We present phi-4, a 14-billion parameter language model developed with a training recipe that is centrally focused on data quality. Unlike most language models, where pre-training is based primarily on organic data sources such as web content or code, phi-4 strategically incorporates synthetic data throughout the training process. While previous models in the Phi family largely distill the capabilities of a teacher model (specifically GPT-4), phi-4 substantially surpasses its teacher model on STEM-focused QA capabilities, giving evidence that our data-generation and post-training techniques go beyond distillation. Despite minimal changes to the phi-3 architecture, phi-4 achieves strong performance relative to its size -- especially on reasoning-focused benchmarks -- due to improved data, training curriculum, and innovations in the post-training scheme.

Statistical downscaling is a technique used in climate modeling to increase the resolution of climate simulations. High-resolution climate information is essential for various high-impact applications, including natural hazard risk assessment. However, simulating climate at high resolution is intractable. Thus, climate simulations are often conducted at a coarse scale and then downscaled to the desired resolution. Existing downscaling techniques are either simulation-based methods with high computational costs, or statistical approaches with limitations in accuracy or application specificity. We introduce Generative Bias Correction and Super-Resolution (GenBCSR), a two-stage probabilistic framework for statistical downscaling that overcomes the limitations of previous methods. GenBCSR employs two transformations to match high-dimensional distributions at different resolutions: (i) the first stage, bias correction, aligns the distributions at coarse scale, (ii) the second stage, statistical super-resolution, lifts the corrected coarse distribution by introducing fine-grained details. Each stage is instantiated by a state-of-the-art generative model, resulting in an efficient and effective computational pipeline for the well-studied distribution matching problem. By framing the downscaling problem as distribution matching, GenBCSR relaxes the constraints of supervised learning, which requires samples to be aligned. Despite not requiring such correspondence, we show that GenBCSR surpasses standard approaches in predictive accuracy of critical impact variables, particularly in predicting the tails (99% percentile) of composite indexes composed of interacting variables, achieving up to 4-5 folds of error reduction.

As waves of loud, car-sized mystery drones continue to buzz over New Jersey, one family reported that the craft changed time on their car's clock. The family of Morris County locals said they were following one of these seemingly terrestrial UFOs in their vehicle, only to experience the odd effect on their car's electronics as the unexplained craft'hovered above them.' 'The clock in their car changed time,' according to one Fox News reporter who spoke to the unnamed family. 'They say the clock went back to normal after they drove off.' While local law enforcement in Morris County has issued a statement asserting that'there is no known threat to public safety' at this time -- the Federal Aviation Administration (FAA) has issued a ban on drone flights over sensitive areas in state.

Recent advances in multimodal training have significantly improved the integration of image understanding and generation within a unified model. This study investigates how vision-language models (VLMs) handle image-understanding tasks, specifically focusing on how visual information is processed and transferred to the textual domain. We compare VLMs that generate both images and text with those that output only text, highlighting key differences in information flow. We find that in models with multimodal outputs, image and text embeddings are more separated within the residual stream. Additionally, models vary in how information is exchanged from visual to textual tokens. VLMs that only output text exhibit a distributed communication pattern, where information is exchanged through multiple image tokens. In contrast, models trained for image and text generation rely on a single token that acts as a narrow gate for the visual information. We demonstrate that ablating this single token significantly deteriorates performance on image understanding tasks. Furthermore, modifying this token enables effective steering of the image semantics, showing that targeted, local interventions can reliably control the model's global behavior.

Understanding the internal computations of large language models (LLMs) is crucial for aligning them with human values and preventing undesirable behaviors like toxic content generation. However, mechanistic interpretability is hindered by polysemanticity -- where individual neurons respond to multiple, unrelated concepts. While Sparse Autoencoders (SAEs) have attempted to disentangle these features through sparse dictionary learning, they have compromised LLM performance due to reliance on post-hoc reconstruction loss. To address this issue, we introduce Mixture of Monosemantic Experts for Transformers (Monet) architecture, which incorporates sparse dictionary learning directly into end-to-end Mixture-of-Experts pretraining. Our novel expert decomposition method enables scaling the expert count to 262,144 per layer while total parameters scale proportionally to the square root of the number of experts. Our analyses demonstrate mutual exclusivity of knowledge across experts and showcase the parametric knowledge encapsulated within individual experts. Moreover, Monet allows knowledge manipulation over domains, languages, and toxicity mitigation without degrading general performance. Our pursuit of transparent LLMs highlights the potential of scaling expert counts to enhance mechanistic interpretability and directly resect the internal knowledge to fundamentally adjust model behavior. The source code and pretrained checkpoints are available at https://github.com/dmis-lab/Monet.

