Subset (2)

questions · 869 rows

Split (1)

train · 869 rows

query_id int64	question string	answer unknown	hallucination_set list	question_type string	evidence list	metadata list
1	Please summarize the paper Quantifying and Mitigating the Impact of Label Errors on Model Disparity Metrics	[ "label error", "group-based disparity metrics", "influence function", "expected calibration error (ECE)", "False Positive Rate (FPR)", "False Negative Rate (FNR)", "Error Rate (ER)", "training input’s label", "sensitivity analysis", "dataset purification", "noise-robust algorithms", "logistic ...	[]	Single-Sum	[ 1 ]	[ { "reviews": [ { "summary": "This paper studies the effect of label error on the model’s disparity metrics (e.g., calibration, FPR, FNR) on both the training and test set. Empirically, the authors have found that label errors have a larger influence on minority groups than on majority groups. To m...
2	Please summarize the paper Suppression helps: Lateral Inhibition-inspired Convolutional Neural Network for Image Classification	[ "Lateral inhibition mechanism", "Deep convolutional networks", "Image classification", "Gaussian low-pass filter", "Learnable channel weight", "AlexNet", "ResNet", "Center-surround pattern", "Neurobiological effect", "Contrast enhancement", "Flexible alternative", "Depthwise convolution", "N...	[]	Single-Sum	[ 2 ]	[ { "reviews": [ { "summary": "The authors propose to add a biologically inspired lateral inhibition mechanism into deep convolutional networks for image recognition. When incorporated into AlexNets and ResNets, LI seems to improve performance on ImageNet classification without increasing trainable ...
3	Please summarize the paper Factorized Fourier Neural Operators	[ "Factorized Fourier Neural Operator (F-FNO)", "separable Fourier representation", "neural operator architecture", "partial differential equations (PDEs)", "chaotic systems", "complex geometries", "residual connections", "teacher forcing", "Markov property", "numerical solvers", "cost-accuracy st...	[]	Single-Sum	[ 3 ]	[ { "reviews": [ { "summary": "In this work, the authors proposed a novel neural operator architecture that factorizes the convolution on Fourier space into separate dimensions. Consequentially, the F-FNO model can scale up to a higher number of layers and achieve smaller errors. The paper has a com...
4	Please summarize the paper DFPC: Data flow driven pruning of coupled channels without data.	[ "Coupled channels", "Data flow driven pruning", "Maximum Score Disagreement mechanism", "Group saliency", "Deep neural networks", "Residual connections", "Pruning algorithm", "Empirical performance", "Time complexity analysis", "Data-free mode", "Path aggregation", "VGG and ResNets", "Image ...	[]	Single-Sum	[ 4 ]	[ { "reviews": [ { "summary": "This paper tackles an important problem of neural network pruning. Specifically, the authors of the paper propose a novel method to prune coupled channels in neural networks. For instance, the layers with skip connections in the ResNet model are considered to be couple...
5	Please summarize the paper TVSPrune - Pruning Non-discriminative filters via Total Variation separability of intermediate representations without fine tuning	[ "Data-free pruning", "Total Variation distance", "class-conditional distributions", "LDIFF score", "IterTVSPrune", "discriminative filters", "sparsification potential", "deep neural networks", "feature maps outputs", "iterative pruning method", "pruning budget", "computational cost", "storag...	[]	Single-Sum	[ 5 ]	[ { "reviews": [ { "summary": "In this paper authors propose a mechanism to prune a convolutional neural network model in a relatively data-free manner i.e., they do not utilize training data or loss function for retraining the pruned model. However unlike the actual data-free pruning techniques the...
6	Please summarize the paper Finding Actual Descent Directions for Adversarial Training	[ "Danskins Descent Direction (DDD)", "adversarial training", "worst-case perturbations", "directional derivative", "minimax formulation", "counterexamples", "robust neural networks", "PGD (Projected Gradient Descent)", "theoretical guarantee", "computational complexity", "inner maximization", "...	[]	Single-Sum	[ 6 ]	[ { "reviews": [ { "summary": "This paper pays attention to the computation of adversarial training, by pointing out that even in the simple case (nonsmooth), the descent direction is not given by the worst-case perturbation, as opposed to common practice in the AT community. The paper then proposes...
7	Please summarize the paper A Study of Biologically Plausible Neural Network: the Role and Interactions of Brain-Inspired Mechanisms in Continual Learning	[ "Dales principle", "Active Dendrites", "Heterogeneous dropout", "Hebbian learning", "Synaptic consolidation", "Experience replay", "Biologically inspired neural networks", "Continual learning (CL)", "Biological plausibility", "Sparse coding", "Incremental learning tasks", "MNIST benchmark", ...	[]	Single-Sum	[ 7 ]	[ { "reviews": [ { "summary": "This paper evaluates previous work on biologically plausible DNNs in the setting of continual learning (CL). Namely, they evaluate ideas around Dale’s principle, Active Dendrites, heterogenous dropout, Hebbian learning, synaptic consolidation and experience replay. The...
8	Please summarize the paper Learning Continuous Normalizing Flows For Faster Convergence To Target Distribution via Ascent Regularizations	[ "ascent continuous normalizing flows", "Wasserstein gradient flows", "maximum likelihood estimation", "variational inference model", "continuous normalizing flows", "monotonically decreasing KL divergence", "target distribution convergence", "density estimation", "unbiased sampling", "variational ...	[]	Single-Sum	[ 8 ]	[ { "reviews": [ { "summary": "This paper discusses ascent regularization for training continuous normalizing flows (CNFs). This is motivated from Wasserstein gradient flows and results in an interesting regularization that encourages the learned model to be similar to the target distribution around...
9	Please summarize the paper pFedKT: Personalized Federated Learning via Knowledge Transfer	[ "personalized federated learning", "knowledge transfer schemes", "historical knowledge transfer", "hypernetwork", "global knowledge transfer", "contrastive learning loss", "local model updates", "generalization performance", "computational efficiency", "non-iid data", "theoretical convergence an...	[]	Single-Sum	[ 9 ]	[ { "reviews": [ { "summary": "This paper aims to improve the performance of personalized federated learning, and, for which, the authors propose two knowledge transfer schemes. In particular, the historical knowledge learned in the local clients is transferred from the hypernetwork, which stores th...
10	Please summarize the paper FARE: Provably Fair Representation Learning	[ "Fairness with Restricted Encoders (FARE)", "Fair representation learning (FRL)", "Demographic parity", "Optimal adversary", "Restricted encoder", "Finite sample analysis", "Decision tree encoder", "Upper bounds on unfairness", "Empirical risk minimization", "Data-processing inequality", "TV dis...	[]	Single-Sum	[ 10 ]	[ { "reviews": [ { "summary": "In this paper authors exploit the use of a restricted encoder to derive a provably fair (group fairness) representation which has the ability to upper bound the unfairness of any down stream classifier. They demonstrate this ability through the use of an optimal advers...
11	Please summarize the paper ONLINE RESTLESS BANDITS WITH UNOBSERVED STATES	[ "TSEETC", "Bayesian regret bound", "Restless Markov Bandit (RMAB)", "Thompson Sampling", "Unobservable states", "Explore-Then-Commit", "Markov chains", "Dirichlet priors", "Regret minimization", "Empirical experiments", "Complexity analysis", "State transition probabilities", "Explore-exploi...	[]	Single-Sum	[ 11 ]	[ { "reviews": [ { "summary": "This paper focuses on solving online restless bandits with unknown parameter and unobservable states by the proposed algorithm TSEETC. A Bayesian regret bound with $O(\\sqrt{T})$ dependency is established, which matches the lower bound dependency on $T$ and improves th...
12	Please summarize the paper Learning to aggregate: A parameterized aggregator to debias aggregation for cross-device federated learning	[ "Federated Learning (FL)", "Meta-learning framework", "Client drift", "Period drift", "Learning-based aggregation strategy", "Parameterized aggregator", "Proxy dataset", "Heterogeneous client data distributions", "Adaptive calibration parameter", "Model aggregation", "Debiasing model aggregation...	[]	Single-Sum	[ 12 ]	[ { "reviews": [ { "summary": "The paper presents a learnable aggregation scheme in the context of federated learning. The paper achieves this using meta-learning to generalize the parameters of the aggregator with a proxy dataset. The paper identifies 'period drift' in the current federated learnin...
