL) with group sequence policy optimization (GSPO), designed to engage in three sequential stages to achieve end-to-end unbiased scene graph generation. During the SFT phase, we propose a relation augmentation strategy by leveraging an MLLM and refined via embedding similarity filtering to alleviate relation sparsity. Subsequently, a stage-aligned reward scheme optimizes the procedural reasoning during RL. Specifically, we propose a novel dual-granularity reward which integrates fine-grained and coarse-grained relation rewards, simultaneously mitigating the long-tail issue via frequency-based adaptive weighting of predicates and improving relation coverage through semantic clustering. Experiments on two benchmarks show that SGG-R ³ achieves superior performance compared to existing methods, demonstrating the effectiveness and generalization of the framework.

High-stakes domains such as finance, law, and biomedicine demand both accurate results and rigorous reasoning. Current reinforcement learning paradigms primarily rely on outcome-based rewards, often overlooking latent logical fallacies in intermediate steps. Leveraging the cognitive asymmetry where falsifying local errors is more efficient than generating global correctness, we propose RADO (Reasoning Audit-Driven Optimization). RADO introduces a specialized audit model augmented with external tools to identify local logical ruptures and calibrate reward signals. By integrating Direct Preference Optimization (DPO) with Group Relative Policy Optimization (GRPO), our framework enables explicit supervision over reasoning paths. Experimental results demonstrate that RADO consistently improves final accuracy while significantly enhancing logical rigor in high-stakes domains.

Parameter-efficient fine-tuning (PEFT) methods, particularly Low-Rank Adaptation (LoRA), offer an efficient way to adapt large language models with reduced computational costs. However, their performance is limited by the small number of trainable parameters. Recent work combines LoRA with the Mixture-of-Experts (MoE), i.e., LoRA-MoE, to enhance capacity, but two limitations remain in hindering the full exploitation of its potential: 1) the influence of downstream tasks when assigning expert numbers, and 2) the uniform rank assignment across all LoRA experts, which restricts representational diversity.To mitigate these gaps, we propose GuiLoMo, a fine-grained layer-wise expert numbers and ranks allocation strategy with GuidedSelection Vectors (GSVs). GSVs are learned via a prior bilevel optimization process to capture both model- and task-specific needs, and are then used to allocate optimal expert numbers and ranks.Experiments on three backbone models across diverse benchmarks show that GuiLoMo consistently achieves superior or comparable performance to all baselines. Further analysis offers key insights into how expert numbers and ranks vary across layers and tasks, highlighting the benefits of adaptive expert configuration. Our code is available at https://anonymous.4open.science/r/GuiLoMo-034.

Large Language Models (LLMs) are widely applied to downstream domains. However, current LLMs for high-stakes domain tasks, such as financial investment and legal QA, typically generate brief answers without reasoning processes and explanations. This limits users’ confidence in making decisions based on their responses. While original CoT shows promise, it lacks self-correction mechanisms during reasoning. This work introduces Domaino1s, which enhances LLMs’ reasoning capabilities on domain tasks through supervised fine-tuning and tree search. We construct CoT-stock-2k and CoT-legal-2k datasets for fine-tuning models that activate domain-specific reasoning steps based on their judgment. Additionally, we propose Selective Tree Exploration to spontaneously explore solution spaces and sample optimal reasoning paths to improve performance. We also introduce PROOF-Score, a new metric for evaluating domain models’ explainability, complementing traditional accuracy metrics with richer assessment dimensions. Extensive experiments on stock investment recommendation and legal reasoning QA tasks demonstrate Domaino1s’s leading performance and explainability. Our code is available at https://anonymous.4open.science/r/Domaino1s-006F/.

Retrieval-augmented language models (RALMs) have recently shown great potential in mitigating the limitations of implicit knowledge in LLMs, such as untimely updating of the latest expertise and unreliable retention of long-tail knowledge. However, since the external knowledge base, as well as the retriever, can not guarantee reliability, potentially leading to the knowledge retrieved not being helpful or even misleading for LLM generation. In this paper, we introduce Supportiveness-based Knowledge Rewriting (SKR), a robust and pluggable knowledge rewriter inherently optimized for LLM generation. Specifically, we introduce the novel concept of “supportiveness”—which represents how effectively a knowledge piece facilitates downstream tasks. Based on supportiveness, we first design a training data curation strategy for our rewriter model, effectively identifying and filtering out poor or irrelevant rewrites to improve data efficacy. We then introduce the direct preference optimization (DPO) algorithm to align the generated rewrites to optimal supportiveness, guiding the rewriter model to summarize augmented content that better improves the final response. Comprehensive evaluations across six popular knowledge-intensive tasks and four LLMs have demonstrated the effectiveness and superiority of SKR. With only 7B parameters, SKR has shown better knowledge rewriting capability over GPT-4.

