Large language model (LLM) watermarks enable authentication of text provenance, curb misuse of machine-generated text, and promote trust in AI systems. Current watermarks operate by changing the next-token predictions output by an LLM. The updated (i.e., watermarked) predictions depend on random side information produced, for example, by hashing previously generated tokens. LLM watermarking is particularly challenging when next-token predictions are near-deterministic. In fact, over 90% of next-token distributions are low-entropy, with more than half of the probability mass on a single token.

Entropy minimization (EM) trains the model to concentrate even more probability mass on its most confident outputs. We show that this simple objective alone, without any labeled data, can substantially improve large language models' (LLMs) performance on challenging math, physics, and coding tasks. We explore three approaches: (1) EM-FT minimizes token-level entropy similarly to instruction finetuning, but on unlabeled outputs drawn from the model; (2) EM-RL: reinforcement learning with negative entropy as the only reward to maximize; (3) EM-INF: inference-time logit adjustment to reduce entropy without any training data or parameter updates. On Qwen-7B, EM-RL, without any labeled data, achieves comparable or better performance than strong RL baselines such as GRPO [68] and RLOO [1] that are trained on 60K labeled examples. Furthermore, EM-INF enables Qwen-32B to match or exceed the performance of proprietary models like GPT-4o, Claude 3 Opus, and Gemini 1.5 Pro on the challenging SciCode benchmark [78], while being 3x more efficient than self-consistency and sequential refinement. Our findings reveal that many pretrained LLMs possess previously underappreciated reasoning capabilities that can be effectively elicited through entropy minimization alone, without any labeled data or even any parameter updates. 1

Plus: Gay bars in San Francisco using face scanners, France quits Palantir, Apple plans to change its private email, and more. Meta is testing face-recognition software built by the United States military and regional police department supplier Rank One, WIRED found in an investigation this week. Meta has been exploring the possibility of adding face recognition tech into its smart glasses, and WIRED previously reported that the app for the glasses contained code --now deleted--that would have enabled the company to activate face-recognition features on the devices. Anthropic is still negotiating with the Trump administration, after apparent White House concerns about the safety of new public model Claude Fable 5 resulted in Anthropic pulling the product off the market entirely. But security experts point out that AI models with advanced capabilities for discovering and exploiting software vulnerabilities--in other words, creating potentially dangerous hacking tools-- will be ubiquitous soon around the world .

Current Large Language Models (LLMs) often undergo supervised fine-tuning (SFT) to acquire tool use capabilities. However, SFT struggles to generalize to unfamiliar or complex tool use scenarios. Recent advancements in reinforcement learning (RL), particularly with R1-like models, have demonstrated promising reasoning and generalization abilities. Yet, reward design for tool use presents unique challenges: multiple tools may be invoked with diverse parameters, and coarse-grained reward signals, such as answer matching, fail to offer the finegrained feedback required for effective learning. In this work, we present the first comprehensive study on reward design for tool selection and application tasks within the RL paradigm. We systematically explore a wide range of reward strategies, analyzing their types, scales, granularity, and temporal dynamics. Building on these insights, we propose a principled reward design tailored for tool use tasks and apply it to train LLMs using RL methods. Empirical evaluations across diverse benchmarks demonstrate that our approach yields robust, scalable, and stable training, achieving a 17% improvement over base models and a 15% gain over SFT models.

What happens if pretending to be someone else becomes your entire life? It is a question at the heart of many of the biggest spy dramas, from Slow Horses to Black Doves - and it is one that TV thriller series The Agency explores more deeply than most. Returning for a second season, the Paramount+ thriller follows CIA operatives living under deep-cover identities. It examines not just the dangers of espionage, but the psychological cost of maintaining a lie for years. Starring Michael Fassbender, Richard Gere and Katherine Waterston, the series is based on acclaimed French drama The Bureau.

Retrieval-Augmented Generation (RAG) provides additional contextual knowledge to complement the parametric knowledge in Large Language Models (LLMs). These two knowledge interweave to enhance the accuracy and timeliness of LLM responses. However, the internal mechanisms by which LLMs utilize these knowledge remain unclear. We propose modeling the forward propagation of knowledge as an entity flow, employing this framework to trace LLMs' internal behaviors when processing mixed-source knowledge. Linear probing utilizes a trainable linear classifier to detect specific attributes in hidden layers.

Fitting Gaussian Processes (GPs) provides interpretable aleatoric uncertainty quantification for estimation of spatio-temporal fields. Spatio-temporal deep learning models, while scalable, typically assume a simplistic independent covariance matrix for the response, failing to capture the underlying correlation structure. However, spatio-temporal GPs suffer from issues of scalability and various forms of approximation bias resulting from restrictive assumptions of the covariance kernel function. We propose STACI, a novel framework consisting of a variational Bayesian neural network approximation of non-stationary spatio-temporal GP along with a novel spatio-temporal conformal inference algorithm. STACI is highly scalable, taking advantage of GPU training capabilities for neural network models, and provides statistically valid prediction intervals for uncertainty quantification. STACI outperforms competing GPs and deep methods in accurately approximating spatio-temporal processes and we show it easily scales to datasets with millions of observations.

Large language model (LLM) benchmarks inform LLM use decisions (e.g., "is this LLM safe to deploy for my use case and context?"). However, benchmarks may be rendered unreliable by various failure modes that impact benchmark bias, variance, coverage, or people's capacity to understand benchmark evidence. Using the National Institute of Standards and Technology's risk management process as a foundation, this research iteratively analyzed 26 popular benchmarks, identifying 57 potential failure modes and 196 corresponding mitigation strategies. The mitigations reduce failure likelihood and/or severity, providing a frame for evaluating "benchmark risk," which is scored to provide a metaevaluation benchmark: BenchRisk. Higher scores indicate that benchmark users are less likely to reach an incorrect or unsupported conclusion about an LLM. All 26 scored benchmarks present significant risk within one or more of the five scored dimensions (comprehensiveness, intelligibility, consistency, correctness, and longevity), which points to important open research directions for the field of LLM benchmarking. The BenchRisk workflow allows for comparison between benchmarks; as an open-source tool, it also facilitates the identification and sharing of risks and their mitigations.

Meritocratic systems, from admissions to hiring, aim to impartially reward skill and effort.

Amazon founder Jeff Bezos delivers a harsh prediction on the future of AI in the workforce, plus OpenAI faces an investigation into data handling and chatbot behavior.