Researchers concluded that LLMs were able to spot the structure of a pun but did not really get the joke. Researchers concluded that LLMs were able to spot the structure of a pun but did not really get the joke. Can't tech a joke: AI does not understand puns, study finds Researchers say results underline large language models' poor grasp of humour, empathy and cultural nuance Comedians who rely on clever wordplay and writers of witty headlines can rest a little easier, for the moment at least, research on AI suggests. Experts from universities in the UK and Italy have been investigating whether large language models (LLMs) understand puns - and found them wanting. The team from Cardiff University, in south Wales, and Ca' Foscari University of Venice concluded that LLMs were able to spot the structure of a pun but did not really get the joke.

The Trump administration's pressure on European regulators is having an impact, with fewer restrictions on Big Tech and canceled measures. Almost everything is on hiatus. The EU AI Act, Digital Services Act, and Digital Markets Act are all at risk. The European Commission is preparing to end the year with virtually no movement on its most important tech policy initiatives. Many measures may even be reversed.

It all seems to be falling apart now! Cunning new tactic women are using to cheat. Trump delivers savage parting shot to'lowlifes' MTG and Thomas Massie while declaring GOP has'never been so united' Gavin Newsom's inner circle on edge as multiple aides receive ominous letter from FBI just days after California governor's chief of staff was indicted Experts discover there are EIGHT different types of long Covid... do you have any of them? Full House's Jodie Sweetin reveals how addiction struggle began at 14 at costar Candace Cameron Bure's wedding Fans turn on RichTok influencer Becca Bloom over shocking comments... as she makes stunning admission about her marriage and her wild extravagance is revealed Morgan was searching for her soulmate in church... then she uncovered the sinister underbelly of Christian dating in MAGA America. Rich moms of Manhattan go to WAR: Innocent comment plunges gilded zip code into anarchy... and everyone's looking over their shoulder Two Texas men's twisted fantasy to recruit homeless to invade remote island, kill its inhabitants and ravage their women WANTED: One VERY tolerant Lady! Picky aristocrat, 79, launches bid to find a wife.

Oscar-winning actress Cate Blanchett and former prime minister Baroness Theresa May are among the six public figures who will guest edit BBC Radio 4's Today programme over the Christmas period. Broadcaster Melvyn Bragg, historian and podcaster Tom Holland, inventor Sir James Dyson and Microsoft's head of artificial intelligence (AI) Mustafa Suleyman will also guest edit shows between 24 December and 31 December. For the past 22 years, the news programme has handed over the editorial reins to guest editors during the festive period. Owenna Griffiths, editor of Today, said: In a rapidly changing world, this year's guest editors will help bring illumination and understanding. She added: Every Christmas on Today, a new set of guest editors take up residence and bring with them a wonderful range of new stories, fresh ideas and, hopefully, a sprinkling of joy.

Is the fall of Pokrovsk inevitable? Is Trump losing patience with Putin? Here's where things stand on Monday, November 24. United States Secretary of State Marco Rubio told reporters in Geneva that "a tremendous amount of progress" was made during talks in the Swiss city on Sunday and that he was "very optimistic" that an agreement could be reached in "a very reasonable period of time, very soon". Rubio also said that specific areas still being worked on from a 28-point peace plan for Ukraine, championed by US President Donald Trump, included the role of NATO and security guarantees for Ukraine.

Machu Picchu, the remains of a 15th Century Inca city, is Peru's most popular tourist destination, and a Unesco world heritage site. Yet a continuing dispute over the buses that take visitors up to the mountain-top site recently saw some 1,400 stranded tourists needing to be evacuated. Cristian Alberto Caballero Chacón is head of operations for bus company Consettur, which for the past 30 years has transported some 4,500 people every day to Machu Picchu from the local town of Aguas Calientes. It is a 20-minute journey, and the only alternative is an arduous, steep, two-hour walk. He admits that in the past few months there have been some conflicts between people from different communities here.