13	Please summarize the paper Deep Reinforcement Learning based Insight Selection Policy	[ "Reinforcement learning framework", "Insight selection problem", "User behavior modeling", "Actionable insights", "Candidate insights scoring", "Health data analysis", "Simulation-based evaluation", "User preferences comprehension", "Multi-dimensional state transitions", "Reward function design", ...	[]	Single-Sum	[ 13 ]	[ { "reviews": [ { "summary": "This work provides a reinforcement learning solution for the insight selection problem and use two experiments to verify the feasibility of the proposed framework. The main claimed contribution is that the framework can provide insights that are both relevant to user p...
14	Please summarize the paper Data Leakage in Tabular Federated Learning	[ "TabLeak", "data leakage attack", "federated learning", "tabular data", "softmax structural priors", "pooled ensembling", "entropy-based uncertainty estimation", "mixed integer programming optimization", "softmax based continuous relaxation", "ensemble strategy", "reconstruction quality assessme...	[]	Single-Sum	[ 14 ]	[ { "reviews": [ { "summary": "This paper considers the data leakage attack in federated learning and focuses on the tabular data. A new method called TabLeak is proposed, which consists of three ingradients: (Section 3.1) softmax structural prios; (Section 3.2) pooled ensembling; and (Section 3.3) ...
15	Please summarize the paper Long-horizon video prediction using a dynamic latent hierarchy	[ "Dynamic Latent Hierarchy (DLH)", "spatiotemporal features", "long-term video prediction", "hierarchical representation learning", "latent variable model", "stochasticity modeling", "mixture of Gaussians", "disentangled temporal dynamics", "variational inference", "multi-object tracking", "KTH A...	[]	Single-Sum	[ 15 ]	[ { "reviews": [ { "summary": "The paper presents a method for hierarchical representation learning of spatiotemporal features in long-term video prediction. The proposed method is called: Dynamic Latent Hierarchy (DLH). The method distinguishes between features that are changing and those that are ...
16	Please summarize the paper SwinZS3: Zero-Shot Semantic Segmentation with a Swin Transformer	[ "Zero-shot semantic segmentation", "Swin transformer", "Pixel-text auxiliary segmentation loss", "Cross-entropy loss", "Regression loss", "Semantic consistency loss", "Language-guided activation fields", "Global feature relations", "Dense language-guided semantic prototypes", "Benchmark results", ...	[]	Single-Sum	[ 16 ]	[ { "reviews": [ { "summary": "The paper proposes a transformer based approach for zero shot semantic segmentation. It makes the use of different loss functions like cross entropy loss for seen classes, regression loss between language and visual features to account for unseen classes, a pixel text ...
17	Please summarize the paper Softened Symbol Grounding for Neuro-symbolic Systems	[ "Neuro-symbolic learning framework", "Symbol grounding problem", "Boltzmann distribution", "MCMC sampling method", "SMT solving", "Annealing mechanism", "Mixed strategy", "Stochastic gradient descent", "Visual Sudoku classification", "Connectivity barrier", "Feasible hidden symbol state", "Neu...	[]	Single-Sum	[ 17 ]	[ { "reviews": [ { "summary": "This paper presents a neuro-symbolic learning framework with an explicit design for addressing the symbol grounding problem. The key idea is softening symbol grounding by using a Boltzmann distribution to represent the entire symbol space, rather than a specific symbo...
18	Please summarize the paper Encoding Recurrence into Transformers	[ "Recurrence in Transformers", "Self-attention with recurrence", "Linear RNN", "Masked linear aggregation", "Gating function", "Temporal patterns", "Sample efficiency", "Inductive bias", "Sequential modeling tasks", "Transformer-RNN combination", "Performance benchmarks", "Learnable masked line...	[ "Overfitting prevention" ]	Single-Sum	[ 18 ]	[ { "reviews": [ { "summary": "The paper tackles the problem of endowing Transformers with the ability to encode information about the past via recurrence. The proposed architecture can leverage the recurrent connections to improve the sample efficiency while maintaining expressivity due to the use ...
19	Please summarize the paper Human-Guided Fair Classification for Natural Language Processing	[ "Individual fairness specifications", "Unsupervised style transfer", "Active learning approaches", "Toxicity classification", "Counterfactual examples", "Word replacement methods", "Similarity model training", "Diverse candidate pairs", "GPT-3 generation", "Empirical studies", "Human fairness in...	[ "Dataset quality control" ]	Single-Sum	[ 19 ]	[ { "reviews": [ { "summary": "This paper introduces a workflow/methodology to generate pairs of\nsimilar sentences that differ only wrt target/protected populations\nsuch as gender or race. The methodology inclues increasingly\nsophisticated steps, such as word replacement, unsupervised style\ntran...
20	Please summarize the paper Proper Scoring Rules for Survival Analysis	[ "Proper scoring rules", "Survival analysis", "Parameter vector w", "EM algorithm", "CQRNN algorithms", "Adapted scoring rules", "Logarithmic scoring rule", "Brier score", "Ranked Probability Score", "Calibration metric", "KL-divergence", "Censored data", "Training objectives", "Estimation ...	[]	Single-Sum	[ 20 ]	[ { "reviews": [ { "summary": "This work is after finding proper scoring rules in survival analysis. They have a parameter vector w whose true specification underpins the proofs their provide for the discussed scoring rules being proper. They approximate the parameter vector w using an EM algorithm ...
21	Please summarize the paper Social Network Structure Shapes Innovation: Experience-sharing in RL with SAPIENS	[ "Multi-agent topology", "Experience sharing", "Dynamic network structure", "Reinforcement learning agents", "DQN learners", "Innovation tasks", "Network interconnect settings", "Conformity and diversity metrics", "Replay buffers", "Social network structures", "Exploration and exploitation", "T...	[]	Single-Sum	[ 21 ]	[ { "reviews": [ { "summary": "The authors present study of the role of multi-agent topology on innovation towards goal of clarifying which social network structures are optimal for which innovation tasks, and which properties of experience sharing improve multi-level innovation. For multi-level hie...
22	Please summarize the paper Mini-batch $k$-means terminates within $O(d/\epsilon)$ iterations	[ "mini-batch k-means algorithm", "convergence rate analysis", "Lloyds algorithm", "batch size", "termination threshold", "upper bound result", "empirical evaluation", "sample complexity", "uniform sampling", "iteration count", "accuracy analysis", "local minima convergence", "sklearn implemen...	[]	Single-Sum	[ 22 ]	[ { "reviews": [ { "summary": "The paper analyzes the convergence rate of mini-batch k-means, namely, running Lloyd's iteration with a uniform sample of points from the data set, rather than using the entire set in each iteration. It gives strong results: with a sample size nearly quadratic in the d...
23	Please summarize the paper Convergence is Not Enough: Average-Case Performance of No-Regret Learning Dynamics	[ "pointwise convergence", "q-replicator dynamics", "Nash Equilibria", "average price of anarchy", "regions of attraction", "symmetric 2x2 coordination games", "social welfare", "no-regret algorithms", "empirical evidence", "bounded price-of-anarchy", "geometric approach", "dynamics convergence"...	[ "generalization to higher dimensions", "special case analysis", "objective function optimization" ]	Single-Sum	[ 23 ]	[ { "reviews": [ { "summary": "This paper proves pointwise convergence of q-replicator dynamics to NE and corresponding bounds on average price of anarchy, generalizing previous works.", "strength_and_weaknesses": "Strengths:\n\nThe results are solid. The motivation and proof ideas are well ...
24	Please summarize the paper Gene finding revisited: improved robustness through structured decoding from learning embeddings	[ "GeneDecoder", "Conditional Random Fields (CRF)", "Long Short-Term Memory (LSTM)", "Dilated Convolutional Layers", "Gene Annotation", "Neural Network Architecture", "Pre-trained Representations", "Gene Prediction", "Benchmarking Gene Prediction Tools", "G3PO Benchmark", "Isoform Separation", "...	[]	Single-Sum	[ 24 ]	[ { "reviews": [ { "summary": "The authors tackle the problem of annotating genes in newly-sequenced genomes. They develop a model called GeneDecoder which uses a combination of a CRF, LSTM and dilated convolutional layers. The authors show that the model relearns several properties of genes, includ...
25	Please summarize the paper Learning Uncertainty for Unknown Domains with Zero-Target-Assumption	[ "Maximum-Entropy Rewarded Reinforcement Learning (MERRL)", "optimal training set selection", "generalization to multiple unknown target domains", "training set entropy", "observational entropy (OE)", "prediction entropy (PE)", "Reinforcement Learning algorithms", "data selection policy network", "A2...	[]	Single-Sum	[ 25 ]	[ { "reviews": [ { "summary": "In this paper, the authors propose to use a Maximum-Entropy Rewarded Reinforcement Learning framework to select training data for NLP tasks, the goal of which is to maximize generalization. The authors experiment with A2C and SAC and experimental results show that the ...
