BudgetThinker: Empowering Budget-aware LLM Reasoning with Control Tokens

Recent advancements in Large Language Models (LLMs) have leveraged increased test-time computation to enhance reasoning capabilities, a strategy that, while effective, incurs significant latency and resource costs, limiting their applicability in real-world time-constrained or cost-sensitive scenarios. This paper introduces BudgetThinker, a novel framework designed to empower LLMs with budget-aware reasoning, enabling precise control over the length of their thought processes. We propose a methodology that periodically inserts special control tokens during inference to continuously inform the model of its remaining token budget. This approach is coupled with a comprehensive two-stage training pipeline, beginning with Supervised Fine-Tuning (SFT) to familiarize the model with budget constraints, followed by a curriculum-based Reinforcement Learning (RL) phase that utilizes a length-aware reward function to optimize for both accuracy and budget adherence. We demonstrate that BudgetThinker significantly surpasses strong baselines in maintaining performance across a variety of reasoning budgets on challenging mathematical benchmarks. Our method provides a scalable and effective solution for developing efficient and controllable LLM reasoning, making advanced models more practical for deployment in resource-constrained and real-time environments.

QIRL: Boosting Visual Question Answering via Optimized Question-Image Relation Learning

Existing debiasing approaches in Visual Question Answering (VQA) primarily focus on enhancing visual learning, integrating auxiliary models, or employing data augmentation strategies. However, these methods exhibit two major drawbacks. First, current debiasing techniques fail to capture the superior relation between images and texts because prevalent learning frameworks do not enable models to extract deeper correlations from highly contrasting samples. Second, they do not assess the relevance between the input question and image during inference, as no prior work has examined the degree of input relevance in debiasing studies. Motivated by these limitations, we propose a novel framework, Optimized Question-Image Relation Learning (QIRL), which employs a generation-based self-supervised learning strategy. Specifically, two modules are introduced to address the aforementioned issues. The Negative Image Generation (NIG) module automatically produces highly irrelevant question-image pairs during training to enhance correlation learning, while the Irrelevant Sample Identification (ISI) module improves model robustness by detecting and filtering irrelevant inputs, thereby reducing prediction errors. Furthermore, to validate our concept of reducing output errors through filtering unrelated question-image inputs, we propose a specialized metric to evaluate the performance of the ISI module. Notably, our approach is model-agnostic and can be integrated with various VQA models. Extensive experiments on VQA-CPv2 and VQA-v2 demonstrate the effectiveness and generalization ability of our method. Among data augmentation strategies, our approach achieves state-of-the-art results.

Feature refinement: An expression-specific feature learning and fusion method for micro-expression recognition

Micro-Expression Recognition has become challenging, as it is extremely difficult to extract the subtle facial changes of micro-expressions. Recently, several approaches proposed several expression-shared features algorithms for micro-expression recognition. However, they do not reveal the specific discriminative characteristics, which lead to sub-optimal performance. This paper proposes a novel Feature Refinement ({FR}) with expression-specific feature learning and fusion for micro-expression recognition. It aims to obtain salient and discriminative features for specific expressions and also predict expression by fusing the expression-specific features. FR consists of an expression proposal module with attention mechanism and a classification branch. First, an inception module is designed based on optical flow to obtain expression-shared features. Second, in order to extract salient and discriminative features for specific expression, expression-shared features are fed into an expression proposal module with attention factors and proposal loss. Last, in the classification branch, labels of categories are predicted by a fusion of the expression-specific features. Experiments on three publicly available databases validate the effectiveness of FR under different protocol. Results on public benchmarks demonstrate that our FR provides salient and discriminative information for micro-expression recognition. The results also show our FR achieves better or competitive performance with the existing state-of-the-art methods on micro-expression recognition.