Recent multimodal large language models achieve promising results on visual reasoning tasks, approaching human-level performance on existing STEM benchmarks. However, these capabilities degrade on complex real-world scenes. We identify two intertwined causes: chain-of-thought reasoning remains largely confined to the linguistic space, and training data rarely requires each step to be anchored in visual evidence. In this work, we introduce Visual Augmentation Primitives, which leverages high-quality instruction-guided generation models to synthesize diverse photorealistic training samples and construct coherent multimodal reasoning chains throughout the generation process, requiring models to unify visual perception with logical reasoning at every step. We further introduce multimodal chain-of-thought rubric-based reinforcement learning, which extends rubric-based reward design from the text-only domain to visual reasoning, scoring intermediate reasoning steps for logical validity, completeness, and consistency, and construct the UNREAL benchmark to systematically evaluate visual reasoning on hyperrealistic scenes across controllable difficulty levels. Experiments show that a 4B model built on this framework surpasses 8B baselines, with the approach scaling consistently to 32B across visual reasoning and VQA tasks. All models, data, and code will be publicly released.

Capturing complex contextual dependencies plays a vital role in Emotion Recognition in Conversations (ERC). Previous studies have predominantly focused on speaker-aware context modeling, overlooking the discourse structure of the conversation. In this paper, we introduce Dual Graph ATtention networks (DualGATs) to concurrently consider the complementary aspects of discourse structure and speaker-aware context, aiming for more precise ERC. Specifically, we devise a Discourse-aware GAT (DisGAT) module to incorporate discourse structural information by analyzing the discourse dependencies between utterances. Additionally, we develop a Speaker-aware GAT (SpkGAT) module to incorporate speaker-aware contextual information by considering the speaker dependencies between utterances. Furthermore, we design an interaction module that facilitates the integration of the DisGAT and SpkGAT modules, enabling the effective interchange of relevant information between the two modules. We extensively evaluate our method on four datasets, and experimental results demonstrate that our proposed DualGATs surpass state-of-the-art baselines on the majority of the datasets.

Visual dialog is challenging since it needs to answer a series of coherent questions based on understanding the visual environment. How to ground related visual objects is one of the key problems. Previous studies utilize the question and history to attend to the image and achieve satisfactory performance, while these methods are not sufficient to locate related visual objects without any guidance. The inappropriate grounding of visual objects prohibits the performance of visual dialog models. In this paper, we propose a novel approach to Learn to Ground visual objects for visual dialog, which employs a novel visual objects grounding mechanism where both prior and posterior distributions over visual objects are used to facilitate visual objects grounding. Specifically, a posterior distribution over visual objects is inferred from both context (history and questions) and answers, and it ensures the appropriate grounding of visual objects during the training process. Meanwhile, a prior distribution, which is inferred from context only, is used to approximate the posterior distribution so that appropriate visual objects can be grounding even without answers during the inference process. Experimental results on the VisDial v0.9 and v1.0 datasets demonstrate that our approach improves the previous strong models in both generative and discriminative settings by a significant margin.

Knowledge selection plays an important role in knowledge-grounded dialogue, which is a challenging task to generate more informative responses by leveraging external knowledge. Recently, latent variable models have been proposed to deal with the diversity of knowledge selection by using both prior and posterior distributions over knowledge and achieve promising performance. However, these models suffer from a huge gap between prior and posterior knowledge selection. Firstly, the prior selection module may not learn to select knowledge properly because of lacking the necessary posterior information. Secondly, latent variable models suffer from the exposure bias that dialogue generation is based on the knowledge selected from the posterior distribution at training but from the prior distribution at inference. Here, we deal with these issues on two aspects: (1) We enhance the prior selection module with the necessary posterior information obtained from the specially designed Posterior Information Prediction Module (PIPM); (2) We propose a Knowledge Distillation Based Training Strategy (KDBTS) to train the decoder with the knowledge selected from the prior distribution, removing the exposure bias of knowledge selection. Experimental results on two knowledge-grounded dialogue datasets show that both PIPM and KDBTS achieve performance improvement over the state-of-the-art latent variable model and their combination shows further improvement.