26	Please summarize the paper Detecting Out-of-Distribution Data with Semi-supervised Graph “Feature" Networks	[ "[Out-of-distribution detection", "Graph-kernel-based method", "Graph embedding algorithms", "Semantic graph construction", "Geometric-learning-based framework", "Near OOD and Far OOD", "Feature extraction technique", "Human-interpretable concepts", "Low-dimensional representations", "Object detec...	[]	Single-Sum	[ 26 ]	[ { "reviews": [ { "summary": "In this paper authors propose a mechanism for deriving low-dimensional representations suitable for effective use of established non-parametric and parametric out-of-distribution data detection methods. Specifically they utilize graphs which represent relationships amo...
27	Please summarize the paper Let Offline RL Flow: Training Conservative Agents in the Latent Space of Normalizing Flow	[ "Conservative Normalizing Flows", "Offline Reinforcement Learning (ORL)", "Bounded uniform distribution", "Latent space optimization", "Action sampling", "Generative model", "Normalizing flows", "Bijective mappings", "Fully invertible layers", "Policy optimization", "Data distribution support", ...	[]	Single-Sum	[ 27 ]	[ { "reviews": [ { "summary": "Offline RL is challenging when the learned RL policy drifts too far from the support of the dataset. Thus, many offline RL methods use some form of constrained or conservative policy update to ensure that the RL policy remains close to the behavior policy of the datase...
28	Please summarize the paper Machine Learning from Explanations	[ "Ground truth explanations", "Input masks", "Image classification models", "Synthetic datasets", "Real-world datasets", "Imbalanced settings", "Two-step training procedure", "Mapping layer", "KL divergence penalty", "Feature extraction layer", "Training sample complexity", "Ablation experiment...	[]	Single-Sum	[ 28 ]	[ { "reviews": [ { "summary": "This paper proposes a method to train more accurate image classification models by leveraging human explanations in the form of input masks to increase test accuracy. While the pitched idea is promising, this work is fundamentally flawed in a number of ways and needs t...
29	Please summarize the paper Functional Risk Minimization	[ "Functional Risk Minimization (FRM)", "Functional Generative Models", "Latent Variable Noise Model", "Maximum Likelihood Estimation (MLE)", "Approximate Algorithm", "Local Laplace Approximation", "Empirical Risk Minimization (ERM)", "Hierarchical Bayesian Model", "Data Point Function Association", ...	[]	Single-Sum	[ 29 ]	[ { "reviews": [ { "summary": "This paper presents a novel framework for supervised learning problems. First, it introduces functional generative models which represent $ P(x,y)$ in terms of a latent variable $ \\theta$ that is sampled independently of $ x $ and then determines $ y $ as a function $...
30	Please summarize the paper Latent Linear ODEs with Neural Kalman Filtering for Irregular Time Series Forecasting	[ "Neural ODE model", "Irregular time series forecasting", "Latent space dynamics", "Linear ODE specification", "Self-consistency", "Forward stability", "Kalman filter-inspired update", "Nonlinear encoder/decoder pair", "Benchmark datasets", "Expressivity of latent dynamics", "Observations mapping...	[]	Single-Sum	[ 30 ]	[ { "reviews": [ { "summary": "The authors propose a novel Neural ODE model that embeds the observations into a latent space with dynamics governed by a linear ODE. \nThey carefully show that the model satisfies self-consistency, which allows forecasting irregularly sampled time series and have some...
31	Please summarize the paper Transformer-based model for symbolic regression via joint supervised learning	[ "transformer-based model", "symbolic regression", "feature extractor", "joint supervised learning mechanism", "contrastive loss", "pointMLP", "recovery rates", "R2 scores", "supervised contrastive learning", "similarity of feature vectors", "expression skeletons", "preorder traversal", "benc...	[ "synthetic datasets", "data augmentation" ]	Single-Sum	[ 31 ]	[ { "reviews": [ { "summary": "This paper proposes a novel transformer-based model for symbolic regression (SR), which produces mathematical expression skeletons from data points. In the proposed method, a feature extractor using pointMLP is added and jointly trained using contrastive loss to realiz...
32	Please summarize the paper Gradient-Based Transfer Learning	[ "Support-query distributional shift problem", "Gradient-based function representation", "Meta-learning styled transfer learning", "Function representation mapping", "High-dimensional parameter space", "Noise cancellation in gradients", "Model-based meta-learning method", "Support and query data separa...	[ "Complexity of chaotic systems" ]	Single-Sum	[ 32 ]	[ { "reviews": [ { "summary": "This paper proposes to solve support-query distributional shift problem which has not been addressed by the previous meta-learning literatures. Instead of assuming that the same function f is used to sample both support and query set, they assume that different functio...
33	Please summarize the paper Coreset for Rational Functions	[ "Coreset construction", "Rational function fitting", "Bicriteria approximation", "Sensitivity sampling", "Time series data", "Approximation loss", "Merge-reduce algorithm", "Sublinear coreset size", "Empirical evaluation", "Approximation algorithms", "Loss preservation", "Polynomial degree", ...	[]	Single-Sum	[ 33 ]	[ { "reviews": [ { "summary": "This paper studies the problem of building a coreset for fitting rational functions to time series data. In particular, suppose $y_1, ..., y_n \\in\\mathbb{R}$ is a time series. Then, in rational function fitting, we want to find a rational function $r$ of degree $k$ s...
34	Please summarize the paper Transformer needs NMDA receptor nonlinearity for long-term memory	[ "NMDA-like nonlinearity", "transformer architecture", "reference memory", "working memory", "place cell representation", "GELU activation function", "recurrent positional encoding", "2D grid navigation", "spatial navigation", "feedforward network", "activation function hyperparameter", "memory...	[ "sparsity of connections" ]	Single-Sum	[ 34 ]	[ { "reviews": [ { "summary": "This paper applies the transformer model to spatial navigation problem in a grid world with labeled grid positions. The task is to predict the label of the next position that is either visited or unvisited. The paper connects this task to working memory and reference m...
35	Please summarize the paper Simple Spectral Graph Convolution from an Optimization Perspective	[ "Label propagation", "Krylov subspace methods", "Least squares fitting", "Polynomial approximation problem", "Chebyshev polynomials", "Contextual stochastic block model", "Heterophilous datasets", "Graph representation", "Node classification datasets", "Residual minimization problem", "Spectral ...	[ "Training data robustness", "Long-range interactions" ]	Single-Sum	[ 35 ]	[ { "reviews": [ { "summary": "The paper proposes a novel approach for shallow graph representation through combining the idea behind label propagation with Krylov subspace methods. Label propagation is applied to the node features, and then the closed form solution is substituted into a least squar...
36	Please summarize the paper QAID: Question Answering Inspired Few-shot Intent Detection	[ "Question Answering inspired Intent Detection (QAID)", "Dual-encoder based retrieval architecture", "Batch contrastive loss", "Late-interaction scores", "Token-level similarity", "Few-shot intent detection", "Contextualized token-level similarity scores", "Self-supervised contrastive pre-training", ...	[]	Single-Sum	[ 36 ]	[ { "reviews": [ { "summary": "The paper proposes QAID, Question Answering inspired Intent Detection system, which models the intention detection classification as a question-answering task. The model uses two stages of training: a pretraining for better query representation and finetuning on few-sh...
37	Please summarize the paper Rethinking the Value of Prompt Learning for Vision-Language Models	[ "Prompt learning paradigm", "Classifier fine-tuning", "Zero-shot classification", "Handcrafted prompts", "Negative prompts", "Random prompts", "Vision-language models", "Robustness to distribution shifts", "Optimality-generalization trade-off", "Parameter-efficient adaptation", "Contrastive lear...	[ "Compositionality in prompts", "Parameter count analysis" ]	Single-Sum	[ 37 ]	[ { "reviews": [ { "summary": "In this work, prompt learning is reexamined, and several unexpected findings that defy accepted notions of the prompt are presented. First , random prompts without learning or fine-grained design may likewise function effectively in zero-shot recognition. Second, direc...
