PAPER PLAINE

When people learn from each other, they naturally pay more attention to confident sources—but this can backfire. A mathematical model of learning agents reveals a dangerous pattern: moderate information sharing helps groups correct early mistakes, but strong social influence can trap entire populations in a shared wrong belief. The model also identifies a paradox: confidence helps people learn better when making private decisions, yet the same confidence makes false ideas spread faster among groups.

Faces we find attractive are easier for our brains to process — they match statistical patterns our visual system encounters most often. Researchers trained artificial neural networks on face images without any beauty labels, then checked whether the networks' internal representations matched human attractiveness ratings. Across multiple datasets, faces rated as attractive aligned with directions in the network's latent space that required the least computational effort to encode, and these "beauty directions" emerged consistently even when networks were retrained from scratch.

Vision-language models excel at analyzing individual images and frames but largely fail to detect when video sequences violate temporal logic — such as when consecutive frames are swapped. Researchers created TimeCatch, a benchmark using simple anomalies like frame swaps and noise insertions, and found that while these AI systems spot obvious corruptions within single frames, they perform near chance-level when asked to notice temporal inconsistencies that humans catch easily.

A new method called PAINT converts standard Landsat satellite images into much richer hyperspectral images—expanding from 7 color bands to 172—while avoiding the massive computational burden that usually makes this task impractical. When applied to land classification tasks, the method improved accuracy from 79% to 92%, demonstrating that these AI-enhanced images capture ground details far better than the originals.

When multiple leveraged exchange-traded funds trade related assets, they create hidden feedback loops that single-product monitoring completely misses. Researchers found that watching each fund individually reports 'safe' while the full system is actually unstable, and in Korea's 2026 crisis, invisible spillover from one fund's rebalancing accounted for 41% of price swings in another asset.

When multiple AI models share their complete solutions with each other, they quickly converge on the same answer within a single round, eliminating the diversity that made having multiple models useful in the first place. Across 11 optimization tasks with equal budgets, having agents work independently produced better results than letting them see each other's full outputs, because interaction caused them to stick with the first solution they encountered rather than exploring different approaches.

Flight engineers have developed a family of related control techniques that tell planes how to respond to pilot commands by using acceleration sensors instead of complex mathematical models. The newer versions, called incremental methods, are simpler and more forgiving of imperfect airplane models, but they create new design tradeoffs that engineers are still learning to navigate.

A new statistical method predicts how long people will watch short videos by modeling the full range of possible watch times—from people who quit immediately to those who watch all the way through. The approach fixes three critical problems in prior methods: collapsed variance, redundant components, and inactive parts of the model. In a real-world test with millions of videos, it improved ranking accuracy and made predictions more stable and interpretable.

Large language models trained to follow generic privacy rules ignore individual user preferences more than half the time. Researchers built a benchmark to test personalized privacy (where different users have different comfort levels with sharing information) and developed a method that adjusts how the AI's internal attention mechanisms work at inference time, significantly improving compliance with each user's specific privacy boundaries.

When AI agents use external tools, a single approved request can accidentally trigger multiple effects if something goes wrong in delivery or recovery. AID-Guard is a new system that tracks authorization through the entire lifecycle — from approval through retries and crashes — ensuring one approved action produces exactly one real-world effect. In testing with Stripe and Resend payment systems, the protocol prevented all 44 simulated attacks while allowing legitimate requests through.

When quantum computers use a computational shortcut called annealing to solve problems, they can get trapped in a dead-end state if the problem crosses a certain type of phase transition. Researchers developed a mathematical framework predicting how many errors pile up during this trap, finding that the error rate follows predictable power-law patterns that shift sharply depending on the problem parameters.

A new method called Re³Cap improves how AI systems describe images by having them search for similar images and captions to catch mistakes. The technique outperforms previous approaches by up to 8.64% on standard image captioning benchmarks, without needing humans to label additional training data.