As large language models become increasingly capable, there is growing concern that they may develop reasoning processes that are encoded or hidden from human oversight. To investigate whether current interpretability techniques can penetrate such encoded reasoning, we construct a controlled testbed by fine-tuning a reasoning model (DeepSeek-R1-Distill-Llama-70B) to perform chain-of-thought reasoning in ROT-13 encryption while maintaining intelligible English outputs. We evaluate mechanistic interpretability methods--in particular, logit lens analysis--on their ability to decode the model's hidden reasoning process using only internal activations. We show that logit lens can effectively translate encoded reasoning, with accuracy peaking in intermediate-to-late layers. Finally, we develop a fully unsupervised decoding pipeline that combines logit lens with automated paraphrasing, achieving substantial accuracy in reconstructing complete reasoning transcripts from internal model representations. These findings suggest that current mechanistic interpretability techniques may be more robust to simple forms of encoded reasoning than previously understood. Our work provides an initial framework for evaluating interpretability methods against models that reason in non-human-readable formats, contributing to the broader challenge of maintaining oversight over increasingly capable AI systems.

Simulating hybrid magnonic quantum systems remains a challenge due to the large disparity between the timescales of the two systems. We present a massively parallel GPU-based simulation framework that enables fully coupled, large-scale modeling of on-chip magnon-photon circuits. T o accelerate design workflows, we develop a physics-informed machine learning surrogate trained on the simulation data, reducing computational cost while maintaining accuracy. This combined approach reveals real-time energy exchange dynamics and reproduces key phenomena such as anti-crossing behavior and the suppression of ferromagnetic resonance under strong electromagnetic fields. By addressing the multiscale and multiphysics challenges in magnon-photon modeling, our framework enables scalable simulation and rapid prototyping of next-generation quantum and spintronic devices. 1 Introduction Hybrid quantum systems, which combine distinct physical platforms, are a promising route toward advanced quantum technologies, as they harness strong interactions that may not be readily achievable in a single platform [1, 2]. These systems take many forms, coupling any two (or more) quantum platforms -- for example, superconducting qubits [3, 4], microwave resonators [5], single spins [6], spin ensembles [4, 7-9], or mechanical resonators [10-12] -- to harness strong interactions. These heterogeneous systems leverage complementary advantages of each component, but their rich multi-physics interactions pose formidable modeling challenges. A prominent example is cavity magnonics, where collective spin excitations (magnons) couple with microwave photons in a resonant cavity to form hybrid magnon-polariton modes when tuned into resonance [13-15]. These states are essential for quantum operations such as mode swapping [16, 17], quantum state storage [4, 18, 19], and dynamic control of energy exchange [19, 20]. The hallmark experimental signature of strong magnon-photon coupling is a pronounced avoided crossing (mode splitting) in the frequency spectrum, in agreement with theoretical predictions [21] and observed in many 3D [13, 22] and on-chip 2D [7, 8, 23] cavity based systems.

Graph unlearning methods aim to efficiently remove the impact of sensitive data from trained GNNs without full retraining, assuming that deleted information cannot be recovered. In this work, we challenge this assumption by introducing the graph unlearning inversion attack: given only black-box access to an unlearned GNN and partial graph knowledge, can an adversary reconstruct the removed edges? We identify two key challenges: varying probability-similarity thresholds for unlearned versus retained edges, and the difficulty of locating unlearned edge endpoints, and address them with TrendAttack. First, we derive and exploit the confidence pitfall, a theoretical and empirical pattern showing that nodes adjacent to unlearned edges exhibit a large drop in model confidence. Second, we design an adaptive prediction mechanism that applies different similarity thresholds to unlearned and other membership edges. Our framework flexibly integrates existing membership inference techniques and extends them with trend features. Experiments on four real-world datasets demonstrate that TrendAttack significantly outperforms state-of-the-art GNN membership inference baselines, exposing a critical privacy vulnerability in current graph unlearning methods.

