Representations learned by pre-training a neural network on a large dataset are increasingly used successfully to perform a variety of downstream tasks. In this work, we take a closer look at how features are encoded in such pre-trained representations. We find that learned representations in a given layer exhibit a degree of diffuse redundancy, i.e., any randomly chosen subset of neurons in the layer that is larger than a threshold size shares a large degree of similarity with the full layer and is able to perform similarly as the whole layer on a variety of downstream tasks. For example, a linear probe trained on 20% of randomly picked neurons from the penultimate layer of a ResNet50 pre-trained on ImageNet1k achieves an accuracy within 5% of a linear probe trained on the full layer of neurons for downstream CIFAR10 classification. We conduct experiments on different neural architectures (including CNNs and Transformers) pretrained on both ImageNet1k and ImageNet21k and evaluate a variety of downstream tasks taken from the VTAB benchmark. We find that the loss & dataset used during pre-training largely govern the degree of diffuse redundancy and the "critical mass" of neurons needed often depends on the downstream task, suggesting that there is a task-inherent redundancy-performance Pareto frontier. Our findings shed light on the nature of representations learned by pre-trained deep neural networks and suggest that entire layers might not be necessary to perform many downstream tasks. We investigate the potential for exploiting this redundancy to achieve efficient generalization for downstream tasks and also draw caution to certain possible unintended consequences.

Post-hoc gradient-based interpretability methods [1, 2] that provide instancespecific explanations of model predictions are often based on assumption (A): magnitude of input gradients--gradients of logits with respect to input--noisily highlight discriminative task-relevant features. In this work, we test the validity of assumption (A) using a three-pronged approach: 1. We develop an evaluation framework, DiffROAR, to test assumption (A) on four image classification benchmarks. Our results suggest that (i) input gradients of standard models (i.e., trained on original data) may grossly violate (A), whereas (ii) input gradients of adversarially robust models satisfy (A) reasonably well.

Previous work [2, 1] has observed that standard training and greedy decoding usually cause models to generate consecutive repetitive texts. These consecutive repetitive texts are redundant and do not convey new information, which is avoided in human language. There are three types of consecutive repetitions: word-level, phrase-level and sentence-level. The phrase-level means that a phrase consisting of several words is repeated consecutively. The sentence in our paper refers to a sequence split by '.!?' is repeated consecutively 2. We calculate the ratio of consecutive repetition in a sequence x as follows.

Tree-based models are used in many high-stakes application domains such as finance and medicine, where robustness and interpretability are of utmost importance. Yet, methods for improving and certifying their robustness are severely under-explored, in contrast to those focusing on neural networks. Targeting this important challenge, we propose deterministic smoothing for decision stump ensembles. Whereas most prior work on randomized smoothing focuses on evaluating arbitrary base models approximately under input randomization, the key insight of our work is that decision stump ensembles enable exact yet efficient evaluation via dynamic programming. Importantly, we obtain deterministic robustness certificates, even jointly over numerical and categorical features, a setting ubiquitous in the real world. Further, we derive an MLE-optimal training method for smoothed decision stumps under randomization and propose two boosting approaches to improve their provable robustness. An extensive experimental evaluation on computer vision and tabular data tasks shows that our approach yields significantly higher certified accuracies than the state-of-the-art for tree-based models. We release all code and trained models at https://github.com/eth-sri/drs.

Topic model evaluation, like evaluation of other unsupervised methods, can be contentious. However, the field has coalesced around automated estimates of topic coherence, which rely on the frequency of word co-occurrences in a reference corpus. Contemporary neural topic models surpass classical ones according to these metrics. At the same time, topic model evaluation suffers from a validation gap: automated coherence, developed for classical models, has not been validated using human experimentation for neural models. In addition, a meta-analysis of topic modeling literature reveals a substantial standardization gap in automated topic modeling benchmarks. To address the validation gap, we compare automated coherence with the two most widely accepted human judgment tasks: topic rating and word intrusion. To address the standardization gap, we systematically evaluate a dominant classical model and two state-of-the-art neural models on two commonly used datasets. Automated evaluations declare a winning model when corresponding human evaluations do not, calling into question the validity of fully automatic evaluations independent of human judgments.

Take the Portland Trail Blazers +2.5 in Game 3 Shocker! Kyle Brandt-Seth Rollins on-set spat was staged Tigers look to exploit Reds' struggles at home as Framber Valdez takes the mound in Cincinnati Watch as Eagles steal Makai Lemon with wild phone call: 'Why is Philly calling me?' Giants' draft pick has intense Jaxson Dart message: 'I'm ready to die for you' Donald Trump uses Pete Rose to justify soldier's alleged shady Maduro bet, and he's not wrong Ex-Michigan football coach Sherrone Moore's mistress reveals he got her pregnant during relationship Giants' bizarre draft decisions leave star player frustrated as true needs go unfulfilled in first round Rueben Bain's short arms and tragic car accident history contributed to his NFL Draft slide Sherrone Moore accuser Paige Shiver speaks out in new interview: he'had complete control over me' Megan Rapinoe calls on traditional WNBA media to be replaced with those who'understand queer culture' The NFL Draft continues to be one of the worst'sporting events' of the year'Fox & Friends' hosts learn country line dancing in Houston Veterans cheer Trump's order on psychedelic drugs to treat PTSD'Fox & Friends' hosts'get their Texas on' with Tecovas boots'Fox & Friends' kicks off the Fox News America 250 Tour in Houston Country artist Rich O'Toole joins'Fox & Friends' in Houston IDF finds'ambulance used by Hezbollah to conceal weapons' Hegseth shuts down reporter's EXTREME question OutKick 49ers GM John Lynch skeptical of Rams' decision to draft QB Ty Simpson with No. 13 overall pick Lynch called Simpson'a good football player' but noted the pick'surprised everybody' The San Francisco 49ers traded out of the NFL Draft's first round on Thursday, so general manager John Lynch didn't have a player to discuss when he met with reporters. No problem, because he started talking players a couple of division rivals drafted. Lynch commented on what the Arizona Cardinals and Los Angeles Rams did. San Francisco 49ers general manager John Lynch speaks at the NFL Scouting Combine at the Indiana Convention Center on Feb. 24, 2026.

Get ready for summer with discounts on robot pool cleaners from Beatbot. National Pool Opening Day is tomorrow, April 25, and summer is almost here, which means pool owners everywhere are getting ready to unveil the horrors of whatever happened during the off-season. Most of the Beatbot lineup is on sale at Amazon and Beatbot's own storefront, with prices starting at $499. Beatbot makes many of the best pool-cleaning robots we've tested, and we've highlighted our top picks below. Note that the discounts are scheduled to end on April 26, though items may sell out sooner.

But the X-class events aren't even close to the most powerful flare on record. More information Adding us as a Preferred Source in Google by using this link indicates that you would like to see more of our content in Google News results. NASA's Solar Dynamics Observatory captured these images of solar flares -- seen as the bright flashes in the top right -- on April 23 and 24, 2026. The images show a subset of extreme ultraviolet light that highlights the extremely hot material in flares and which is colorized in in gold and blue on the left and teal on the right. Breakthroughs, discoveries, and DIY tips sent six days a week.