Weaknesses: 1: Limited novelty: CLEARER uses multi-scale search space that consists of three types of modules: parallel module, transition module, and fusion module. All of these modules were originally proposed in [2, 1].The authors did not cite these works when mentioning the said modules throughout the paper. It seems inconvenient, as for every new task we would have a different architecture. However, they did not provide any analysis/insights of what makes it specific for image restoration. For instance, what makes it suitable for image denoising and image deraining, OR why it would not work for any other applications such as semantic segmentation?

The paper initially received mixed ratings, but the reviewers were convinced by the rebuttal. Overall, their assessment is positive; the area chair agrees with their assessment and recommends an accept.

In Distributed Manipulator Systems (DMS), decentralization is a highly desirable property as it promotes robustness and facilitates scalability by distributing computational burden and eliminating singular points of failure. However, current DMS typically utilize a centralized approach to sensing, such as single-camera computer vision systems. This centralization poses a risk to system reliability and offers a significant limiting factor to system size. In this work, we introduce a decentralized approach for sensing and in a Distributed Manipulator Systems using Neural Cellular Automata (NCA). Demonstrating a decentralized sensing in a hardware implementation, we present a novel inductive sensor board designed for distributed sensing and evaluate its ability to estimate global object properties, such as the geometric center, through local interactions and computations. Experiments demonstrate that NCA-based sensing networks accurately estimate object position at 0.24 times the inter sensor distance. They maintain resilience under sensor faults and noise, and scale seamlessly across varying network sizes. These findings underscore the potential of local, decentralized computations to enable scalable, fault-tolerant, and noise-resilient object property estimation in DMS

Weaknesses: The paper is not very novel or significant in its contribution. It compiles two regularization methods to mitigate two long-standing problems in differentiable NAS, however, the proposed methods are not very novel. NAS-Bench is not a very well established benchmark that not many people are very familiar with. It is not fair to compare with existing work on NAS-bench, as most of them were not optimized on NAS-Bench. For instance, the DARTS work may work equally well with proper hyperparameter tuning and regularization. With the existing DARTS hyperparmeters, search on NAS-bench converges to networks with only identity/skip operation.

The reviewers generally found this paper to be a good contribution to the NAO/NAS field, with a good motivation and strong results. There were concerns on the novelty of the work, but after considering the author's response, particularly in relation to EWC, I think the work is sufficiently novel, especially given the relatively new domain. I would encourage the authors to include the clarifications and comparison to related work from the rebuttal in the main paper. The biggest issue that still lingers is the fact that NAS-Bench-201 is a very small benchmark. The most positive reviewer strongly encourages the authors to apply their technique to a larger benchmark such as NAS-Bench-1Shot1.

Conducting efficient performance estimations of neural architectures is a major challenge in neural architecture search (NAS). To reduce the architecture training costs in NAS, one-shot estimators (OSEs) amortize the architecture training costs by sharing the parameters of one "supernet" between all architectures. Recently, zero-shot estimators (ZSEs) that involve no training are proposed to further reduce the architecture evaluation cost. Despite the high efficiency of these estimators, the quality of such estimations has not been thoroughly studied. In this paper, we conduct an extensive and organized assessment of OSEs and ZSEs on five NAS benchmarks: NAS-Bench-101/201/301, and NDS ResNet/ResNeXt-A. Specifically, we employ a set of NAS-oriented criteria to study the behavior of OSEs and ZSEs, and reveal their biases and variances.

This paper proposes a novel search space for neural architecture post-processing that seeks to reduce resource consumption of trained models without sacrificing performance. Following the author feedback, all reviewers scored this paper above the threshold. They also continue to highlight crucial improvements, that I hope the authors will address for the camera-ready version.

This paper proposes a strategy for an efficient deep network architecture search (here for image classification, but a the general idea would apply for other tasks as well). The proposed strategy is will motivated and involves a data sampling stage at each step. Here, an active querying strategy can be employed and the authors evaluate their strategy with three different active sampling strategies. They show that their strategy improves over active learning (with the same active query strategies) with a fixed architecture. However, the reviewers have rightly pointed out that a comparison with other architecture search strategies would also have been in place.

Weaknesses: Similar to this work, many recent NAS approaches perform architecture search in the continuous representation space, and gradient descent (GD) is one of the most commonly used approaches for architecture search in the continuous space. However, this work only considers RL and BO as the search algorithms. I am curious to know how competitive GD is compared to RL and BO. In Figure 4, I am not sure why the right plot has more blank space compared to the left plot. An explanation is needed to help the readers understand the insights behind those plots.

Existing Neural Architecture Search (NAS) methods either encode neural architectures using discrete encodings that do not scale well, or adopt supervised learning-based methods to jointly learn architecture representations and optimize architecture search on such representations which incurs search bias. Despite the widespread use, architecture representations learned in NAS are still poorly understood. We observe that the structural properties of neural architectures are hard to preserve in the latent space if architecture representation learning and search are coupled, resulting in less effective search performance. In this work, we find empirically that pre-training architecture representations using only neural architectures without their accuracies as labels improves the downstream architecture search efficiency. To explain this finding, we visualize how unsupervised architecture representation learning better encourages neural architectures with similar connections and operators to cluster together. This helps map neural architectures with similar performance to the same regions in the latent space and makes the transition of architectures in the latent space relatively smooth, which considerably benefits diverse downstream search strategies.