38	Please summarize the paper Disentangled Feature Swapping Augmentation for Weakly Supervised Semantic Segmentation	[ "Weakly supervised semantic segmentation", "Data augmentation technique", "Disentangling feature representation", "Foreground-background separation", "Localization maps", "Pseudo groundtruth", "mIoU improvement", "Two-way swapping method", "Spurious correlation", "Disentanglement loss", "Feature...	[]	Single-Sum	[ 38 ]	[ { "reviews": [ { "summary": "This paper studies the phenomenal that WSSS performance will degrade with dataset biases whee a specific target object frequently appears in the background. To resolve it, the paper proposed an augmentation method that disentangles the target object and background-rela...
39	Please summarize the paper Distributed Least Square Ranking with Random Features	[ "Distributed learning", "Pairwise ranking", "Random features", "DRank-RF", "DLSRank", "DLSRank-C", "Convergence analysis", "Communication strategy", "Theoretical assessments", "Numerical experiments", "Efficient pairwise ranking", "Kernel learning", "Experimental comparison", "Convergence ...	[]	Single-Sum	[ 39 ]	[ { "reviews": [ { "summary": "The authors study the statistical properties of pairwise ranking using distributed learning and random features ( DRank-RF) and establish its convergence analysis in probability. Numerical results confirm the practical aspects of the theory.", "strength_and_wea...
40	Please summarize the paper Doing Fast Adaptation Fast: Conditionally Independent Deep Ensembles for Distribution Shifts	[ "Conditionally Independent Deep Ensembles", "conditional mutual information (CMI)", "shortcut learning datasets", "ensemble learning", "output distributions", "diversity of ensembles", "predictive signals", "loss components", "confident-prediction regularization", "data generation process", "dis...	[]	Single-Sum	[ 40 ]	[ { "reviews": [ { "summary": "This paper quantifies a notion of diversity for deep ensembles that facilitates efficient estimation. The authors show that it is sufficient to enforce conditional independence on the output distributions of the classifiers. This leads to their main contribution concer...
41	Please summarize the paper Solving stochastic weak Minty variational inequalities without increasing batch size	[ "stochastic extragradient-type algorithms", "stochastic weak Minty variational inequality", "bias-corrected stochastic extragradient (BCSEG+) algorithm", "Minty variational inequality (MVI)", "constant stepsizes", "bounded batchsizes", "inclusion problems", "unconstrained smooth case", "constrained ...	[]	Single-Sum	[ 41 ]	[ { "reviews": [ { "summary": "For the first time, the authors introduces a family of stochastic extragradient-type algorithms that positively solves a class of nonconvex-nonconcave problems which can be cast as stochastic weak Minty variational inequality (MVI). In the monotone setting, extragradie...
42	Please summarize the paper Diversity Boosted Learning for Domain Generalization with a Large Number of Domains	[ "Diversity boosted two-level sampling framework", "Domain generalization problem", "Spurious correlations", "Inverse DANN", "Determinantal Point Process (DPP)", "Domain similarity matrix", "Data-side impacts", "Object-side spurious correlations", "Causal correlations", "Robust model training", "...	[]	Single-Sum	[ 42 ]	[ { "reviews": [ { "summary": "This paper:\n* solves the domain generalization problem\n* presents several observations that diversity helps mitigate serious correlations\n* proposes a sampling method that helps train robust models\n* conducts experiment on Rotated MNIST and Rotated Fashion MNIST to...
43	Please summarize the paper Towards Performance-maximizing Network Pruning via Global Channel Attention	[ "Global channel attention", "Channel saliency", "Static pruning", "Dynamic pruning", "Learn-to-rank algorithm", "Bayesian-based regularization", "Channel attention prior", "Network pruning framework", "Channel importance computation", "Global rank of channel saliencies", "Inter-channel relations...	[]	Single-Sum	[ 43 ]	[ { "reviews": [ { "summary": "The proposed work initially obtains a majority vote-based prior on the global rank of channel saliencies before forcing each sample-level channel saliency to match the global prior. In this way, the proposed work aims to use the platform of static pruning yet match the...
44	Please summarize the paper Adaptive Block-wise Learning for Knowledge Distillation	[ "Adaptive Block-wise Learning", "Knowledge Distillation (KD)", "Bi-level optimization scheme", "Auxiliary networks", "Layer-wise learnable parameters", "Local error signals", "Teacher-student knowledge allocation", "Gradient propagation", "Knowledge contribution balancing", "Homogeneous and hetero...	[]	Single-Sum	[ 44 ]	[ { "reviews": [ { "summary": "The manuscript observes the problem of fixed contributions of ground truth knowledge and teacher knowledge at different blocks of the student networks during knowledge distillation training. The author proposes a bi-level optimization scheme to balance the knowledge on...
45	Please summarize the paper Curriculum-based Co-design of Morphology and Control of Voxel-based Soft Robots	[ "Curriculum-based co-design method", "Voxel-Based Soft Robots (VSR)", "Reinforcement Learning (RL)", "Predefined curriculum", "Self-attention mechanism", "Neural Cellular Automata (NCA)", "Transformer-based control policy", "End-to-end training", "Design and control policies", "Ablation studies", ...	[]	Single-Sum	[ 45 ]	[ { "reviews": [ { "summary": "This paper introduces a new curriculum-based method for co-designing morphology and control of voxel-based soft robots. This curriculum-based method expands the design space from a small size to the target size using reinforcement learning with a predefined curriculum....
46	Please summarize the paper Object-Centric Learning with Slot Mixture Models	[ "Dual-level retrieval mechanism", "Gaussian mixture model (GMM)", "Slot-based model", "Slot mixture model (SMM)", "Expectation-maximization (EM) algorithm", "Object-centric representations", "Clustering center representation", "Density function learning", "Incremental contribution mechanism", "Set...	[]	Single-Sum	[ 46 ]	[ { "reviews": [ { "summary": "This paper proposes to combine the slot-based model with the gaussian mixture model (GMM) to improve the object-centric model. It explicitly represents the slot as the clustering center and uses the distance between slots to learn the mixture model. The experiments sho...
47	Please summarize the paper WiNeRT: Towards Neural Ray Tracing for Wireless Channel Modelling and Differentiable Simulations	[ "Neural surrogate", "Ray tracing simulation", "Wireless signal propagation", "Ray-Surface Interaction", "Differentiable neural ray tracer", "Implicit neural network", "Time-angle channel impulse response", "Hybrid heuristic/learned ray tracer", "Computational complexity", "Inverse problems", "Ne...	[]	Single-Sum	[ 47 ]	[ { "reviews": [ { "summary": "This paper proposes a neural network based solution to heuristically solve a wireless signal (physics) rendering problem. Given the environment set-up and configurations of the transmitter and receivers, the pre-trained network is able to simulate the wireless signal p...
48	Please summarize the paper Pocket-specific 3D Molecule Generation by Fragment-based Autoregressive Diffusion Models	[ "FragDiff", "3D molecule generation", "diffusion model", "autoregressive model", "local-to-global generation", "E(3)-equivariant graph neural networks", "molecular fragments", "atom types and bond types", "focal predictor", "molecule discriminator", "structure similarity metrics", "binding aff...	[]	Single-Sum	[ 48 ]	[ { "reviews": [ { "summary": "A general framework called FragDiff for pocket-specific 3D molecule generation is introduced. In particular, the generation process is executed in a local-to-global style. Namely, the diffusion model is adopted to generate the local fragment from scratch, while the aut...
49	Please summarize the paper Towards scalable and non-IID robust Hierarchical Federated Learning via Label-driven Knowledge Aggregator	[ "Hierarchical Federated Learning (FL)", "Label-Driven Knowledge Distillation (LKD)", "Non-IID data handling", "Full-stack FL (F2L)", "Global server aggregation", "Computational efficiency in FL", "Generalization gap reduction", "Knowledge distillation parameters", "Root dataset dependency", "Scala...	[]	Single-Sum	[ 49 ]	[ { "reviews": [ { "summary": "This paper proposes a hierarchical FL framework with a new label-driven distillation method to handle non-iid FL scenarios", "strength_and_weaknesses": "S1. The proposal of a hierarchical structure for FL is reasonable.\n\nW1. The paper is very hard to follow, ...
50	Please summarize the paper LS-IQ: Implicit Reward Regularization for Inverse Reinforcement Learning	[ "Least Squares Inverse Q Learning (LS-IQ)", "Implicit reward regularization", "Inverse reinforcement learning (IRL)", "Chi-squared divergence", "Mixture distribution", "Absorbing states", "Regularization critic", "Q-function parameterization", "Online imitation learning", "Entropy-regularized leas...	[]	Single-Sum	[ 50 ]	[ { "reviews": [ { "summary": "The paper studies the problem of imitation learning, building on the recent IQ-learn framework. Instead of an adversarial reward-policy loss like GAIL, IQ-learn instead parameterizes the Q-function so that the policy can be directly extracted. While IQ-learn works fine...
