Vision-Language Models seamlessly discriminate among arbitrary semantic categories, yet they still suffer from poor generalization when presented with challenging examples. For this reason, Episodic Test-Time Adaptation (TTA) strategies have recently emerged as powerful techniques to adapt VLMs in the presence of a single unlabeled image. The recent literature on TTA is dominated by the paradigm of prompt tuning by Marginal Entropy Minimization, which, relying on online backpropagation, inevitably slows down inference while increasing memory. In this work, we theoretically investigate the properties of this approach and unveil that a surprisingly strong TTA method lies dormant and hidden within it.

Reconstructing past events requires reasoning across long time horizons. To figure out what happened, humans draw on prior knowledge about the world and human behavior and integrate insights from various sources of evidence including visual, language, and auditory cues. We introduce MARPLE, a benchmark for evaluating long-horizon inference capabilities using multi-modal evidence. Our benchmark features agents interacting with simulated households, supporting vision, language, and auditory stimuli, as well as procedurally generated environments and agent behaviors. Inspired by classic "whodunit" stories, we ask AI models and human participants to infer which agent caused a change in the environment based on a step-by-step replay of what actually happened.

Although it is often argued that this representational format is useful in learning to solve many real-world down-stream tasks, there is little empirical evidence that supports this claim. In this paper, we conduct a large-scale study that investigates whether disentangled representations are more suitable for abstract reasoning tasks. Using two new tasks similar to Raven's Progressive Matrices, we evaluate the usefulness of the representations learned by 360 state-of-the-art unsupervised disentanglement models. Based on these representations, we train 3600 abstract reasoning models and observe that disentangled representations do in fact lead to better down-stream performance. In particular, they enable quicker learning using fewer samples.

Traditional unlearnable strategies have been proposed to prevent unauthorized users from training on the 2D image data. With more 3D point cloud data containing sensitivity information, unauthorized usage of this new type data has also become a serious concern. To address this, we propose the first integral unlearnable framework for 3D point clouds including two processes: (i) we propose an unlearnable data protection scheme, involving a class-wise setting established by a categoryadaptive allocation strategy and multi-transformations assigned to samples; (ii) we propose a data restoration scheme that utilizes class-wise inverse matrix transformation, thus enabling authorized-only training for unlearnable data. This restoration process is a practical issue overlooked in most existing unlearnable literature, i.e., even authorized users struggle to gain knowledge from 3D unlearnable data. Both theoretical and empirical results (including 6 datasets, 16 models, and 2 tasks) demonstrate the effectiveness of our proposed unlearnable framework. Our code is available at https://github.com/CGCL-codes/UnlearnablePC.

To verify the effectiveness of PABEE on Computer Vision, we follow the experimental settings in Shallow-Deep [5], we conduct experiments on two image classification datasets, CIFAR-10 and CIFAR-100 [55]. We use ResNet-56 [10] as the backbone and compare PABEE with BranchyNet [26] and Shallow-Deep [5]. After every two convolutional layers, an internal classifier is added. We set the batch size to 128 and use SGD optimizer with learning rate of 0.1. Table 6: Experimental results (median of 5 runs) of ResNet based models on CIFAR-10 and CIFAR-100 datasets.

We would like to thank the reviewers for their comments, and take the opportunity to answer their questions below. R1: (1) We thank the reviewer for the relevant [Amari et al., 2000] reference, which we will cite and discuss. Similarly, [Amari et al., 2000] considers single-layer networks Further, we examined the method's accuracy relative to recent techniques, and extended it to R2: (1) Regarding λ, we selected a small value so that the Hessian is not dominated by the dampening. Please see Appendix S5 for ablation studies. V1 results increase from 63.87 64.59, etc.).

Reviewer 1: Thank you for the insightful analysis and acknowledgement of our effort. We will split the table to improve readability and test data. The model is clearly expressive enough as training and test accuracy are near-perfect. Reviewer 3: 1. XMC datasets have been well-researched and improvements "by couple of % points" are significant. In Sec 2.2 and Theorem 2.1, we rigorously showed the existence of a perfect accuracy For example, compare P@k and PSP@k of PfastreXML and FastXML in Table 3 and Table 4.

Mechanistic interpretability methods aim to identify the algorithm a neural network implements, but it is difficult to validate such methods when the true algorithm is unknown.

Mechanistic interpretability methods aim to identify the algorithm a neural network implements, but it is difficult to validate such methods when the true algorithm is unknown.

Autonomous agents that control user interfaces to accomplish human tasks are emerging. Leveraging LLMs to power such agents has been of special interest, but unless fine-tuned on human-collected task demonstrations, performance is still relatively low. In this work we study whether fine-tuning alone is a viable approach for building real-world UI control agents.