Generative reward models (GRMs) have emerged as a promising approach for aligning Large Language Models (LLMs) with human preferences by offering greater representational capacity and flexibility than traditional scalar reward models. However, GRMs face two major challenges: reliance on costly human-annotated data restricts scalability, and self-training approaches often suffer from instability and vulnerability to reward hacking. To address these issues, we propose ConsistRM, a self-training framework that enables effective and stable GRM training without human annotations. ConsistRM incorporates the Consistency-Aware Answer Reward, which produces reliable pseudo-labels with temporal consistency, thereby providing more stable model optimization. Moreover, the Consistency-Aware Critique Reward is introduced to assess semantic consistency across multiple critiques and allocates fine-grained and differentiated rewards. Experiments on five benchmark datasets across four base models demonstrate that ConsistRM outperforms vanilla Reinforcement Fine-Tuning (RFT) by an average of 1.5%. Further analysis shows that ConsistRM enhances output consistency and mitigates position bias caused by input order, highlighting the effectiveness of consistency-aware rewards in improving GRMs.Our implementation is available at https://github.com/yuliangCarmelo/ConsistRM.

Reward Models (RMs) are critical components in the Reinforcement Learning from Human Feedback (RLHF) pipeline, directly determining the alignment quality of Large Language Models (LLMs). Recently, Generative Reward Models (GRMs) have emerged as a superior paradigm, offering higher interpretability and stronger generalization than traditional scalar RMs. However, existing methods for GRMs focus primarily on outcome-level supervision, neglecting analytical process quality, which constrains their potential. To address this, we propose ReflectRM, a novel GRM that leverages self-reflection to assess analytical quality and enhance preference modeling. ReflectRM is trained under a unified generative framework for joint modeling of response preference and analysis preference. During inference, we use its self-reflection capability to identify the most reliable analysis, from which the final preference prediction is derived. Experiments across four benchmarks show that ReflectRM consistently improves performance, achieving an average accuracy gain of +3.7 on Qwen3-4B. Further experiments confirm that response preference and analysis preference are mutually reinforcing. Notably, ReflectRM substantially mitigates positional bias, yielding +10.2 improvement compared with leading GRMs and establishing itself as a more stable evaluator. Our code is available at https://github.com/yuliangCarmelo/ReflectRM.

With the increasing availability of multimodal content on social media, consisting primarily of text and images, multimodal named entity recognition (MNER) has gained a wide-spread attention. A fundamental challenge of MNER lies in effectively aligning different modalities. However, the majority of current approaches rely on word-based sequence labeling framework and align the image and text at inconsistent semantic levels (whole image-words or regions-words). This misalignment may lead to inferior entity recognition performance. To address this issue, we propose an effective span-based method, named SMNER, which achieves a more consistent multimodal alignment from the perspectives of information-theoretic and cross-modal interaction, respectively. Specifically, we first introduce a cross-modal information bottleneck module for the global-level multimodal alignment (whole image-whole text). This module aims to encourage the semantic distribution of the image to be closer to the semantic distribution of the text, which can enable the filtering out of visual noise. Next, we introduce a cross-modal attention module for the local-level multimodal alignment (regions-spans), which captures the correlations between regions in the image and spans in the text, enabling a more precise alignment of the two modalities. Extensive ex- periments conducted on two benchmark datasets demonstrate that SMNER outperforms the state-of-the-art baselines.

Fact verification is an essential tool to mitigate the spread of false information online, which has gained a widespread attention recently. However, a fact verification in the question-answering dialogue is still underexplored. In this paper, we propose a neural network based approach called question-answering dialogue based fact verification with mixture of experts (QaDialMoE). It exploits questions and evidence effectively in the verification process and can significantly improve the performance of fact verification. Specifically, we exploit the mixture of experts to focus on various interactions among responses, questions and evidence. A manager with an attention guidance module is implemented to guide the training of experts and assign a reasonable attention score to each expert. A prompt module is developed to generate synthetic questions that make our approach more generalizable. Finally, we evaluate the QaDialMoE and conduct a comparative study on three benchmark datasets. The experimental results demonstrate that our QaDialMoE outperforms previous approaches by a large margin and achieves new state-of-the-art results on all benchmarks. This includes the accuracy improvements on the HEALTHVER as 84.26%, the FAVIQ A dev set as 78.7%, the FAVIQ R dev set as 86.1%, test set as 86.0%, and the COLLOQUIAL as 89.5%. To our best knowledge, this is the first work to investigate a question-answering dialogue based fact verification, and achieves new state-of-the-art results on various benchmark datasets.