51	Please summarize the paper Online black-box adaptation to label-shift in the presence of conditional-shift	[ "label shift adaptation", "conditional shift", "online learning", "OOD validation set", "empirical investigation", "non-invertible confusion matrix", "label shift condition", "regression settings", "heuristic methods", "distribution shifts", "covariate shift", "general distribution shifts", ...	[ "model selection" ]	Single-Sum	[ 56 ]	[ { "reviews": [ { "summary": "It is well known that the performance of machine learning models is highly dependent on the distribution of the data on which it is evaluated: model performance deteriorates when tested on data generated from a distribution shifted with respect to the training data gen...
52	Please summarize the paper RuDar: Weather Radar Dataset for Precipitation Nowcasting with Geographical and Seasonal Variability	[ "Weather radar dataset", "Precipitation nowcasting task", "Geographical and climatic conditions", "Baseline methods", "Ablation study experiments", "Cross evaluation", "State-of-the-art deep learning methods", "Uncertainty estimation", "Mean Squared Error (MSE)", "Data acquisition process", "Eva...	[]	Single-Sum	[ 57 ]	[ { "reviews": [ { "summary": "1) This paper propsed a different weather prediction model by inducing a new dataset [above russia] which is adopted to evaluate all current models such as ConvLSTM etc. \n\n2) The results looks promising on this new datasets.", "strength_and_weaknesses": "Pro:...
53	Please summarize the paper Learning Representations for Reinforcement Learning with Hierarchical Forward Models	[ "Hierarchical models", "Latent representations", "Temporal abstraction", "Model-free reinforcement learning", "Communication network", "Latent dynamics model", "Sample efficiency", "Modified SAC agent", "Ablation studies", "Dynamic models", "Temporal coarseness levels", "Forward models", "Re...	[]	Single-Sum	[ 58 ]	[ { "reviews": [ { "summary": "Learn a hierarchy of latent models where each level takes in the latent state and a length n concatenation of actions and predicts the state after n steps, instead of simply predicting single step transitions. Information is passed from the layers predicting longer tem...
54	Please summarize the paper Sleep Tracking Apps' Design Choices: A Review	[ "Mental Health or Wellbeing", "Sleep", "Clinical Health", "Preparation", "Behavior Change", "Measure Taken to Protect Privacy", "Broad Ethical Discussion", "Informing the discussion of results", "Empirical" ]	[]	Single-Sum	[ 59 ]	[ { "Retrieval Scope": { "simple": [ 59 ], "middle": [ 19, 20, 28, 110, 127, 147, 202, 206, 231, 293, 295, 328, 359, 388, 59, 60, 61, 62, ...
55	Please summarize the paper Graceful Interactions and Social Support as Motivational Design Strategies to Encourage Women in Exercising	[ "Physical Activity", "Social Interactions", "Clinical Health", "Collection", "A Specific Goal (e.g., Pregnancy)", "Social Connection", "Creative Self-Expression", "Documenting or integrating into routine", "Socially or with sharing", "A need for more motivation or engagement" ]	[]	Single-Sum	[ 60 ]	[ { "Retrieval Scope": { "simple": [ 60 ], "middle": [ 19, 20, 28, 110, 127, 147, 202, 206, 231, 293, 295, 328, 359, 388, 59, 60, 61, 62, ...
56	Please summarize the paper GeniAuti: Toward Data-Driven Interventions to Challenging Behaviors of Autistic Children through Caregivers’ Trackingn	[ "Mental Health or Wellbeing", "Challenging Behaviors", "Clinical Health", "Behavior Change", "Documenting or integrating into routine", "Planning, making sense of data, or collecting accurate data", "Emotional burden of tracking", "Socially or with sharing", "Negative Consequences of Tracking", "T...	[]	Single-Sum	[ 61 ]	[ { "Retrieval Scope": { "simple": [ 61 ], "middle": [ 19, 20, 28, 110, 127, 147, 202, 206, 231, 293, 295, 328, 359, 388, 59, 60, 61, 62, ...
57	Please summarize the paper Understanding People's Perceptions of Approaches to Semi-Automated Dietary Monitoring	[ "Food", "Both Clinical Health and Wellness", "Behavior Change", "Habit Awareness", "Management of a Chronic Condition", "Curiosity", "Having a Record", "Documenting or integrating into routine", "Abandonment or lapsing", "Emotional burden of tracking", "Broad Privacy Discussion" ]	[]	Single-Sum	[ 62 ]	[ { "Retrieval Scope": { "simple": [ 62 ], "middle": [ 19, 20, 28, 110, 127, 147, 202, 206, 231, 293, 295, 328, 359, 388, 59, 60, 61, 62, ...
58	Please summarize the paper Identifying factors that inhibit self-care behavior among individuals with severe spinal cord injury	[ "Clinical Health", "Documenting or integrating into routine", "Planning, making sense of data, or collecting accurate data", "Measure Taken to Protect Privacy", "Informing the discussion of results", "Recruited by a Domain Expert", "Recruitment Aide", "Documenting or integrating into routine", "Aban...	[]	Single-Sum	[ 63 ]	[ { "Retrieval Scope": { "simple": [ 63 ], "middle": [ 19, 20, 28, 110, 127, 147, 202, 206, 231, 293, 295, 328, 359, 388, 59, 60, 61, 62, ...
59	Please summarize the paper A Model of Socially Sustained Self-Tracking for Food and Diet	[ "Both Clinical Health and Wellness", "Behavior Change", "Planning, making sense of data, or collecting accurate data", "Measure Taken to Protect Privacy", "Differing or evolving goals", "A Model of Personal Informatics", "From a Snowball Sample" ]	[]	Single-Sum	[ 64 ]	[ { "Retrieval Scope": { "simple": [ 64 ], "middle": [ 19, 20, 28, 110, 127, 147, 202, 206, 231, 293, 295, 328, 359, 388, 59, 60, 61, 62, ...
60	Please summarize the paper Thresholded Lexicographic Ordered Multi-Objective Reinforcement Learning	[ "Lexicographic multi-objective problems", "Policy gradient algorithm", "Lexicographic projection algorithm", "Thresholded lexicographic ordered objectives", "Multi-objective reinforcement learning (MORL)", "Heuristic projection of gradients", "Constrained objective", "Unconstrained objective", "Grad...	[]	Single-Sum	[ 101 ]	[ { "reviews": [ { "summary": "The authors discuss shortcomings of existing approaches for thresholded lexicographic ordered multiobijective problems in reinforcement learning. Additionally, the authors provide a policy gradient algorithm that performs well on this class of problems.", "str...
61	Please summarize the paper HOW SAMPLING AFFECTS TRAINING: AN EFFECTIVE SAMPLING THEORY STUDY FOR LONG-TAILED IMAGE CLASSIFICATION	[ "Effective sampling theory", "Jitter sampling strategy", "Long-tailed image classification", "Effective sample utilization", "Total number of effective samples", "Sampling with/without replacement", "Decoupling representation and classifier", "Long-tail learning benchmarks", "Sampling approaches com...	[]	Single-Sum	[ 102 ]	[ { "reviews": [ { "summary": "This paper studies long-tailed image classification problem. The authors identify two key factors that affect the performance of long-tailed image classification: (1) the total number of effective samples and (2) the effective sample utilization. Based on this finding...
62	Please summarize the paper EquiMod: An Equivariance Module to Improve Visual Instance Discrimination	[ "Equivariance regularizer", "Self-supervised learning", "Data augmentation", "Invariant baselines", "Equivariance projection head", "Equivariant predictor", "Latent space representation", "Contrastive visual representation learning", "Linear probe improvements", "Augmentation parameters", "Equim...	[ "local minima avoidance" ]	Single-Sum	[ 103 ]	[ { "reviews": [ { "summary": "This paper proposes an equivariance regularizer as a modification to the usual invariance-inducing self-supervised losses. This is an interesting approach to enabling equivariance as there is no need to have a special architecture as prior work. The authors are able to...
