Different distribution shifts require different algorithmic and operational interventions. Methodological research must be grounded by the specific shifts they address.

Modern large language models (LLMs) such as GPT, Claude, and Gemini have transformed the way we learn, work, and communicate. Y et, their ability to produce highly human-like text raises serious concerns about misinformation and academic integrity, making it an urgent need for reliable algorithms to detect LLMgenerated content. In this paper, we start by presenting a geometric approach to demystify rewrite-based detection algorithms, revealing their underlying rationale and demonstrating their generalization ability. Building on this insight, we introduce a novel rewrite-based detection algorithm that adaptively learns the distance between the original and rewritten text. Theoretically, we demonstrate that employing an adaptively learned distance function is more effective for detection than using a fixed distance. Empirically, we conduct extensive experiments with over 100 settings, and find that our approach demonstrates superior performance over baseline algorithms in the majority of scenarios. In particular, it achieves relative improvements from 57.8% to 80.6% over the strongest baseline across different target LLMs (e.g., GPT, Claude, and Gemini). The past few years have witnessed the emergence and rapid development of large language models (LLMs) such as GPT (Hurst et al., 2024), DeepSeek (Liu et al., 2024), Claude (Anthropic, 2024), Gemini (Comanici et al., 2025), Grok (xAI, 2025) and Qwen (Y ang et al., 2025). Their impact is everywhere, from education, academia and software development to healthcare and everyday life (Arora & Arora, 2023; Chan & Hu, 2023; Hou et al., 2024). On one side of the coin, LLMs can support users with conversational question answering, help students learn more effectively, draft emails, write computer code, prepare presentation slides and more. On the other side, their ability to closely mimic human-written text also raises serious concerns, including the generation of biased or harmful content, the spread of misinformation in the news ecosystem, and the challenges related to authorship attribution and intellectual property (Dave et al., 2023; Fang et al., 2024; Messeri & Crockett, 2024; Mahajan et al., 2025; Laurito et al., 2025). Addressing these concerns requires effective algorithms to distinguish between human-written and LLM-generated text, which has become an active and popular research direction in recent literature (see Crothers et al., 2023; Wu et al., 2025, for reviews).

Little girl's appalling question to nanny who was having affair with her father hours after mother's brutal murder New Idaho murder photos lay bare the humiliating truth about arrogant Bryan Kohberger's pathetic attempt to plot'the perfect crime' Why'controlling' Nicola Peltz also made an enemy of the Hadids: Before Brooklyn, she dated Anwar. Now family insiders reveal what made her'FLIP'... and humiliating comment they still whisper about her Hoda Kotb mercilessly mocked by NBC staff: Insiders slam her as'perpetual pest' they'just want to go away'... as her'exhausting' demands are laid bare Prince Harry says British troops who died in Afghanistan deserve'respect' in backlash against Donald Trump's jibe at UK's war dead The 12 cities that will see'catastrophic' damage by crippling winter storm MAGA supporters slam Today show after Dylan Dreyer makes on-air slip up during weather forecast: 'Did y'all hear that?' Yankees icon Derek Jeter reveals what he misses most about New York after selling $6million castle... as he gives rare glimpse into family life Meghan Trainor's teary photo with her new baby born via surrogate has sparked an almost unsayable thought. Most women won't admit it... but I will: CAROLINE BULLOCK DJ Fat Tony now reveals Nicola Peltz's entire family stormed out of wedding after THAT dance and how Victoria Beckham draped her arms around Brooklyn American Idol star Nutsa Buzaladze resurfaces with'unbelievable' nose job - see her now Boy, 5, filmed being snatched off Minnesota street by ICE is now a THOUSAND miles from home... as family deny JD Vance's claim that father abandoned him These stars insist secret notes and bizarre daily mutterings made them famous. Here's science that proves manifesting is real... and how you can do it too America's top celebrities are often asked about the secret to their success, and many have honestly claimed that the practice of'manifestation' turned their wildest dreams into reality. A-listers including Oprah Winfrey, Ariana Grande, Will Smith and Arnold Schwarzenegger have all said they essentially imagined what they desired most and were able to achieve it solely through positive thinking and focused goal-setting.

Is it illegal to own an axolotl? A recent pet seizure at Chicago's O'Hare Airport illustrates ongoing confusion. Many pet axolotls are crossbred with other salamanders to create their unique coloration. Breakthroughs, discoveries, and DIY tips sent six days a week. The axolotl () is a confusing creature, and not simply because it looks like a real-life Pokémon .

January 3 marked the return of US military intervention in Latin America. While the events unfolded between Caracas and Brooklyn, social networks had already fabricated their own reality. A fire is seen in the distance at Fort Tiuna, Venezuela's largest military complex, following a series of explosions in Caracas on January 3, 2026. Geopolitics are being reduced to videos lasting just a few minutes. Social media has surpassed traditional media, not only in the speed with which it is created and shared, but also in its ability to frame our reality. People have the illusion of knowing what is happening and why within just a few hours--or less--of major world events. But reality is more complicated.

Text-based causal inference increasingly employs textual data as proxies for unobserved confounders, yet this approach introduces a previously undertheorized source of bias: treatment leakage. Treatment leakage occurs when text intended to capture confounding information also contains signals predictive of treatment status, thereby inducing post-treatment bias in causal estimates. Critically, this problem can arise even when documents precede treatment assignment, as authors may employ future-referencing language that anticipates subsequent interventions. Despite growing recognition of this issue, no systematic methods exist for identifying and mitigating treatment leakage in text-as-confounder applications. This paper addresses this gap through three contributions. First, we provide formal statistical and set-theoretic definitions of treatment leakage that clarify when and why bias occurs. Second, we propose four text distillation methods -- similarity-based passage removal, distant supervision classification, salient feature removal, and iterative nullspace projection -- designed to eliminate treatment-predictive content while preserving confounder information. Third, we validate these methods through simulations using synthetic text and an empirical application examining International Monetary Fund structural adjustment programs and child mortality. Our findings indicate that moderate distillation optimally balances bias reduction against confounder retention, whereas overly stringent approaches degrade estimate precision.

Ensuring that deep learning models are well-calibrated in terms of their predictive uncertainty is essential in maintaining their trustworthiness and reliability, yet despite increasing advances in foundation model research, the relationship between such large language models (LLMs) and their calibration remains an open area of research. In this work, we look at a critical gap in the calibration of LLMs within multilingual settings, in an attempt to better understand how the data scarcity can potentially lead to different calibration effects and how commonly used techniques can apply in these settings. Our analysis on two multilingual benchmarks, over 29 and 42 languages respectively, reveals that even in low-resource languages, model confidence can increase significantly after instruction-tuning on high-resource language SFT datasets. However, improvements in accuracy are marginal or non-existent, resulting in mis-calibration, highlighting a critical shortcoming of standard SFT for multilingual languages. Furthermore, we observe that the use of label smoothing to be a reasonable method alleviate this concern, again without any need for low-resource SFT data, maintaining better calibration across all languages. Overall, this highlights the importance of multilingual considerations for both training and tuning LLMs in order to improve their reliability and fairness in downstream use.