Vision transformers (ViTs) can be trained using various learning paradigms, from fully supervised to self-supervised. Diverse training protocols often result in significantly different feature spaces, which are usually compared through alignment analysis. However, current alignment measures quantify this relationship in terms of a single scalar value, obscuring the distinctions between common and unique features in pairs of representations that share the same scalar alignment. We address this limitation by combining alignment analysis with concept discovery, which enables a breakdown of alignment into single concepts encoded in feature space. This fine-grained comparison reveals both universal and unique concepts across different representations, as well as the internal structure of concepts within each of them. Our methodological contributions address two key prerequisites for concept-based alignment: 1) For a description of the representation in terms of concepts that faithfully capture the geometry of the feature space, we define concepts as the most general structure they can possibly form - arbitrary manifolds, allowing hidden features to be described by their proximity to these manifolds. 2) To measure distances between concept proximity scores of two representations, we use a generalized Rand index and partition it for alignment between pairs of concepts. We confirm the superiority of our novel concept definition for alignment analysis over existing linear baselines in a sanity check. The concept-based alignment analysis of representations from four different ViTs reveals that increased supervision correlates with a reduction in the semantic structure of learned representations.

New footage of multiple eerie'triangle' craft flying above New Jersey has only compounded the mystery for locals. At least five or possibly six of the unidentified drones were captured in the new, 50-second cell phone video, which one commenter declared was'the clearest video yet.' One drone, heard roaring in the skies as it moved through the darkness, appeared to have a cluster of white lights on its underbelly and red lights blinking at the tips of its wings and tail. Another drone came into frame that resembled a classic'black triangle' UFO or the triangular TR-3B, which beamed bright white lights from its nose, wingtips and tail. Since mid-November, a wave of unexplained drone sightings above central Jersey has left both law enforcement and the general public watching the skies, hunting for clues on what these mysterious night flights might be.

Social robots are becoming more and more perceptible in public service settings. For engaging people in a natural environment a smooth social interaction as well as acceptance by the users are important issues for future successful Human-Robot Interaction (HRI). The type of verbal communication has a special significance here. In this paper we investigate the effects of spoken language varieties of a non-standard / regional language compared to standard language. More precisely we compare a human dialog with a humanoid social robot Pepper where the robot on the one hand is answering in High German and on the other hand in Low German, a regional language that is understood and partly still spoken in the northern parts of Germany. The content of what the robot says remains the same in both variants. We are interested in the effects that these two different ways of robot talk have on human interlocutors who are more or less familiar with Low German in terms of perceived warmth, competence and possible discomfort in conversation against a background of cultural identity. To measure these factors we use the Robotic Social Attributes Scale (RoSAS) on 17 participants with an age ranging from 19 to 61. Our results show that significantly higher warmth is perceived in the Low German version of the conversation.

How well can Multimodal Large Language Models (MLLMs) understand composite images? Composite images (CIs) are synthetic visuals created by merging multiple visual elements, such as charts, posters, or screenshots, rather than being captured directly by a camera. While CIs are prevalent in real-world applications, recent MLLM developments have primarily focused on interpreting natural images (NIs). Our research reveals that current MLLMs face significant challenges in accurately understanding CIs, often struggling to extract information or perform complex reasoning based on these images. We find that existing training data for CIs are mostly formatted for question-answer tasks (e.g., in datasets like ChartQA and ScienceQA), while high-quality image-caption datasets, critical for robust vision-language alignment, are only available for NIs. To bridge this gap, we introduce Composite Captions (CompCap), a flexible framework that leverages Large Language Models (LLMs) and automation tools to synthesize CIs with accurate and detailed captions. Using CompCap, we curate CompCap-118K, a dataset containing 118K image-caption pairs across six CI types. We validate the effectiveness of CompCap-118K by supervised fine-tuning MLLMs of three sizes: xGen-MM-inst.-4B and LLaVA-NeXT-Vicuna-7B/13B. Empirical results show that CompCap-118K significantly enhances MLLMs' understanding of CIs, yielding average gains of 1.7%, 2.0%, and 2.9% across eleven benchmarks, respectively.

For the first time, razor sharp teeth of tyrannosaurs were found in the Bexhill-on-Sea region of coastal East Sussex, England. These teeth indicate that a bevy of carnivorous dinosaurs including tyrannosaurs, spinosaurs, and members of the Velociraptor family stalked this region about 135 million years ago. The findings are detailed in a study published December 5 in the journal Papers in Palaeontology. Breakthroughs, discoveries, and DIY tips sent every weekday. By signing up you agree to our Terms of Service and Privacy Policy.