63	Please summarize the paper Manipulating Multi-agent Navigation Task via Emergent Communications	[ "Multi-agent navigation framework", "Emergent communication setting", "Vision-language navigation task", "Reinforcement learning model", "Gridworld environment", "Language emergence", "Observation space design", "Bidirectional communication", "Complexity of visual and language components", "Graph ...	[ "gridworld environment", "neural network architecture", "language evolution analysis", "ablation studies", "environment complexity", "graph networks" ]	Single-Sum	[ 104 ]	[ { "reviews": [ { "summary": "The paper proposes a multi-agent navigation framework as a new benchmark to assess the emergent language that is thus evolved between the agents.", "strength_and_weaknesses": "The paper is not clearly written and some sections are hard to understand. The relate...
64	Please summarize the paper Task-Aware Information Routing from Common Representation Space in Lifelong Learning	[ "Continual learning model", "Global workspace theory", "Common representation space", "Task-specific attention modules", "Catastrophic forgetting", "Self-regulated neurogenesis", "Experience replay", "Class-Incremental Learning", "Task-Incremental Learning", "Feature selectors", "Functional regu...	[ "task action space" ]	Single-Sum	[ 105 ]	[ { "reviews": [ { "summary": "In this paper, the authors introduced TAMiL, a continual-learning model inspired by the global workspace theory that can learn multiple tasks without catastrophic forgetting by constructing a common representation space across tasks. By combining previous approaches on...
65	Please summarize the paper CodeBPE: Investigating Subtokenization Options for Large Language Model Pretraining on Source Code	[ "Subtokenization strategies", "UnigramLM", "BPE tokenization", "Tokenization schemes", "Vocabulary size restriction", "Input length compression", "PLBART-like models", "Code generation", "Code summarization", "Code clone detection", "Downstream tasks", "Punctuation combination tokenization", ...	[]	Single-Sum	[ 106 ]	[ { "reviews": [ { "summary": "The authors of this paper propose various subtokenization strategies affecting input string length in a way so as to improve the efficiency of LLMs when trained on source code and when applied to downstream tasks such as code generation, code summarization and code clo...
66	Please summarize the paper Transport with Support: Data-Conditional Diffusion Bridges	[ "Iterative Smoothing Bridge (ISB)", "Schrodinger bridge", "sparse observations", "forward and backward drift functions", "Iterative Proportional Fitting procedure (IPFP)", "particle filtering", "constrained stochastic processes", "computationally efficient framework", "theoretical formulation", "d...	[]	Single-Sum	[ 107 ]	[ { "reviews": [ { "summary": "In this paper, the authors propose to add sparse constraints to the original Schrodinger bridges through optimal control. Specifically, the paper assumes that there exist some intermediate sparse samples during the diffusion process. By modifying the Iterative Proporti...
67	Please summarize the paper Randomized Sharpness-Aware Training for Boosting Computational Efficiency in Deep Learning	[ "Randomized Sharpness-Aware Training (RST)", "Sharpness-Aware Minimization (SAM)", "G-RST", "Bernoulli trial", "Parameterized Bernoulli distribution", "Non-convex stochastic cases", "Convergence properties", "Image classification tasks", "Computational efficiency", "Scheduling function", "Perfor...	[]	Single-Sum	[ 108 ]	[ { "reviews": [ { "summary": "This paper presents new training methods called RST and G-RST to extend the geometry inspired training method SAM and improve its computational efficiency.\nBased on randomized gradient boosting RST randomly selects between SGD and SAM with probability based on paramet...
68	Please summarize the paper Self-Supervised Off-Policy Ranking via Crowd Layer	[ "Off-Policy Evaluation (OPE)", "pairwise policy representation", "transformer architecture", "crowd layer", "policy ranking", "self-supervised learning", "noisy labels", "policy ranking pipeline", "end-to-end method", "hyper-parameter tuning", "ablation studies", "relative rank evaluation", ...	[]	Single-Sum	[ 109 ]	[ { "reviews": [ { "summary": "This paper proposes a new method for ranking of offline RL policies with off-policy evaluation (OPE). The ranking is produced with a model that 1) learns a pairwise policy representation with a transformer architecture, 2) uses a crowd layer to aggregate OPE scores of ...
69	Please summarize the paper Humanly Certifying Superhuman Classifiers	[ "Superhuman performance certification", "Human annotator performance bounds", "Inter-annotator agreement", "Machine learning model accuracy", "Finite sample analysis", "Confidence intervals construction", "Majority voting aggregation", "Positively correlated annotations", "Oracle accuracy estimation...	[]	Single-Sum	[ 110 ]	[ { "reviews": [ { "summary": "This paper proposes an approach to certify whether a given machine learning model achieves super-human performance when the dataset labels are (possibly erroneous) human annotations and not (unobserved) ground-truth labels. The proposed approached relies on proving the...
70	Please summarize the paper Share Your Representation Only: Guaranteed Improvement of the Privacy-Utility Tradeoff in Federated Learning	[ "CENTAUR algorithm", "differential privacy", "federated learning", "linear representation learning", "utility-privacy tradeoff", "Gaussian mechanism", "personalized model", "shared representation", "empirical performance", "statistical heterogeneity", "communication efficiency", "rate improvem...	[ "non-private gradient" ]	Single-Sum	[ 111 ]	[ { "reviews": [ { "summary": "The paper presents a new algorithm for DP federated learning (with trusted central server) under model personalization assumptions. The algorithm (CENTAUR) works by having the clients only send a subset of their parameters to the central server to be trained privately....
71	Please summarize the paper Quantized Disentangled Representations for Object-Centric Visual Tasks	[ "Slot Attention", "Vector Quantization", "Object-Centric Representation", "Disentangled Representations", "Set Prediction Task", "Disentanglement Metrics", "Latent Space", "Learnable Codebooks", "DQCF-micro", "DQCF-macro", "CLEVR Dataset", "Empirical Evaluation", "Quantitative Analysis", "...	[ "Beta-VAE Style Loss", "Cross Products and Dot Products" ]	Single-Sum	[ 112 ]	[ { "reviews": [ { "summary": "This work proposed to combine the idea of slot-attention and vector quantization to learn discrete object-centric representation of visual scenes. The proposed model utilizes slot-attention to decompose the image into a set of object-centric slots, and then transform e...
72	Please summarize the paper Supervised Random Feature Regression via Projection Pursuit	[ "Kernelized Ridge regression", "Random features", "Deep-learning inspired preprocessing", "Expressive kernels", "Generalized Spectral Kernels", "Supervised random features (SRF) regression", "Kernel methods", "Two-layer neural network", "Analytical tractability", "Kernel neural networks", "Impli...	[]	Single-Sum	[ 113 ]	[ { "reviews": [ { "summary": "The paper proposes an approach for boosting the effectiveness of kernelized Ridge regression by first learning a set of random features through a deep-learning inspired preprocessing step, then applying kernelized Ridge regression to the learned features.", "st...
73	Please summarize the paper Graph Spline Networks for Efficient Continuous Simulation of Dynamical Systems	[ "Graph neural networks (GNN)", "Orthogonal spline collocation (OSC)", "Continuous spatial-temporal solutions", "Differentiable OSC", "Message-passing neural networks", "Partial differential equations (PDE)", "End-to-end differentiable model", "Dynamic systems modeling", "Heat equation", "Damped wa...	[ "Physics-Informed Neural Networks (PINN)" ]	Single-Sum	[ 114 ]	[ { "reviews": [ { "summary": "The aurhors provide GRAPHSPLINENETS, a novel deep learning approach to speed up simulation of physical systems with spatio-temporal continuous outputs by exploiting the synergy between graph neural networks (GNN) and orthogonal spline collocation (OSC).\n\nTwo differen...
74	Please summarize the paper xTrimoABFold: Improving Antibody Structure Prediction without Multiple Sequence Alignments	[ "Antibody structure prediction", "xTrimoABFold", "antibody language model", "multi-modal sequence search technique", "AlphaFold2 architecture", "competitive performance", "runtime reduction", "template searching", "PDB database", "EvoFormer", "structure module", "RMSD performance", "CDR regi...	[ "recurrent neural networks" ]	Single-Sum	[ 115 ]	[ { "reviews": [ { "summary": "Antibody structure prediction is a highly sought after problem in industrial drug discovery pipelines. As with standard proteins, accurate predictions can lead to a better understanding of an antibody's function. Unfortunately, current methods for antibody structure pr...
75	Please summarize the paper Few-Shot Domain Adaptation For End-to-End Communication	[ "Gaussian Mixture Density Network (MDN)", "Few-shot domain adaptation", "Channel distribution changes", "Adapter layer", "Affine transform", "Regularization objective", "Feature transformation formulation", "Symbol error rate (SER)", "Domain-invariant features", "Non-domain shifted dataset", "Pe...	[ "unsupervised domain adaptation" ]	Single-Sum	[ 124 ]	[ { "reviews": [ { "summary": "- The paper addresses the problem of handling domain-shifts that arises in generative learnt channel models in E2E communication systems in a few-shot setting.\n- The proposed domain adaptation approach is tailored around a Mixture Density Network (MDN) representing th...
