Reasoning has substantially improved the performance of large language models (LLMs) on complicated tasks. Central to the current reasoning studies, Process Reward Models (PRMs) offer a fine-grained evaluation of intermediate reasoning steps and guide the reasoning process. However, extending PRMs to multimodal large language models (MLLMs) introduces challenges. Since multimodal reasoning covers a wider range of tasks compared to text-only scenarios, the resulting distribution shift from the training to testing sets is more severe, leading to greater generalization difficulty. Training a reliable multimodal PRM, therefore, demands large and diverse datasets to ensure sufficient coverage.

Transformer language models have demonstrated impressive generalization capabilities in natural language domains, yet we lack a fine-grained understanding of how such generalization arises. In this paper, we investigate length generalization--the ability to extrapolate from shorter to longer inputs--through the lens of \textit{task transfer}. We find that length generalization can be \textit{transferred} across related tasks. That is, training a model with a longer and related auxiliary task can lead the model to generalize to unseen and longer inputs from some other target task. We demonstrate this length generalization transfer across a diverse suite of algorithmic tasks, including arithmetic operations, string transformations, and maze navigation. Our results show that transformer models can inherit generalization capabilities from similar tasks when trained jointly. Moreover, we observe similar transfer effects in pretrained language models, suggesting that pretraining equips models with reusable computational scaffolding that facilitates extrapolation in downstream settings. Finally, we provide initial mechanistic evidence that length generalization transfer correlates with the re-use of the same attention heads between the tasks. Together, our findings deepen our understanding of how transformers generalize to out-of-distribution inputs and highlight the compositional reuse of inductive structure across tasks.

Pre-trained Vision-Language Models (VLMs) exhibit strong generalization capabilities, enabling them to recognize a wide range of objects across diverse domains without additional training. However, they often retain irrelevant information beyond the requirements of specific target downstream tasks, raising concerns about computational efficiency and potential information leakage. This has motivated growing interest in approximate unlearning, which aims to selectively remove unnecessary knowledge while preserving overall model performance. Existing approaches to approximate unlearning have primarily focused on {\em class unlearning}, where a VLM is retrained to fail to recognize specified object classes while maintaining accuracy for others. However, merely forgetting object classes is often insufficient in practical applications.

Over decades, Markov chain Monte Carlo (MCMC) methods have been widely studied, with a typical application being the quantification of posterior uncertainties in Bayesian system identification of structural dynamic models. To address the issue of excessively low sampling efficiency in generic MCMC methods when applied to specific problems, researchers developed several MCMC algorithms that integrate trainable neural networks to replace and enhance their critical components. Later, meta-learning MCMC methods emerged to reduce training time. However, they require considerable similarity between test and training tasks, while their sampling efficiency is constrained by trade-off-simplified network designs. This paper proposes the Adaptive Principal-Component (PC) Meta-learning Stochastic Gradient Hamiltonian Monte Carlo (APM-SGHMC) algorithm. It adaptively rotates coordinate axes in the parameter space to align with the PC directions of the current posterior samples, ensuring rotation-invariance of sampling performance with respect to the posterior distribution. By incorporating translation-invariance, scale-invariance, and rotation-invariance in a unified framework, APM-SGHMC enables universal samplers to acquire generalizable knowledge across diverse Bayesian system identification tasks using minimalistic tasks while eliminating the constraints imposed by network design trade-offs on sampling efficiency. Practical feasibility issues are also addressed. Two Bayesian system identification case studies demonstrate its effectiveness and universality: our method overcomes the case-by-case limitations of traditional data-driven approaches, achieving zero-shot generalization across structurally distinct models without retraining and maintaining consistent superior performance across all scenarios.

Domain generalization (DG) aims to transfer the learning task from a single or multiple source domains to unseen target domains. To extract and leverage the information which exhibits sufficient generalization ability, we propose a simple yet effective approach of Adversarial Teacher-Student Representation Learning, with the goal of deriving the domain generalizable representations via generating and exploring out-of-source data distributions. Our proposed framework advances Teacher-Student learning in an adversarial learning manner, which alternates between knowledge-distillation based representation learning and novel-domain data augmentation.

In the rapidly evolving domain of vision-based deep reinforcement learning (RL), a pivotal challenge is to achieve generalization capability to dynamic environmental changes reflected in visual observations.Our work delves into the intricacies of this problem, identifying two key issues that appear in previous approaches for visual RL generalization: (i) imbalanced saliency and (ii) observational overfitting.Imbalanced saliency is a phenomenon where an RL agent disproportionately identifies salient features across consecutive frames in a frame stack. Observational overfitting occurs when the agent focuses on certain background regions rather than task-relevant objects.To address these challenges, we present a simple yet effective framework for generalization in visual RL (SimGRL) under dynamic scene perturbations.First, to mitigate the imbalanced saliency problem, we introduce an architectural modification to the image encoder to stack frames at the feature level rather than the image level.Simultaneously, to alleviate the observational overfitting problem, we propose a novel technique called shifted random overlay augmentation, which is specifically designed to learn robust representations capable of effectively handling dynamic visual scenes.Extensive experiments demonstrate the superior generalization capability of SimGRL, achieving state-of-the-art performance in benchmarks including the DeepMind Control Suite.

Pre-trained vision-language models (VLMs) such as CLIP have shown excellent performance for zero-shot classification. Based on CLIP, recent methods design various learnable prompts to evaluate the zero-shot generalization capability on a base-to-novel setting. This setting assumes test samples are already divided into either base or novel classes, limiting its application to realistic scenarios. In this paper, we focus on a more challenging and practical setting: generalized zero-shot learning (GZSL), i.e., testing with no information about the base/novel division. To address this challenging zero-shot problem, we introduce two unique designs that enable us to classify an image without the need of knowing whether it comes from seen or unseen classes.