One of Meta's Offices Was Briefly Overtaken by a Rogue Squirrel The animal escaped after apparently arriving inside a package at Meta's Bangkok office, injuring one employee before finally being caught. Meta's year so far hasn't exactly been a picnic. A squirrel apparently got loose inside a building in Bangkok, Thailand, where some of the tech giant's regional teams are based. The critter spent at least 20 minutes darting past staff, according to an internal memo seen by WIRED. It noted that the squirrel minorly injured a janitor before finally being caught.

People visit the Delta Electronics booth during the annual Computex trade show in Taipei, Taiwan, on June 3, 2026. Thailand's stock market is having the best year among Southeast Asian peers, as investors discover an unlikely source of exposure to the global artificial-intelligence boom. Much of that gain has come from Delta Electronics (Thailand). The maker of power systems for AI data centers has surged more than 80% this year and became Thailand's first $100 billion company, large enough to be worth more than the next four largest Thai stocks combined. While the country lacks the semiconductor champions of Taiwan or South Korea, investors are increasingly recognizing its role in supplying the infrastructure behind AI. "Thailand isn't a pure AI market, but its exposure to data centers, electronics, power systems and digital infrastructure gives investors a new way to view Thai equities beyond the traditional tourism, banks and domestic consumption cycle," Bloomberg Intelligence Strategist Sufianti said in a note. Delta's rise is the clearest evidence of that shift.

Recent advances in Large Language Models (LLMs) have highlighted the critical importance of extending context length, yet the quadratic complexity of attention mechanisms poses significant challenges for efficient long-context modeling. KV cache compression has emerged as a key approach to address this challenge. Through extensive empirical analysis, we reveal a fundamental yet previously overlooked asymmetry in KV caches: while adjacent keys receive similar attention weights (local homogeneity), adjacent values demonstrate distinct heterogeneous distributions. This key-value asymmetry reveals a critical limitation in existing compression methods that treat keys and values uniformly. To address the limitation, we propose a training-free compression framework (AsymKV) that combines homogeneity-based key merging with a mathematically proven lossless value compression. Extensive experiments demonstrate that AsymKV consistently outperforms existing long-context methods across various tasks and base models.

In this work, we demonstrate that affine mappings between residual streams of language models is a cheap way to effectively transfer represented features between models. We apply this technique to transfer the weights of Sparse Autoencoders (SAEs) between models of different sizes to compare their representations. We find that small and large models learn similar representation spaces, which motivates training expensive components like SAEs on a smaller model and transferring to a larger model at a FLOPs savings. In particular, using a small-to-large transferred SAE as initialization can lead to 50% cheaper training runs when training SAEs on larger models. Next, we show that transferred probes and steering vectors can effectively recover ground truth performance. Finally, we dive deeper into feature-level transferability, finding that semantic and structural features transfer noticeably differently while specific classes of functional features have their roles faithfully mapped. Overall, our findings illustrate similarities and differences in the linear representation spaces of small and large models and demonstrate a method for improving the training efficiency of SAEs.

Follow this author to personalize your feed and get instant alerts. Follow Go to your personalized feed WHY FOLLOW? Smart Alerts: Get notified about major news as it happens. Back in 2017, a few days after turning 14, Jeeno Thitikul won the Ladies European Thailand Championship in her native Thailand as an amateur. Nearly a decade later, Thitikul returned in February as the world's top-ranked player and won her eighth LPGA title, and her first in her home country.

The real image, which the police station has since shared, shows the officers in normal clothes and no female officer in the picture at all. The real image, which the police station has since shared, shows the officers in normal clothes and no female officer in the picture at all. Picture was created by administrator in charge of station's Facebook account who wanted to create'friendlier image' It was an arresting image and an irresistible story. A group of tough Thai police officers - five men and one woman - all wearing elaborate festival-style dresses, surrounding a drug dealer they had caught while undercover. The image, released by local police, was so compelling that it found its way on to the front page of the UK's Daily Star, as well as in picture stories in the Telegraph, the Sun and the New York Post. The Sun wrote: "The burly crew of five men and one woman slipped into skin tight sequins and feathers for the covert mission in Thailand ."

Thailand risks losing ground to regional rivals such as Vietnam and Indonesia, which have moved more aggressively to streamline regulatory regimes and court foreign capital. Thai Prime Minister Anutin Charnvirakul's government plans a sweeping reform of more than 7,000 business regulations, aiming to cut bureaucratic hurdles and accelerate investment as it tries to compete for global capital. The planned rollback of ministerial rules and secondary regulations, many of which have accumulated into a significant burden on companies, marks a concerted push to reposition Thailand as a more competitive destination for multinational firms reconfiguring supply chains. The effort was detailed in a government statement Monday and comes as Thailand risks losing ground to regional rivals such as Vietnam and Indonesia, which have moved more aggressively to streamline regulatory regimes and court foreign capital. "Regulations intended to guide have, in practice, become costs," said government spokeswoman Rachada Dhnadirek.

Science Dinosaurs A new dinosaur dubbed the'Last Titan of Thailand' weighed more than 9 elephants Say hello to the'Nagatitan.' More information Adding us as a Preferred Source in Google by using this link indicates that you would like to see more of our content in Google News results. This sauropod lived in present-day Thailand during the Early Cretaceous period. Breakthroughs, discoveries, and DIY tips sent six days a week. Not long before an asteroid crashed into Earth and wiped out most of the dinosaurs, a long-necked dinosaur the size of nine adult Asian elephants may have been near a windy river peacefully eating plants.

Deep learning models for 12-lead electrocardiogram (ECG) analysis achieve high diagnostic performance but lack the intuitive interpretability required for clinical integration. Standard feature attribution methods are limited by the inherent difficulty in mapping abstract waveform fluctuations to physical anatomical pathologies. To resolve this, we propose a cross-modal method that projects feature attributions from high-performance 12-lead ECG models onto the CineECG 3D anatomical space. Our study reveals that while models trained directly on CineECG signals suffer from reduced accuracy and incoherent attributions, the proposed mapping mechanism effectively recovers clinically relevant feature rankings. Validated against a ground-truth dataset of 20 cases annotated by domain experts, the mapped explanations yield a Dice score of 0.56, significantly outperforming the 0.47 baseline of standard 12-lead attributions. These findings indicate that cross-modal averaging mapping effectively filters attribution instability and improves the localization of pathological features, combining the diagnostic expressiveness of standard ECG with the intuitive clarity of anatomical visualization.

Inference-time LLM alignment methods, particularly activation steering, offer an alternative to fine-tuning by directly modifying activations during generation. Existing methods, however, often rely on non-anticipative interventions that ignore how perturbations propagate through transformer layers and lack online error feedback, resulting in suboptimal, open-loop control. To address this, we show empirically that, despite the nonlinear structure of transformer blocks, layer-wise dynamics across multiple LLM architectures and scales are well-approximated by locally-linear models. Exploiting this property, we model LLM inference as a linear time-varying dynamical system and adapt the classical linear quadratic regulator to compute feedback controllers using layer-wise Jacobians, steering activations toward desired semantic setpoints in closed-loop with minimal computational overhead and no offline training. We also derive theoretical bounds on setpoint tracking error, enabling formal guarantees on steering performance. Using a novel adaptive semantic feature setpoint signal, our method yields robust, fine-grained behavior control across models, scales, and tasks, including state-of-the-art modulation of toxicity, truthfulness, refusal, and arbitrary concepts, surpassing baseline steering methods. Our code is available at: https://github.com/trustworthyrobotics/lqr-activation-steering