Historically, the focus in entanglement distillation has predominantly been on the distillable entanglement, and the framework assumes complete knowledge of the initial state. In this paper, we study the reliability function of entanglement distillation, which specifies the optimal exponent of the decay of the distillation error when the distillation rate is below the distillable entanglement. Furthermore, to capture greater operational significance, we extend the framework from the standard setting of known states to a black-box setting, where distillation is performed from a set of possible states. We establish an exact finite blocklength result connecting to composite correlated hypothesis testing without any redundant correction terms. Based on this, the reliability function of entanglement distillation is characterized by the regularized quantum Hoeffding divergence. In the special case of a pure initial state, our result reduces to the error exponent for entanglement concentration derived by Hayashi et al. in 2003. Given full prior knowledge of the state, we construct a concrete optimal distillation protocol. Additionally, we analyze the strong converse exponent of entanglement distillation. While all the above results assume the free operations to be non-entangling, we also investigate other free operation classes, including PPT-preserving, dually non-entangling, and dually PPT-preserving operations.

Large-scale pre-trained models have achieved remarkable success in various computer vision tasks. A standard approach to leverage these models is to fine-tune all model parameters for downstream tasks, which poses challenges in terms of computational and storage costs. Recently, inspired by Natural Language Processing (NLP), parameter-efficient transfer learning has been successfully applied to vision tasks. However, most existing techniques primarily focus on single-task adaptation, and despite limited research on multi-task adaptation, these methods often exhibit suboptimal training/inference efficiency. In this paper, we first propose an once-for-all Vision Multi-Task Adapter (VMT-Adapter), which strikes approximately O(1) training and inference efficiency w.r.t task number. Concretely, VMT-Adapter shares the knowledge from multiple tasks to enhance cross-task interaction while preserves task-specific knowledge via independent knowledge extraction modules. Notably, since task-specific modules require few parameters, VMT-Adapter can handle an arbitrary number of tasks with a negligible increase of trainable parameters. We also propose VMT-Adapter-Lite, which further reduces the trainable parameters by learning shared parameters between down- and up-projections. Extensive experiments on four dense scene understanding tasks demonstrate the superiority of VMT-Adapter(-Lite), achieving a 3.96% (1.34%) relative improvement compared to single-task full fine-tuning, while utilizing merely ～1% (0.36%) trainable parameters of the pre-trained model.

We aim at finetuning a vision-language model without hurting its out-of-distribution (OOD) generalization. We address two types of OOD generalization, i.e., i) domain shift such as natural to sketch images, and ii) zero-shot capability to recognize the category that was not contained in the finetune data. Arguably, the diminished OOD generalization after finetuning stems from the excessively simplified finetuning target, which only provides the class information, such as “a photo of a [CLASS]”. This is distinct from the process in that CLIP was pretrained, where there is abundant text supervision with rich semantic information. Therefore, we propose to compensate for the finetune process using auxiliary supervision with rich semantic information, which acts as anchors to preserve the OOD generalization. Specifically, two types of anchors are elaborated in our method, including i) text-compensated anchor which uses the images from the finetune set but enriches the text supervision from a pretrained captioner, ii) image-text-pair anchor which is retrieved from the dataset similar to pretraining data of CLIP according to the downstream task, associating with the original CLIP text with rich semantics. Those anchors are utilized as auxiliary semantic information to maintain the original feature space of CLIP, thereby preserving the OOD generalization capabilities. Comprehensive experiments demonstrate that our method achieves in-distribution performance akin to conventional finetuning while attaining new state-of-the-art results on domain shift and zero-shot learning benchmarks.

The task of Human-Object Interaction (HOI) detection is to detect humans and their interactions with surrounding objects, where transformer-based methods show dominant advances currently. However, these methods ignore the relationship among humans, objects, and interactions: 1) human features are more contributive than object ones to interaction prediction; 2) interactive information disturbs the detection of objects but helps human detection. In this article, we propose aHuman and Object Disentangling Network(HODN) to model the HOI relationships explicitly, where humans and objects are first detected by two disentangling decoders independently and then processed by an interaction decoder. Considering that human features are more contributive to interaction, we propose aHuman-Guide Linkingmethod to make sure the interaction decoder focuses on the human-centric regions with human features as the positional embeddings. To handle the opposite influences of interactions on humans and objects, we propose aStop-Gradient Mechanismto stop interaction gradients from optimizing the object detection but to allow them to optimize the human detection. Our proposed method achieves competitive performance on both the V-COCO and the HICO-Det datasets. It can be combined with existing methods easily for state-of-the-art results.

Multi-modality pre-trained models (PTMs) have considerably boosted the performance on a broad range of computer vision topics. Still, they have not been explored purposefully in open set recognition (OSR) scenarios when applying PTMs to downstream recognition tasks. Directly fine/prompt tuning PTMs on closed-set classification tasks will inevitably suffer from data bias and always learn more or less target class-irrelevant cooccurring contextual information, which leads to over-confident predictions on unknown samples. In this paper, we propose a simple yet effective approach, termed Anti-Associative Prompt Tuning(A2Pt), toward learning compact and accurate class-related representation with few class-irrelevant associations from context using multi-modal priors. Specifically, a cross-modal guided activation module is adopted to refine the class-aware representation and suppress the associations from co-occurring contexts by involving text-modal information. We further design an anti-association calibration module to obtain compact class-aware and class-irrelevant representations, respectively, by introducing two additional object functions. Extensive experiments on publicly available benchmarks, including CIFAR series, Tiny-ImageNet, and ImageNet-21K-P, show that the proposed(A2Pt)achieves substantial and consistent performance gains compared with both SOTA OSR and PTM prompt tuning approaches.

Weakly supervised object localization (WSOL) remains an open problem given the deficiency of finding object extent information using a classification network. Although prior works struggled to localize objects through various spatial regularization strategies, we argue that how to extract object structural information from the trained classification network is neglected. In this paper, we propose a two-stage approach, termed structure-preserving activation (SPA), toward fully leveraging the structure information incorporated in convolutional features for WSOL. First, a restricted activation module (RAM) is designed to alleviate the structure-missing issue caused by the classification network on the basis of the observation that the unbounded classification map and global average pooling layer drive the network to focus only on object parts. Second, we designed a post-process approach, termed self-correlation map generating (SCG) module to obtain structure-preserving localization maps on the basis of the activation maps acquired from the first stage. Specifically, we utilize the high-order self-correlation (HSC) to extract the inherent structural information retained in the learned model and then aggregate HSC of multiple points for precise object localization. Extensive experiments on two publicly available benchmarks including CUB-2002011 and ILSVRC show that the proposed SPA achieves substantial and consistent performance gains compared with baseline approaches. Code and models are available at github.com/Panxjia/SPA CVPR2021.

We study the problem of weakly supervised grounded image captioning. That is, given an image, the goal is to automatically generate a sentence describing the context of the image with each noun word grounded to the corresponding region in the image. This task is challenging due to the lack of explicit fine-grained region word alignments as supervision. Previous weakly supervised methods mainly explore various kinds of regularization schemes to improve attention accuracy. However, their performances are still far from the fully supervised ones. One main issue that has been ignored is that the attention for generating visually groundable words may only focus on the most discriminate parts and can not cover the whole object. To this end, we propose a simple yet effective method to alleviate the issue, termed as partial grounding problem in our paper. Specifically, we design a distributed attention mechanism to enforce the network to aggregate information from multiple spatially different regions with consistent semantics while generating the words. Therefore, the union of the focused region proposals should form a visual region that encloses the object of interest completely. Extensive experiments have demonstrated the superiority of our proposed method compared with the state-of-the-arts.