We consider the adaptive control problem for discrete-time, nonlinear stochastic systems with linearly parameterised uncertainty. Assuming access to a parameterised family of controllers that can stabilise the system in a bounded set within an informative region of the state space when the parameter is well-chosen, we propose a certainty equivalence learning-based adaptive control strategy, and subsequently derive stability bounds on the closed-loop system that hold for some probabilities. We then show that if the entire state space is informative, and the family of controllers is globally stabilising with appropriately chosen parameters, high probability stability guarantees can be derived.

We investigate whether socio-economic indicators like household wealth leave recoverable imprints in satellite imagery (capturing physical features) and Internet-sourced text (reflecting historical/economic narratives). Using Demographic and Health Survey (DHS) data from African neighborhoods, we pair Landsat images with LLM-generated textual descriptions conditioned on location/year and text retrieved by an AI search agent from web sources. We develop a multimodal framework predicting household wealth (International Wealth Index) through five pipelines: (i) vision model on satellite images, (ii) LLM using only location/year, (iii) AI agent searching/synthesizing web text, (iv) joint image-text encoder, (v) ensemble of all signals. Our framework yields three contributions. First, fusing vision and agent/LLM text outperforms vision-only baselines in wealth prediction (e.g., R-squared of 0.77 vs. 0.63 on out-of-sample splits), with LLM-internal knowledge proving more effective than agent-retrieved text, improving robustness to out-of-country and out-of-time generalization. Second, we find partial representational convergence: fused embeddings from vision/language modalities correlate moderately (median cosine similarity of 0.60 after alignment), suggesting a shared latent code of material well-being while retaining complementary details, consistent with the Platonic Representation Hypothesis. Although LLM-only text outperforms agent-retrieved data, challenging our Agent-Induced Novelty Hypothesis, modest gains from combining agent data in some splits weakly support the notion that agent-gathered information introduces unique representational structures not fully captured by static LLM knowledge. Third, we release a large-scale multimodal dataset comprising more than 60,000 DHS clusters linked to satellite images, LLM-generated descriptions, and agent-retrieved texts.

In dynamic noncooperative games, each player makes conjectures about other players' reactions before choosing a strategy. However, resulting equilibria may be multiple and do not always lead to desirable outcomes. These issues are typically addressed separately, for example, through opponent modelling and incentive design. Drawing inspiration from conjectural variations games, we propose an incentive design framework in which a coordinator first computes an equilibrium by optimizing a predefined objective function, then communicates this equilibrium as a target for the players to reach. In a centralized setting, the coordinator also optimizes the conjectures to steer the players towards the target. In decentralized settings, players independently compute conjectures and update their strategies based on individual targets. We provide a guarantee of equilibrium existence in both cases. This framework uses conjectures not only to guide the system towards desirable outcomes but also to decouple the game into independent optimization problems, enabling efficient computation and parallelization in large-scale settings. We illustrate our theoretical results on classical representative noncooperative games, demonstrating its application potential.

As Large Language Models (LLMs) evolve into personal assistants with access to sensitive user data, they face a critical privacy challenge: while prior work has addressed output-level privacy, recent findings reveal that LLMs often leak private information through their internal reasoning processes, violating contextual privacy expectations. These leaky thoughts occur when models inadvertently expose sensitive details in their reasoning traces, even when final outputs appear safe. The challenge lies in preventing such leakage without compromising the model's reasoning capabilities, requiring a delicate balance between privacy and utility. We introduce Steering Activations towards Leakage-free Thinking (SALT), a lightweight test-time intervention that mitigates privacy leakage in model's Chain of Thought (CoT) by injecting targeted steering vectors into hidden state. We identify the high-leakage layers responsible for this behavior. Through experiments across multiple LLMs, we demonstrate that SALT achieves reductions including $18.2\%$ reduction in CPL on QwQ-32B, $17.9\%$ reduction in CPL on Llama-3.1-8B, and $31.2\%$ reduction in CPL on Deepseek in contextual privacy leakage dataset AirGapAgent-R while maintaining comparable task performance and utility. Our work establishes SALT as a practical approach for test-time privacy protection in reasoning-capable language models, offering a path toward safer deployment of LLM-based personal agents.