Large language models (LLMs) integrated with retrieval-augmented generation (RAG) have improved factuality by grounding outputs in external evidence. However, they remain susceptible to unfaithful generation, where outputs contradict retrieved context despite its relevance and accuracy. Existing approaches aiming to improve faithfulness primarily focus on enhancing the utilization of external context, but often overlook the persistent influence of internal parametric knowledge during generation. In this work, we investigate the internal mechanisms behind unfaithful generation and identify a subset of mid-to-deep feed-forward networks (FFNs) that are disproportionately activated in such cases. Building on this insight, we propose Parametric Knowledge Muting through FFNSuppression (ParamMute), a framework that improves contextual faithfulness by suppressing the activation of unfaithfulness-associated FFNs and calibrating the model toward retrieved knowledge. To evaluate our approach, we introduce CoFaithfulQA, a benchmark specifically designed to evaluate faithfulness in scenarios where internal knowledge conflicts with accurate external evidence. Experimental results show that ParamMute significantly enhances faithfulness across both CoFaithfulQA and the established ConFiQA benchmark, achieving substantial reductions in reliance on parametric memory. These findings underscore the importance of mitigating internal knowledge dominance and provide a new direction for improving LLM trustworthiness in RAG.

Neural networks trained with ERM (empirical risk minimization) sometimes learn unintended decision rules, in particular when their training data is biased, i.e., when training labels are strongly correlated with undesirable features. To prevent a network from learning such features, recent methods augment training data such that examples displaying spurious correlations (i.e., bias-aligned examples) become a minority, whereas the other, bias-conflicting examples become prevalent. However, these approaches are sometimes difficult to train and scale to real-world data because they rely on generative models or disentangled representations. We propose an alternative based on mixup, a popular augmentation that creates convex combinations of training examples. Our method, coined SelecMix, applies mixup to contradicting pairs of examples, defined as showing either (i) the same label but dissimilar biased features, or (ii) different labels but similar biased features. Identifying such pairs requires comparing examples with respect to unknown biased features. For this, we utilize an auxiliary contrastive model with the popular heuristic that biased features are learned preferentially during training. Experiments on standard benchmarks demonstrate the effectiveness of the method, in particular when label noise complicates the identification of bias-conflicting examples.

Surface electromyography (sEMG) at the wrists could enable natural, keyboard free text entry, yet the state of the art emg2qwerty baseline still misrecognizes 51.8\% of characters zero shot on unseen users and 7.0\% after user specific fine tuning. We trace much of these errors to mismatched cross user signal statistics, fragile reliance on high order feature dependencies, and the absence of architectural inductive biases aligned with the bilateral nature of typing. To address these issues, we introduce three simple modifications: (i) Rolling Time Normalization which adaptively aligns input distributions across users; (ii) Aggressive Channel Masking, which encourages reliance on low order feature combinations more likely to generalize across users; and (iii) a Split and Share encoder that processes each hand independently with weight shared streams to reflect the bilateral symmetry of the neuromuscular system. Combined with a five fold reduction in spectral resolution (33$\rightarrow$6 frequency bands), these components yield a compact Split-and-Share model, SplashNet mini, which uses only the parameters and 0.6 the FLOPs of the baseline while reducing character error rate (CER) to 36.4\% zero shot and 5.9\% after fine tuning. An upscaled variant, SplashNet ( parameters, 1.15 FLOPs of the baseline), further lowers error to 35.7\% and 5.5\%, representing 31\% and 21\% relative improvements in the zero-shot and finetuned settings, respectively. SplashNet therefore establishes a new state-of-the-art without requiring additional data.

The hypothesis that Convolutional Neural Networks (CNNs) are inherently texture-biased has shaped much of the discourse on feature use in deep learning. We revisit this hypothesis by examining limitations in the cue-conflict experiment by Geirhos et al. To address these limitations, we propose a domain-agnostic framework that quantifies feature reliance through systematic suppression of shape, texture, and color cues, avoiding the confounds of forced-choice conflicts. By evaluating humans and neural networks under controlled suppression conditions, we find that CNNs are not inherently texture-biased but predominantly rely on local shape features. Nonetheless, this reliance can be substantially mitigated through modern training strategies or architectures (ConvNeXt, ViTs). We further extend the analysis across computer vision, medical imaging, and remote sensing, revealing that reliance patterns differ systematically: computer vision models prioritize shape, medical imaging models emphasize color, and remote sensing models exhibit a stronger reliance on texture.

Results averaged over ResNet18 and ResNet50.

Revealed: The message Norwegian PM sent Trump that sparked president's outburst saying Nobel Peace Prize snub justified Greenland land-grab The Kristi Noem photo that reveals why anti-ICE mob stormed Minnesota church as terrified child worshipper sobbed in father's arms We were $460,000 in debt. We weren't high earners and paid it off using simple but life-changing tricks... anyone can do it Idyllic city was hit by a surge in cancers and miscarriages when Trump's'beautiful baby' arrived. Joseph Gordon-Levitt was the hottest actor in Hollywood... then vanished: Unearthing family tragedy that sparked disappearance and has left'lasting' scars Mayor of gorgeous Oregon city that's home to Nike HQ explains simple reasons why it is thriving while neighboring Portland circles the drain Spanish train disaster victims flew through windows and were found hundreds of yards away, with more than 39 feared dead - as mystery over what caused'truly strange' crash grows Dark side of America's favorite vacation hotspot... where women are subjected to the most horrific sex attacks imaginable Dietitian reveals the game-changing supplements that work like Ozempic... and will super-charge your weight loss without side-effects Pierce Brosnan fans defend star's wife Keely Shaye Smith, 62, after cruel troll posts a photo of her when she met the 72-year-old Bond star alongside a recent snap as a'reminder to avoid marriage' My husband was acting odd for months. But nothing prepared me for what was hidden under the couch... undeniable proof of my worst fear John Barrowman breaks down in tears while cradling his dog's body after the beloved pet'waited until I got home' before dying peacefully in his arms Country singer Karley Scott Collins responds to rumors she's living with Keith Urban after Nicole Kidman split Is AI making us STUPID? Latest episode of Daily Mail's Wellness Explained examines what our increasing reliance on ChatGPT is doing to our brains Artificial Intelligence ( AI) chatbots like ChatGPT are now a daily part of life for millions of people - but what is that really doing to our brains?

As Europe's longstanding alliance with the US falters, its push to become a self-sufficient AI superpower has become more urgent. As the relationship between the US and its European allies shows signs of strain, AI labs across the continent are searching for inventive ways to close the gap with American rivals that have so far dominated the field. With rare exceptions, US-based firms outstrip European competitors across the AI production line--from processor design and manufacturing, to datacenter capacity, to model and application development. Likewise, the US has captured a massive proportion of the money pouring into AI, reflected in the performance last year of its homegrown stocks and the growth of its econonmy . The belief in some quarters is that the US-based leaders --Nvidia, Google, Meta, OpenAI, Anthropic, and the like--are already so entrenched as to make it impossible for European nations to break their dependency on American AI, mirroring the pattern in cloud services.

In many classification datasets, the task labels are spuriously correlated with some input attributes. Classifiers trained on such datasets often rely on these attributes for prediction, especially when the spurious correlation is high, and thus fail togeneralize whenever there is a shift in the attributes' correlation at deployment. If we assume that the spurious attributes are known a priori, several methods have been proposed to learn a classifier that is invariant to the specified attributes. However, in real-world data, information about spurious attributes is typically unavailable. Therefore, we propose a method that automatically identifies spurious attributes by estimating their causal effect on the label and then uses a regularization objective to mitigate the classifier's reliance on them.