Contrastive learning has recently shown great potential to improve text-based knowledge graph completion (KGC). In this paper, we propose to learn a more semantically structured entity representation space in text-based KGC via hard negatives mining. Specifically, we novelly leverage a sequence-to-sequence architecture to generate high-quality hard negatives. These negatives are sampled from the same decoding distributions as the anchor (or correct entity), inherently being semantically close to the anchor and thus enjoying good hardness. A self-information-enhanced contrasting strategy is further incorporated into the Seq2Seq generator to systematically diversify the produced negatives. Extensive experiments on three KGC benchmarks demonstrate the sound hardness and diversity of our generated negatives and the resulting performance superiority on KGC.

Many recent sentence-level event detection efforts focus on enriching sentence semantics, e.g., via multi-task or prompt-based learning. Despite the promising performance, these methods commonly depend on label-extensive manual annotations or require domain expertise to design sophisticated templates and rules. This paper proposes a new paradigm, named dialogue-based explanation, to enhance sentence semantics for event detection. By saying dialogue-based explanation of an event, we mean explaining it through a consistent information-intensive dialogue, with the original event description as the start utterance. We propose three simple dialogue generation methods, whose outputs are then fed into a hybrid attention mechanism to characterize the complementary event semantics. Extensive experimental results on two event detection datasets verify the effectiveness of our method and suggest promising research opportunities in the dialogue-based explanation paradigm.

Event detection aims to detect events from the text by identifying and classifying event triggers (the most representative words). Most of the existing works rely heavily on complex downstream networks and require sufficient training data. Thus, those models may be structurally redundant and perform poorly when data is scarce. Prompt-based models are easy to build and are promising for few-shot tasks. However, current prompt-based methods may suffer from low precision because they have not introduced event-related semantic knowledge (e.g., part of speech, semantic correlation, etc.). To address these problems, this paper proposes a Knowledge-injected Prompt Tuning (KiPT) model. Specifically, the event detection task is formulated into a condition generation task. Then, knowledge-injected prompts are constructed using external knowledge bases, and a prompt tuning strategy is leveraged to optimize the prompts. Extensive experiments indicate that KiPT outperforms strong baselines, especially in few-shot scenarios.

Answering natural language questions on knowledge graphs (KGQA) remains a great challenge in terms of understanding complex questions via multi-hop reasoning. Previous efforts usually exploit large-scale entity-related text corpus or knowledge graph (KG) embeddings as auxiliary information to facilitate answer selection. However, the rich semantics implied in off-the-shelf relation paths between entities is far from well explored. This paper proposes improving multi-hop KGQA by exploiting relation paths’ hybrid semantics. Specifically, we integrate explicit textual information and implicit KG structural features of relation paths based on a novel rotate-and-scale entity link prediction framework. Extensive experiments on three existing KGQA datasets demonstrate the superiority of our method, especially in multi-hop scenarios. Further investigation confirms our method’s systematical coordination between questions and relation paths to identify answer entities.

Natural Language Sentence Matching (NLSM) serves as the core of many natural language processing tasks. 1) Most previous work develops a single specific neural model for NLSM tasks. 2) There is no previous work considering adversarial attack to improve the performance of NLSM tasks. 3) Adversarial attack is usually used to generate adversarial samples that can fool neural models. In this paper, we first find a phenomenon that different categories of samples have different vulnerabilities. Vulnerability is the difficulty degree in changing the label of a sample. Considering the phenomenon, we propose a general two-stage training framework to enhance neural models with Vulnerability via Adversarial Attack (VAA). We design criteria to measure the vulnerability which is obtained by adversarial attack. VAA framework can be adapted to various neural models by incorporating the vulnerability. In addition, we prove a theorem and four corollaries to explain the factors influencing vulnerability effectiveness. Experimental results show that VAA significantly improves the performance of neural models on NLSM datasets. The results are also consistent with the theorem and corollaries. The code is released on https://github.com/rzhangpku/VAA.