76	Please summarize the paper HyPHEN: A Hybrid Packing Method and Optimizations for Homomorphic Encryption-Based Neural Network	[ "Hybrid packing method", "RAConv", "CAConv", "Fully Homomorphic Encryption (FHE)", "CKKS implementation", "convolution layers", "ResNet architecture", "homomorphic rotations", "computational complexity", "latency improvements", "multiplex packing", "private inference", "low-degree polynomial...	[]	Single-Sum	[ 125 ]	[ { "reviews": [ { "summary": "The paper presents a new HCNN implementation to perform private inference (PI) called HyPHEN. It proposes a replication-based convolution method (RAConv), which is innovatively alternated with the channel-aligned convolution method (CAConv). It also proposes a hybrid p...
77	Please summarize the paper Causal Inference for Knowledge Graph Completion	[ "Causal Knowledge Graph Completion (KGC)", "Causal inference techniques", "Artificial confounders", "Group theory perspective", "Knowledge graph construction", "Causal graph modeling", "Frequency penalty", "Inverse propensity scoring", "Evaluation metrics", "Empirical experiments", "Bias correct...	[]	Single-Sum	[ 126 ]	[ { "reviews": [ { "summary": "This paper proposed a causal KGC model to alleviate the data bias issues that come from the data collection process, and further proposed an evaluation method to evaluate the causal KGC models on observation datasets. Moreover, the authors also provided a different per...
78	Please summarize the paper Formal Specifications from Natural Language	[ "Natural language to formal specification", "Regular expressions (regex)", "First-order logic (FOL)", "Linear temporal logic (LTL)", "Fine-tuning pretrained language models", "Generalization capacity", "Synthetic datasets", "T5 model", "Out-of-distribution instances", "Translation performance", ...	[]	Single-Sum	[ 127 ]	[ { "reviews": [ { "summary": "The paper focuses on the problem of translating natural language into formal specifications, in particular, regular expressions (regex), first order logic (FOL), and linear temporal logical (LTL). Instead of training deep models from scratch for this problem, the autho...
79	Please summarize the paper DELTA: Diverse Client Sampling for Fasting Federated Learning	[ "Diverse client sampling strategies", "Federated Learning (FL)", "Convergence analysis", "Optimal client sampling strategies", "Partial client participation", "Gradient diversity", "Stochastic gradient approximation", "Federated importance sampling (FedIS)", "DELTA client sampling scheme", "Varian...	[]	Single-Sum	[ 128 ]	[ { "reviews": [ { "summary": "This paper proposes a new method to improve previous (cluster-based) important client sampling methods in federated learning. The new method is motivated by the insight that it would be beneficial to select clients from diverse groups. Convergence analysis are also pro...
80	Please summarize the paper Incremental Predictive Coding: A Parallel and Fully Automatic Learning Algorithm	[ "Incremental predictive coding (iPC)", "Expectation-Maximization (EM)", "Backpropagation (BP)", "Hierarchical Gaussian model", "Biological plausibility", "Localized EM algorithm", "Time-complexity analysis", "Convergence guarantees", "Generalization performance", "CPU implementation", "Predictiv...	[]	Single-Sum	[ 129 ]	[ { "reviews": [ { "summary": "This paper describes a variant of predictive coding, named incremental predictive coding (iPC), based on incremental EM, which it is argued should be considered a biologically plausible approach to learning in the brain. The complexity of iPC is considered in relation...
81	Please summarize the paper Learning Geometric Representations of Interactive Objects	[ "Equivariant representation learning framework", "Geometric-aware disentangled representations", "Agent-environment interaction", "Latent space decoupling", "Contrastive loss", "Equivariant loss term", "Isometric representation", "Agent and object states", "Theoretical foundation", "Empirical stud...	[]	Single-Sum	[ 130 ]	[ { "reviews": [ { "summary": "The paper addresses the task of estimating the states (translation) of the agent and the object that it interacts with using image evidence. Authors assume that the states and properties of the agent and the object are unknown but the actions are known. The key contrib...
82	Please summarize the paper Improve distance metric learning by learning positions of class centers	[ "skewed mean function", "distance metric learning", "global distribution", "local relationships", "class centers", "energy functions", "deep metric learning", "extensive experiments", "better accuracies", "uniform distributions", "regularization term", "feature representations", "discriminat...	[ "dynamic updates" ]	Single-Sum	[ 131 ]	[ { "reviews": [ { "summary": "This paper presents a method to expand the distance between centers of different classes. However, the contribution is not clear, and the article is an unfinished work. In section 2, nothing is revisited but the title. The writing is terrible. There are too many spelli...
83	Please summarize the paper The guide and the explorer: smart agents for resource-limited iterated batch reinforcement learning	[ "Iterated batch reinforcement learning", "Dyna-style Guide & Explore strategy", "Model-free reinforcement learning", "Decision-time planning", "Exploration model", "Temperature parameter", "Guide policy", "Model-based exploration", "Sample efficiency", "Resource utilization", "Acrobot task", "...	[ "Synthetic problems" ]	Single-Sum	[ 132 ]	[ { "reviews": [ { "summary": "This algorithm proposes an algorithm for iterated batch reinforcement learning. The algorithm uses model-free RL to learn a guide policy, and then uses decision-time planning to improve the policy. The decision-time planning uses some exploration method and a rollout p...
84	Please summarize the paper FairGBM: Gradient Boosting with Fairness Constraints	[ "FairGBM", "Gradient-Boosted Decision Trees (GBDT)", "fairness constraints", "proxy Lagrangian approach", "dual-player game formulation", "predictive performance", "fairness metrics", "in-processing learning framework", "empirical results", "benchmark datasets", "constrained optimization method"...	[]	Single-Sum	[ 133 ]	[ { "reviews": [ { "summary": "The paper proposed a new FairGBM method to train GBDT under fairness constraints that shows little impact to predictive performance but improved fairness metrics as compared to unconstrained GBDT. The major challenge in the Fairness constraint is from the non-different...
85	Please summarize the paper How (Un)Fair is Text Summarization?	[ "Bias analysis in text summarization", "Content bias", "Structural bias", "Position bias", "Style bias", "Representation score R(T", "g)", "Synthetic dataset", "CNN/DM news summarization dataset", "Abstractive and extractive summarization models", "Summarization algorithms", "Underrepresentati...	[]	Single-Sum	[ 134 ]	[ { "reviews": [ { "summary": "This paper investigates the presence bias introduced by text summarization models. The authors propose to measure bias in two dimensions: content and structure. They define content bias as tendency to mention a specific group (e.g., gender, religion, etc) in a text, wh...
86	Please summarize the paper Simulating Task-Free Continual Learning Streams From Existing Datasets	[ "task-free continual learning", "Beta distribution", "class distribution assignment", "permutation index", "benchmark for task-free continual learning", "maximum entropy principle", "continuous distribution changes", "evaluation protocol", "online stream learning", "conceptual algorithms", "syst...	[ "real-time evaluation frameworks", "stream of tasks", "data shape anomalies" ]	Single-Sum	[ 135 ]	[ { "reviews": [ { "summary": "This paper presents a framework for task-free continual learning. The framework consists of a method for converting any dataset to a task-free continual learning problem where information on the class is not required and is not explicit, but rather examples from differ...
87	Please summarize the paper Online Bias Correction for Task-Free Continual Learning	[ "Online Bias Correction", "Task-free continual learning", "Recency bias", "Experience replay mechanism", "Prediction bias quantification", "Surrogate classifier", "Final model layer adjustment", "Data sampling bias", "Bias metric", "Ablation study", "Neural network training", "Cumulative regre...	[ "computational overhead reduction" ]	Single-Sum	[ 136 ]	[ { "reviews": [ { "summary": "Paper proposes a method to correct a recency bias in replay-based task-free continual learning, but separately optimising the final connected layer of network from the rest of the network. Focuses on continual learning in vision with evaluations in the area.", ...
