Oral assessments, more security checks and speedier marking are all on the cards as generative artificial intelligence (AI) could transform exams for the next generation of students. As the 2025 exam season drew to a close with GCSE students picking up their results on Thursday, after mostly sitting traditional pen and paper exams, AI is already changing the landscape. Exam preparation is undergoing a revolution, with students increasingly creating personal AI tutors, available around the clock to generate learning materials to suit individual needs that potentially lead to better results. "Using AI can give a student a much better understanding of a subject because they can ask those questions they wouldn't ask in class, or at odd hours, without being judged," said Dr Andrew Rogoyski of the Surrey Institute for People-Centred AI. "It really took off this summer," said Sandra Leaton Gray, a professor of education futures at University College London's Institute of Education. "So they're able to talk to it about the marking frameworks that are in use and upload those, and then they're able to do sample answers on their own. And then they're able to say to the AI: 'How would you improve the answer?' It's like having a tireless tutor."

Google has just released a report detailing how much energy its Gemini apps use for each query. In total, the median prompt--one that falls in the middle of the range of energy demand--consumes 0.24 watt-hours of electricity, the equivalent of running a standard microwave for about one second. The company also provided average estimates for the water consumption (five drops per query) and carbon emissions associated with a text prompt to Gemini. It's the most transparent estimate yet from a Big Tech company with a popular AI product, and the report includes detailed information about how the company calculated its final estimate. Earlier this year, MIT Technology Review published a comprehensive series on AI and energy, at which time none of the major AI companies would reveal their per-prompt energy usage.

They make counterfactual predictions by using observed actions to learn the underlying utility function (a.k.a.

Children are remarkably fast in learning new skills; How do they do this?

The framework of feedback graphs is a generalization of sequential decision-making with bandit or full information feedback. In this work, we study an extension where the directed feedback graph is stochastic, following a distribution similar to the classical Erdős-Rényi model. Specifically, in each round every edge in the graph is either realized or not with a distinct probability for each edge.

The single-step formulation does not account for changes in the student's internal state over In the multi-step formulation, effort put towards studying accumulates in the form of knowledge. We demonstrate this by revisiting the classroom example. The student's grade is then the summation of all scores across time. B.1 Agent's best-response effort sequence A rational agent solves the following optimization to determine his best-response effort policy: { e Recall that the agent's score A dominated effort policy is formally defined as follows: Lemma C.1 Next we look at the complementary slackness condition. From Lemma D.1, we know the form a rational agent's effort Substituting this into Equation 6, we obtain the following characterization of the principal's assessment policy: { E.1 The set of incentivizable effort policies is convex Proof.

A recent line of research investigates how strategic agents may respond to such scoring tools to receive favorable assessments. While prior work has focused on the short-term strategic interactions between a decision-making institution (modeled as a principal) and individual decision-subjects (modeled as agents), we investigate interactions spanning multiple time-steps . In particular, we consider settings in which the agent's effort investment

Equipped with this algorithm, we demonstrate reincarnating RL's gains

Quantile (and, more generally, KL) regret bounds, such as those achieved by NormalHedge (Chaudhuri, Freund, and Hsu 2009) and its variants, relax the goal of competing against the best individual expert to only competing against a majority of experts on adversarial data.