In the face of sea ice loss, some of the bears on the Norwegian archipelago are gaining weight. Three polar bear cubs gather around their tranquilized mother. She had a litter of three cubs (an unusual brood size) and the smallest cub only weighed 11 pounds (five kilograms). Breakthroughs, discoveries, and DIY tips sent six days a week. The Arctic's polar bears () are often the poster species for the perils of climate change .

Rare polar bear adoption could save cub's life The cubs were born into a well-studied'celebration' of polar bears in Canada. Breakthroughs, discoveries, and DIY tips sent every weekday. Scientists in Churchill, Manitoba, Canada (aka the polar bear capital of the world) have confirmed that a wild female polar bear has adopted a cub that is not her own. This rare behavior was captured on cameras during the polar bear's annual migration along Western Hudson Bay . Researchers from Environment and Climate Change Canada and Polar Bears International spotted the mother bear (designated as bear X33991) during spring 2025, when she came out of her maternity den.

In this work, we unveil and study idiosyncrasies in Large Language Models (LLMs) -- unique patterns in their outputs that can be used to distinguish the models. To do so, we consider a simple classification task: given a particular text output, the objective is to predict the source LLM that generates the text. We evaluate this synthetic task across various groups of LLMs and find that simply fine-tuning existing text embedding models on LLM-generated texts yields excellent classification accuracy. Notably, we achieve 97.1% accuracy on held-out validation data in the five-way classification problem involving ChatGPT, Claude, Grok, Gemini, and DeepSeek. Our further investigation reveals that these idiosyncrasies are rooted in word-level distributions. These patterns persist even when the texts are rewritten, translated, or summarized by an external LLM, suggesting that they are also encoded in the semantic content. Additionally, we leverage LLM as judges to generate detailed, open-ended descriptions of each model's idiosyncrasies. Finally, we discuss the broader implications of our findings, particularly for training on synthetic data and inferring model similarity. Code is available at https://github.com/locuslab/llm-idiosyncrasies.

The use of large language models (LLMs) as evaluators has garnered significant attention due to their potential to rival human-level evaluations in long-form response assessments. However, current LLM evaluators rely heavily on static, human-defined criteria, limiting their ability to generalize across diverse generative tasks and incorporate context-specific knowledge. In this paper, we propose a novel Self-Assessing LLM framework that integrates Context-Aware Criteria (SALC) with dynamic knowledge tailored to each evaluation instance. This instance-level knowledge enhances the LLM evaluator's performance by providing relevant and context-aware insights that pinpoint the important criteria specific to the current instance. Additionally, the proposed framework adapts seamlessly to various tasks without relying on predefined human criteria, offering a more flexible evaluation approach. Empirical evaluations demonstrate that our approach significantly outperforms existing baseline evaluation frameworks, yielding improvements on average 4.8% across a wide variety of datasets. Furthermore, by leveraging knowledge distillation techniques, we fine-tuned smaller language models for criteria generation and evaluation, achieving comparable or superior performance to larger models with much lower cost. Our method also exhibits a improvement in LC Win-Rate in AlpacaEval2 leaderboard up to a 12% when employed for preference data generation in Direct Preference Optimization (DPO), underscoring its efficacy as a robust and scalable evaluation framework.

But have you ever wondered what a fight between a hippopotamus and a great white shark might look like? Now, scientists have set the record straight, after using artificial intelligence (AI) to simulate battles between the most terrifying animals on Earth. Somewhat surprisingly, the simulations suggest that a hippo would beat a great white shark - and could even take down a polar bear. However, the ultimate champion of the animal kingdom is the African Elephant, according to researchers from Animal Matchup. In honour of World Animal Day, experts from Animal Match set out to settle the debate - which animal is the strongest?

Polar bears and brown bears were still mating with each other long after they had split into two distinct species, a new study has found. The two species are known to have separated up to 1.6 million years ago, yet new genomic evidence suggests they have inherited traits from each other much more recently. Scientists from the USA, Mexico and Finland analysed the genomes of 64 modern polar and brown bears, as well as that of an ancient polar bear that lived up to 130,000 years ago. While evidence of evidence of hybridisation was found in both brown and polar bear genomes, the latter carried a particularly strong signature of DNA from brown bears. As global warming continued to melt Arctic sea ice, the two bear species may run into each other more frequently, their shared evolutionary history could become more significant.

A new artificial intelligence (AI) system is about to be used to predict ice loss in the Arctic, a study reveals. The deep learning tool, called IceNet was created by scientists at the British Antarctic Survey (BAS) and has been trained with the past four decades of satellite data from the region. It's almost 95 per cent accurate in predicting whether sea ice will be present two months ahead – better than the leading physics-based model previously used by BAS – but it's been trained to predict as far as six months ahead. Sea ice in both the north and south poles naturally expands in the winter and shrinks in the summer. But sea ice is very hard to predict because it has'very complex interactions' with the atmosphere above and the ocean below.

Genghis Khan got his dying wish: despite attempts by archaeologists and scientists to find the Mongolian ruler's final resting place, the location remains a secret 800 years after his death. The search for his tomb, though, has inspired an innovative project that could help protect polar bears. "I randomly tuned into the radio one night and heard an expert talking about the use of synthetic aperture radar [SAR] to look for Genghis Khan's tomb," says Tom Smith, associate professor in plant and wildlife sciences at Brigham Young University (BYU) in Utah. "They were using SAR to penetrate layers of forest canopy in upper Mongolia, looking for the ruins of a burial structure." Talking to engineers, including BYU's Dr David Long, Smith learned that SAR is used by the military to detect enemy camps, tanks and vehicles hidden beneath camouflage and is being studied as a potential tool for finding avalanche survivors.

The "Planet Earth Report" connects you to headline news on the science, technology, discoveries, people and events changing our planet and the future of the human species. We have a new global tally of the insect apocalypse. "Scary Known Unknown" –A Vast Hidden Asteroid Population Close to Sun Elizabeth Warren wants to ban the US from using nuclear weapons first –This 12-word bill could change how we use nuclear weapons. Bill Gates tweeted out a chart and sparked a huge debate about global poverty–Has global poverty declined dramatically? Intelligent Machines –Trump has a plan to keep America first in artificial intelligence.

Scarcely a day goes by without another headline about neural networks: some new task that deep learning algorithms can excel at, approaching or even surpassing human competence. As the application of this approach to computer vision has continued to improve, with algorithms capable of specialized recognition tasks like those found in medicine, the software is getting closer to widespread commercial use--for example, in self-driving cars. Our ability to recognize patterns is a huge part of human intelligence: if this can be done faster by machines, the consequences will be profound. Yet, as ever with algorithms, there are deep concerns about their reliability, especially when we don't know precisely how they work. State-of-the-art neural networks will confidently--and incorrectly--classify images that look like television static or abstract art as real-world objects like school-buses or armadillos. Specific algorithms could be targeted by "adversarial examples," where adding an imperceptible amount of noise to an image can cause an algorithm to completely mistake one object for another.