88	Please summarize the paper A Simple Contrastive Learning Objective for Alleviating Neural Text Degeneration	[ "Contrastive token learning", "Unlikelihood training", "Cross-entropy loss function", "Language modeling", "Open-domain dialogue generation", "Negative token suppression", "Positive token reinforcement", "Perplexity metric", "Diversity metrics", "Fine-tuning language models", "Experimental evalu...	[]	Single-Sum	[ 137 ]	[ { "reviews": [ { "summary": "This paper proposes a contrastive learning method to balance the learning of positive and negative tokens in text generation tasks (e.g., language modeling and open-domain dialogue generation tasks).", "strength_and_weaknesses": "Strength \n1. The paper is...
89	Please summarize the paper Enriching Online Knowledge Distillation with Specialist Ensemble	[ "online knowledge distillation framework", "diversified ensemble of teachers", "label prior shift", "Post-Compensated Softmax", "importance sampling", "teacher diversity", "teacher logits", "aggregated classifier", "empirical analysis", "specialist models", "CIFAR", "ImageNet", "peer-based d...	[]	Single-Sum	[ 138 ]	[ { "reviews": [ { "summary": "The authors propose to diversify the models in an ensemble forming the teachers in distillation. The specific setup the authors consider is on-line distillation (where the teachers are trained in parallel with the student) and peer-based (where the models share parts ...
90	Please summarize the paper Improved Gradient Descent Optimization Algorithm based on Inverse Model-Parameter Difference	[ "Adaptive learning rate method", "Stochastic gradient descent algorithm", "Inverse parameter difference", "Convergence proof", "Empirical studies", "Regret bound analysis", "First-order optimizer", "Training loss", "CIFAR-10/100 datasets", "Tiny-ImageNet", "Element-wise step sizes", "Neural ne...	[]	Single-Sum	[ 139 ]	[ { "reviews": [ { "summary": "This paper proposes an adaptive learning rate method to improved the stochastic gradient descent algorithm by leveraging the difference of two consecutive iterations. More concretely, the authors proposes that the learning rate of each parameter should be inversely pro...
91	Please summarize the paper Variational Learning ISTA	[ "unrolled ISTA-type algorithms", "compressed sensing", "adaptive measurements matrix", "sparsifying dictionary", "VLISTA", "probabilistic formulation", "out of distribution (OOD) detection", "variational learning", "signal recovery performance", "data-driven approach", "time varying sensing matr...	[]	Single-Sum	[ 140 ]	[ { "reviews": [ { "summary": "The authors present two unrolled ISTA-type algorithms for compressed sensing. The first approach is amenable to adaptive measurements matrix while learning the sparsifying dictionary per layer. The second approach called as VLISTA places a probability distribution the ...
92	Please summarize the paper Moment Distributionally Robust Probabilistic Supervised Learning	[ "DRO objective", "uncertainty set", "primal-dual formulation", "distributionally robust probabilistic supervised learning", "moment ambiguity sets", "adversarial training approach", "minimax problem", "dual formulation", "gradient descent algorithm", "empirical distribution", "first-order featur...	[]	Single-Sum	[ 141 ]	[ { "reviews": [ { "summary": "The paper proposes an algorithm to compute a DRO objective where the uncertainty set is defined with deviations in moments of the representation. The algorithm allows one to compute the minimax DRO loss using a primal-dual formulation and reasoning about the stationary...
93	Please summarize the paper CLEP: Exploiting Edge Partitioning for Graph Contrastive Learning	[ "Contrastive learning with edge partitioning (CLEP)", "Graph generative models", "Graph contrastive learning", "Intra- and inter-graph information", "Community-aware edge partition", "Latent community structure", "Cumulative latent node interactions", "Probabilistic framework", "Variational inferenc...	[]	Single-Sum	[ 142 ]	[ { "reviews": [ { "summary": "In this paper, author introduce a probabilistic framework called contrastive learning with edge partitioning (CLEP) that integrates generative modeling and graph contrastive learning. CLEP models edge generation by cumulative latent node interactions over multiple mutu...
94	Please summarize the paper Meta-Learning the Inductive Biases of Simple Neural Circuits	[ "Inductive bias", "Meta-learning framework", "Architectural design choices", "Function space features", "Adversarial training strategy", "High-dimensional end-to-end learning", "Bootstrapped learning framework", "Dimension collapse", "Learning rate variation", "Low-dimensional toy problems", "Su...	[]	Single-Sum	[ 143 ]	[ { "reviews": [ { "summary": "The authors present a meta-learning framework to infer a system's or circuit's inductive bias. In their work, the authors claim that their method connects architectural design choices to function space features. This work demonstrates the usability of the method in inf...
95	Please summarize the paper Accelerating spiking neural network training using the $d$-block model	[ "d-block model", "spiking neural networks (SNNs)", "accelerated computing", "sequential operations", "stochastic absolute refractory periods", "recurrent conductance latencies", "fast vectorized operations", "training time reduction", "state-of-the-art (SOTA) performance", "event datasets", "mem...	[]	Single-Sum	[ 144 ]	[ { "reviews": [ { "summary": "This work extends the 1-block model in Taylor et al. (2022) to the d-block model. Compared with the LIF model, the d-block model achieves accelerated computing on GPU by using fewer sequential operations.", "strength_and_weaknesses": "Strength:\n\nThe proposed ...
96	Please summarize the paper RG: OUT-OF-DISTRIBUTION DETECTION WITH REACTIVATE GRADNORM	[ "Gradient-based OOD detection", "Decomposition G(x)=UV", "Energy-based score", "Feature norm", "Clipped feature vector", "Out-of-distribution detection", "Reactivate Gradnorm", "Sensitivity analysis", "Hyperparameter selection method", "ImageNet benchmarks", "U and V terms", "Addition-based me...	[]	Single-Sum	[ 145 ]	[ { "reviews": [ { "summary": "The authors revisit gradient-based OOD detection from the perspective of backpropagation and extend the decomposition G(x)=UV of GradNorm to more loss functions. Here G is a gradient-based detection score, V is the feature norm and U represents output information. Acco...
97	Please summarize the paper Don’t fear the unlabelled: safe semi-supervised learning via debiasing	[ "Debiased semi-supervised learning", "Empirical Risk Minimization (ERM)", "Debiasing term", "Theoretical guarantees", "Calibration improvement", "CIFAR-10", "CIFAR-100", "FixMatch algorithm", "Unlabeled data weighting", "Control variates", "Constrained optimization", "Generalization bound", ...	[ "Radamacher complexity", "Statistical significance testing" ]	Single-Sum	[ 146 ]	[ { "reviews": [ { "summary": "This work identifies that the existing approaches in semi-supervised learning minimize a biased risk and hence, are devoid of theoretical guarantees unless there are strong assumptions. This works gives a method to de-bias the loss with a simple estimator which also al...
98	Please summarize the paper Gandalf : Data Augmentation is all you need for Extreme Classification	[ "Extreme multi-label classification", "Data augmentation techniques", "Gandalf", "LabelMix", "Label correlation graph", "Label features", "Query-label interpolation", "Label co-occurrence matrix", "Label-to-label augmentation", "Empirical evaluation", "Benchmark datasets", "Soft-labels", "De...	[]	Single-Sum	[ 147 ]	[ { "reviews": [ { "summary": "Extreme classification (or deep information retrieval) has been a popular research field that matches an input text (query) to a label that often is also a text. This paper focuses on a sub-field where both the input text and the label text are short. The proposed trai...
99	Please summarize the paper Learning a Data-Driven Policy Network for Pre-Training Automated Feature Engineering	[ "Automated feature engineering", "Markov decision process", "Transferability", "Policy networks", "Data-driven MDP setup", "Feature engineering actions", "Reinforcement learning algorithm", "Tabular datasets", "Computational costs", "Human behavior patterns", "Pre-training technique", "State-o...	[]	Single-Sum	[ 148 ]	[ { "reviews": [ { "summary": "This paper proposes a first of its kind architecture framework for automated feature engineering called Fetch, a system based on brand new data driven Markov decision process. The authors identify critical gaps by stating the current methods for AutoFe are insufficient...
100	Please summarize the paper Attention Flows for General Transformers	[ "Attention flow network", "Encoder-decoder transformers", "Max flow problem", "Shapley values", "Positional independence", "Token importance quantification", "Attention flow formulation", "Auto-regressive structure", "Game theoretical concepts", "Flow network construction", "Attention values nor...	[]	Single-Sum	[ 149 ]	[ { "reviews": [ { "summary": "This work extends the attention flow method proposed in Abnar and Zuidema 2020 to encoder-decoder and decoder-only transformers. The major contribution is based on the observation that later predicted words have more incoming edges than earlier words, such that to ensu...

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.