In visual question answering (VQA) context, users often pose ambiguous questions to visual language models (VLMs) due to varying expression habits. Existing research addresses such ambiguities primarily by rephrasing questions. These approaches neglect the inherently interactive nature of user interactions with VLMs, where ambiguities can be clarified through user feedback. However, research on interactive clarification faces two major challenges: (1) Benchmarks are absent to assess VLMs’ capacity for resolving ambiguities through interaction; (2) VLMs are trained to prefer answering rather than asking, preventing them from seeking clarification. To overcome these challenges, we introduce ClearVQA benchmark, which targets three common categories of ambiguity in VQA context, and encompasses various VQA scenarios. Furthermore, we propose an automated pipeline to generate ambiguity-clarification question pairs, enabling VLMs to ask reasonable clarification questions and generate more accurate and specific answers based on user feedback, as demonstrated by experimental results.

Self-critique mechanisms significantly improve the performance of language models in complex reasoning tasks by giving them the ability to correct errors, conduct induction and deduction, and switch thinking insights. However, synthetic data methods often require human-introduced errors or sampling of the model’s reasoning results from the previous moment, and the current output distribution of the model cannot be obtained, makes the data for critique and reasoning face the problem of distribution shifts. In this work, we propose an on-policy reinforcement learning framework to synchronize the reasoning and critique capabilities of language models. To alleviate reward hacking caused by outcome-based supervision, we design a deliberate reward framework for different purposes. The reward framework not only supervises the model reasoning process based on the results, but also uses Monte Carlo sampling to give appropriate rewards to the critique content according to the success rate of the model’s correction after critique. In addition, we introduce a rule-based reward function to impose penalties on the model when it generates hallucinatory critiques. When our approach is applied to the DeepSeek-Math-7B-Base and Qwen2.5-7B-Base models, model performance improves 5.40 and 3.66 points, respectively, compared to the best baseline approach. This validates the significant advantages of our method in improving model’s reasoning and self-critique capability. Code will be made available at https://github.com/rbao2018/SCOP

Simultaneous Machine Translation (SiMT) generates target translation before receiving the whole source sentence and faces a serious hallucination problem. In contrast, traditional offline machine translation (OMT) models exhibit significantly fewer hallucinations. Motivated by this disparity, we propose Knowledge Distillation for SiMT (KD-SiMT), a simple yet effective method that utilizes the OMT model to mitigate hallucinations in SiMT. Experiments on Zh→En and De→En tasks demonstrate that KD-SiMT effectively reduces hallucinations and enhances the SiMT performance. Furthermore, we systematically investigate the deficiencies in SiMT models related to serious hallucinations and the effect of KD-SiMT. Specifically, we design targeted tasks and metrics to quantitatively evaluate the components in SiMT models from the perspectives of model structure and knowledge acquisition. Our analyses reveal that inaccurate source representations and imbalanced cross-attention are more likely to occur in SiMT models when generating hallucinations, while KD-SiMT alleviates these issues. Besides, we find that KD-SiMT equips SiMT models with sufficient faithfulness knowledge in training, thus reducing hallucinations.

Simultaneous Machine Translation (SiMT) generates target outputs while receiving stream source inputs and requires a read/write policy to decide whether to wait for the next source token or generate a new target token, whose decisions form a decision path. Existing SiMT methods, which learn the policy by exploring various decision paths in training, face inherent limitations. These methods not only fail to precisely optimize the policy due to the inability to accurately assess the individual impact of each decision on SiMT performance, but also cannot sufficiently explore all potential paths because of their vast number. Besides, building decision paths requires unidirectional encoders to simulate streaming source inputs, which impairs the translation quality of SiMT models. To solve these issues, we propose Self-Modifying State Modeling (SM²), a novel training paradigm for SiMT task. Without building decision paths, SM² individually optimizes decisions at each state during training. To precisely optimize the policy, SM² introduces Self-Modifying process to independently assess and adjust decisions at each state. For sufficient exploration, SM² proposes Prefix Sampling to efficiently traverse all potential states. Moreover, SM² ensures compatibility with bidirectional encoders, thus achieving higher translation quality. Experiments show that SM² outperforms strong baselines. Furthermore, SM² allows offline machine translation models to acquire SiMT ability with fine-tuning.

Visual question answering (VQA) tasks, often performed by visual language model (VLM), face challenges with long-tail knowledge. Recent retrieval-augmented VQA (RA-VQA) systems address this by retrieving and integrating external knowledge sources. However, these systems still suffer from redundant visual information irrelevant to the question during retrieval. To address these issues, in this paper, we propose LLM-RA, a novel method leveraging the reasoning capability of a large language model (LLM) to identify key visual entities, thus minimizing the impact of irrelevant information in the query of retriever. Furthermore, key visual entities are independently encoded for multimodal joint retrieval, preventing cross-entity interference. Experimental results demonstrate that our method outperforms other strong RA-VQA systems. In two knowledge-intensive VQA benchmarks, our method achieves the new state-of-the-art performance among those with similar scale of parameters and even performs comparably to models with 1-2 orders larger parameters.