The flexibility of a power generation unit determines how quickly and often it can ramp up or down. In energy models, it depends on assumptions on the technical characteristics of the unit, such as its installed capacity or turbine technology. In this paper, we learn the empirical flexibility of gas units from their electricity generation, revealing how real-world limitations can lead to substantial differences between units with similar technical characteristics. Using a novel deep clustering approach, we transform 5 years (2019-2023) of unit-level hourly generation data for 49 German units from 100 MWp of installed capacity into low-dimensional embeddings. Our unsupervised approach identifies two clusters of peaker units (high flexibility) and two clusters of non-peaker units (low flexibility). The estimated ramp rates of non-peakers, which constitute half of the sample, display a low empirical flexibility, comparable to coal units. Non-peakers, predominantly owned by industry and municipal utilities, show limited response to low residual load and negative prices, generating on average 1.3 GWh during those hours. As the transition to renewables increases market variability, regulatory changes will be needed to unlock this flexibility potential.

Comparison of fine-tuning models with reinforcement learning (RL) and supervised fine-tuning (SFT) reveals that, despite similar performance at a new task, RL preserves prior knowledge and capabilities significantly better. We find that the degree of forgetting is determined by the distributional shift, measured as the KL-divergence between the fine-tuned and base policy evaluated on the new task. Our analysis reveals that on-policy RL is implicitly biased towards KL-minimal solutions among the many that solve the new task, whereas SFT can converge to distributions arbitrarily far from the base model. We validate these findings through experiments with large language models and robotic foundation models and further provide theoretical justification for why on-policy RL updates lead to a smaller KL change. We term this principle $\textit{RL's Razor}$: among all ways to solve a new task, RL prefers those closest in KL to the original model.

We investigate the mistake-bound model of online learning with caps on the number of arithmetic operations per round. We prove general bounds on the minimum number of arithmetic operations per round that are necessary to learn an arbitrary family of functions with finitely many mistakes. We solve a problem on agnostic mistake-bounded online learning with bandit feedback from (Filmus et al, 2024) and (Geneson \& Tang, 2024). We also extend this result to the setting of operation caps.

In applications across agriculture, ecology, and human development, machine learning with satellite imagery (SatML) is limited by the sparsity of labeled training data. While satellite data cover the globe, labeled training datasets for SatML are often small, spatially clustered, and collected for other purposes (e.g., administrative surveys or field measurements). Despite the pervasiveness of this issue in practice, past SatML research has largely focused on new model architectures and training algorithms to handle scarce training data, rather than modeling data conditions directly. This leaves scientists and policymakers who wish to use SatML for large-scale monitoring uncertain about whether and how to collect additional data to maximize performance. Here, we present the first problem formulation for the optimization of spatial training data in the presence of heterogeneous data collection costs and realistic budget constraints, as well as novel methods for addressing this problem. In experiments simulating different problem settings across three continents and four tasks, our strategies reveal substantial gains from sample optimization. Further experiments delineate settings for which optimized sampling is particularly effective. The problem formulation and methods we introduce are designed to generalize across application domains for SatML; we put special emphasis on a specific problem setting where our coauthors can immediately use our findings to augment clustered agricultural surveys for SatML monitoring in Togo.

Despite advances in improving large language model (LLM) to refuse to answer malicious instructions, widely used LLMs remain vulnerable to jailbreak attacks where attackers generate instructions with distributions differing from safety alignment corpora. New attacks expose LLMs' inability to recognize unseen malicious instructions, highlighting a critical distributional mismatch between training data and real-world attacks that forces developers into reactive patching cycles. To tackle this challenge, we propose IMAGINE, a synthesis framework that leverages embedding space distribution analysis to generate jailbreak-like instructions. This approach effectively fills the distributional gap between authentic jailbreak patterns and safety alignment corpora. IMAGINE follows an iterative optimization process that dynamically evolves text generation distributions across iterations, thereby augmenting the coverage of safety alignment data distributions through synthesized data examples. Based on the safety-aligned corpus enhanced through IMAGINE, our framework demonstrates significant decreases in attack success rate on Qwen2.5, Llama3.1, and Llama3.2 without compromising their utility.

First Lady Melania Trump made a rare public appearance at the White House on Thursday, telling the crowd that "the robots are here" and that it is "our responsibility to prepare America's children" for the AI-driven decades ahead. "Our future is no longer science fiction," she said. "During this primitive stage, it is our duty to treat AI as we would our own children – empowering, but with watchful guidance." The event - a meeting of a White House AI education task force established earlier this year - was one of so far only a handful of public events for a first lady who has proven both elusive and influential since her husband returned to the White House. Born Melanija Knavs in Slovenia, the 55-year-old first lady and former fashion model was, for a time, often described as an "enigma" - less public than her predecessors, with fewer speeches and public engagements. Her relative absence during long stretches of her husband's ultimately successful 2024 campaign even prompted a flurry of news articles with headlines asking "Where is Melania?".

The United States Patent and Trademark Office has rejected Neuralink's attempt to trademark the product names Telepathy and Telekinesis, citing pending applications by another person for the same trademarks. Neuralink, the brain implant company co-founded by Elon Musk, filed to trademark the names in March. But in letters sent to Neuralink in August, the trademark office is refusing to allow the applications to move forward. It says Wesley Berry, a computer scientist and co-founder of tech startup Prophetic, previously filed trademark applications for Telepathy in May 2023 and Telekinesis in August 2024. Prophetic is building a wearable headset to induce lucid dreaming, but only Berry is the author of the trademark applications, not Prophetic.

First lady Melania Trump is hosting an artificial intelligence meeting with top industry leaders, including Google CEO Sundar Pichai, Thursday, as she stresses the importance of managing AI's growth "responsibly." The White House Task Force on Artificial Intelligence Education will meet for the second time in the East Room of the White House Thursday afternoon. The first lady will host the meeting alongside members of the task force and private sector leaders. "I predict AI will represent the single largest growth category in our nation during the Trump Administration -- and I won't be surprised if AI becomes known as the greatest engine of progress in the history of the United States of America," the first lady said. First lady Melania Trump is hosting an artificial intelligence meeting with top industry leaders, as she stresses the importance of managing AI's growth "responsibly."

Fox News contributor Jonathan Morris and Pastor Robert Jeffress react to the president unveiling new guidance on public school prayer. The battle over prayer in school is raging in Texas right now, with Attorney General Ken Paxton vowing to defend any school district that introduces the controversial practice under a recent state law expanding religious expression in education. For the entirety of my life, and I'm old, the prohibition on public school-sponsored prayer seemed like settled Constitutional science, owing to a 1962 Supreme Court decision barring what had previously been a widespread and normal practice. In the past, I agreed with this form of separation of church and state. For me it was almost a question of better safe than sorry regarding the rights of minority religions, and importantly, I believed that Christian moral values were so ingrained in our culture that 30 seconds a day of praying could be forsaken.

In response to the resignation, a spokesperson from the Department for Science, Innovation and Technology said: "The technology secretary has been clear on the need for the institute to deliver value for money and maximum impact for taxpayers, and we will continue our work to support that ambition." Staff who wrote the whistleblowing complaint have told the BBC Dr Innes' resignation was the "first step". "With the rest of our 100m public funding still at stake, the priority now is to ensure the leadership overhaul that should follow - board and executive alike - can command the confidence of staff, government, regulators and, most importantly, the nation," they said. The Turing Institute said its board was now looking to appoint a new CEO who will oversee "the next phase" to "step up its work on defence, national security and sovereign capabilities". Its work had once focused on AI and data science research in environmental sustainability, health and national security, but moved on to other areas such as responsible AI.