In my opinion this paper is quite good, although I do not consider myself an expert on point processes, which should be taken into account when reading this review. I consider it in the top 50% of NIPS papers as a lower bound, admitting that I do not know enough to evaluated it further. I have asked the committee to ensure that at least one person who can judge the proofs sufficiently be assigned to this paper and they promised me this would be the case. I do hope if is published as I would like to try using it. Overall I think the main issue with this paper is that it would more comfortably fit in about 20 pages.

Paper revolves over the observation that in e-commerce world customers rarely purchase two items that belong to the same category (e.g. Therefore, they claim that a robust recommendation system should incorporate both utility and time utility. An additional problem that is tackled in the paper is that many e-commerce systems have no explicit negative feedback to learn from (for example one can see only what items customer purchased - positives - and no explicit negatives in form of items user did not like). I believe that the second problem they mention is not as big of a concern as advertised by authors. In absence of any explicit negative signal good replacements are long dwell-time clicks that did not end up in purchase, as well as cart additions that did not end up in the final purchase or returns. Many companies are also implementing swipe to dismiss that is useful for collecting explicit negative signal and can be applied to any e-commerce site easily.

Can't you just say "visually intuitive (given evidence such as in Figure 1)" or something like that? Updated review: My original score of 3 was a strong (and accurate) statement, voicing my opinions about what was not included in the paper, making it incomplete and inappropriate for acceptance. After reading the authors' response, I believe they agree with my assessment and will rectify this in the final version of the paper. If that is indeed corrected, then the paper is a good submission, justifying a score of 7 (in my opinion), which I stated in my original review. I will check back in on the final version and raise an objection to the AE and editors if this is these promises are not kept.

Summary: The paper proposed 4 new variants of Augmented Lagrangian methods, which called NAPALA (non-ergodic alternating proximal augmented Lagrangian algorithms) to solve non-smooth constrained convex optimization problems under different problem structure assumptions. The first algorithm only requires f and g to be convex but neither necessarily strongly convex nor smooth. Especially, its parallel variant (8) has more advantages. The second algorithm requires either f or g to be strongly convex. Its variant (12) also allows to have parallel steps.

The paper investigates the problem of inferring an agent's belief of the system dynamics of an MDP, given demonstrations of its behavior and the reward function it was optimizing. Knowledge of this internal belief can be used for Inverse Reinforcement Learning of an unknown task in the same environment. Furthermore, given the action provided by the agent, its intended action on the true dynamics can be inferred. This allows for assistive tele-operation, by applying the intended actions to the system instead of the provided ones. The proposed method models the agent using the model derived in maximum causal entropy inverse reinforcement learning.

Summary: This work propose an approach to model symmetries in deformable object categories in an unsupervised manner. This approach has been demonstrated to work for objects with bilateral symmetry (identifying symmetries in human faces using CelebA dataset, cats using cats head dataset, on cars with a synthetic car dataset), and finally for rotational symmetry on a protein structure. Pros: Overview of the problem and associated challenges. The proposed approach seems a natural way to establish dense correspondences for non-rigid objects given two views of same object category (say example in Figure-3). In my opinion, correspondences for non-rigid/deformable objects is far more important problem than symmetry (with a potential impact on numerous problems including non rigid 3D reconstruction, wide baseline disparity estimation, human analysis etc).

Future multi-spacecraft missions require robust autonomous trajectory optimization capabilities to ensure safe and efficient rendezvous operations. This capability hinges on solving non-convex optimal control problems in real time, although traditional iterative methods such as sequential convex programming impose significant computational challenges. To mitigate this burden, the Autonomous Rendezvous Transformer introduced a generative model trained to provide near-optimal initial guesses. This approach provides convergence to better local optima (e.g., fuel optimality), improves feasibility rates, and results in faster convergence speed of optimization algorithms through warm-starting. This work extends the capabilities of ART to address robust chance-constrained optimal control problems. Specifically, ART is applied to challenging rendezvous scenarios in Low Earth Orbit (LEO), ensuring fault-tolerant behavior under uncertainty. Through extensive experimentation, the proposed warm-starting strategy is shown to consistently produce high-quality reference trajectories, achieving up to 30% cost improvement and 50% reduction in infeasible cases compared to conventional methods, demonstrating robust performance across multiple state representations. Additionally, a post hoc evaluation framework is proposed to assess the quality of generated trajectories and mitigate runtime failures, marking an initial step toward the reliable deployment of AI-driven solutions in safety-critical autonomous systems such as spacecraft.

In this paper, we propose a new approach to train deep learning models using game theory concepts including Generative Adversarial Networks (GANs) and Adversarial Training (AT) where we deploy a double-oracle framework using best response oracles. GAN is essentially a two-player zero-sum game between the generator and the discriminator. The same concept can be applied to AT with attacker and classifier as players. Training these models is challenging as a pure Nash equilibrium may not exist and even finding the mixed Nash equilibrium is difficult as training algorithms for both GAN and AT have a large-scale strategy space. Extending our preliminary model DO-GAN, we propose the methods to apply the double oracle framework concept to Adversarial Neural Architecture Search (NAS for GAN) and Adversarial Training (NAS for AT) algorithms. We first generalize the players' strategies as the trained models of generator and discriminator from the best response oracles. We then compute the meta-strategies using a linear program. For scalability of the framework where multiple network models of best responses are stored in the memory, we prune the weakly-dominated players' strategies to keep the oracles from becoming intractable. Finally, we conduct experiments on MNIST, CIFAR-10 and TinyImageNet for DONAS-GAN. We also evaluate the robustness under FGSM and PGD attacks on CIFAR-10, SVHN and TinyImageNet for DONAS-AT. We show that all our variants have significant improvements in both subjective qualitative evaluation and quantitative metrics, compared with their respective base architectures.

Large language model (LLM)-based AI delegates are increasingly utilized to act on behalf of users, assisting them with a wide range of tasks through conversational interfaces. Despite their advantages, concerns arise regarding the potential risk of privacy leaks, particularly in scenarios involving social interactions. While existing research has focused on protecting privacy by limiting the access of AI delegates to sensitive user information, many social scenarios require disclosing private details to achieve desired outcomes, necessitating a balance between privacy protection and disclosure. To address this challenge, we conduct a pilot study to investigate user preferences for AI delegates across various social relations and task scenarios, and then propose a novel AI delegate system that enables privacy-conscious self-disclosure. Our user study demonstrates that the proposed AI delegate strategically protects privacy, pioneering its use in diverse and dynamic social interactions.

Research ideation involves broad exploring and deep refining ideas. Both require deep engagement with literature. Existing tools focus primarily on idea broad generation, yet offer little support for iterative specification, refinement, and evaluation needed to further develop initial ideas. To bridge this gap, we introduce IdeaSynth, a research idea development system that uses LLMs to provide literature-grounded feedback for articulating research problems, solutions, evaluations, and contributions. IdeaSynth represents these idea facets as nodes on a canvas, and allow researchers to iteratively refine them by creating and exploring variations and composing them. Our lab study (N=20) showed that participants, while using IdeaSynth, explored more alternative ideas and expanded initial ideas with more details compared to a strong LLM-based baseline. Our deployment study (N=7) demonstrated that participants effectively used IdeaSynth for real-world research projects at various ideation stages from developing initial ideas to revising framings of mature manuscripts, highlighting the possibilities to adopt IdeaSynth in researcher's workflows.