Things to Do in L.A. Tap to enable a layout that focuses on the article. Mail handlers use long tools to drag packages out of a bin onto a conveyer belt for sorting at the Los Angeles Processing & Distribution Center. This is read by an automated voice. Please report any issues or inconsistencies here . There is still time to get your packages and gifts shipped to friends and loved ones for the holiday season, but you need to hurry.

Tokyo Metropolitan Police stations receive around 4.4 million lost items annually. With the vision of creating a world where every lost item is found, Find Inc. runs Find (stylized as) a cloud-based lost-and-found service powered by proprietary AI technology. First introduced by Keio Corporation in 2023, the service is now used by 33 companies nationwide. We spoke with Ryu Wada, Director and COO of Find Inc., about how technology is transforming the future of lost items. Find Inc. is a Tokyo-based startup that provides a cloud service dedicated to lost items. The company was founded in December 2021 by Representative Director Akira Takashima and COO Wada.

The sidewalk outside Majed Zeidan's grocery store in San Francisco's Mission District has stayed filled with flowers, candles, memorials and pictures since his cat was crushed under a Waymo in late October. A month later, a Waymo reportedly crushed a dog. Amid the pictures of the cat, a visitor had placed a poster that said, "save the cat, kill the car". That's when Zeidan knew Kit Kat, his bodega cat, had become the face of the simmering discontent over San Francisco's growing number of self-driving cars. Residents became increasingly comfortable riding one, costumed Halloween parade goers clambered on its rooftops and danced, and pedestrians occasionally banged its bonnet to get it to give way to them.

The final language of the annual bill that funds the US military is in. It removes provisions that would have helped ensure service members' ability to fix their own equipment. US lawmakers have removed provisions in the National Defense Authorization Act for 2026 that would have ensured military members' right to repair their own equipment. The final language of the NDAA was shared by the House Armed Services Committee on Sunday, after weeks of delays pushed the annual funding bill to the end of the year. Among a host of other language changes made as part of reconciling different versions of the legislation drafted by the Senate and the House of Representatives, two provisions focused on the right to repair--Section 836 of the Senate bill and Section 863 of the House bill--have both been removed.

The companies, products and trends that fared the best over the last 12 months. Every December, the Engadget staff compiles a list of the year's biggest winners . We scour over articles from the previous 12 months to determine the people, companies, products and trends that made the most impact over the course of the year. Not all of that influence is positive, however, and some selections may also appear on our list of biggest losers. Still, sit back and enjoy our picks for the biggest winners of 2025.

What is in the 28-point US plan for Ukraine? 'Ukraine is running out of men, money and time' Can the US get all sides to end the war? Why is Europe opposing Trump's peace plan? Firefighters evacuated residents and their pets from a nine-storey apartment building in Ukraine's Sumy region after a Russian drone strike. The strikes come as Ukrainian President Volodymyr Zelenskyy met with leaders of the UK, France and Germany in London to discuss the US peace plan.

With artificial intelligence putting productivity on hyperspeed, the painstaking but often slow nature of dealing with other countries, as well as policymaking, is also forced to speed up. But a panel at the forefront of these changes at the BRIDGE Summit in Abu Dhabi--which convenes creators, policymakers, investors, technologists, media institutions, and cultural leaders around the world to discuss the future of media--said that breaking things fast is not without consequences. "Decision makers are being asked to make decisions very quickly on the basis of information that may not be verified or verifiable," Elizabeth Churchill, a professor of Human-Computer Interaction from the Mohamed Bin Zayed University of Artificial Intelligence, told moderator Nikhil Kumar, an executive editor at TIME, which is a media partner of the BRIDGE Summit. Churchill, who held senior roles in firms like Google and Yahoo, said she returned to academia to explore transparent and "interrogable" AI tools and content that is effectively watermarked--so that decision-makers know at a glance if information is trustworthy. She said current shortfalls in information quality are "very much a design problem that sits at the surface of all of the tools that we use and in diplomacy conversations many different people are using."