In this natural setting, it is not common to see the same data sample multiple times, or even twice at times.

'It brings you closer to the natural world': the rise of the Merlin birdsong identifying app W hen Natasha Walter first became curious about the birds around her, she recorded their songs on her phone and arduously tried to match each song with online recordings. After a friend recommended Merlin Bird ID, a free app, she tried it in her London garden and was delighted to discover the birds she assumed were female blackbirds - "this is how bad a birder I was" - were actually song thrushes and mistle thrushes. "I'm obsessed with Merlin - it's wonderful and it's been a joy to me," says Walter, a writer and human rights activist. "This is what AI and machine-learning have been invented for. Merlin is having a moment. The app, developed by the Cornell Lab of Ornithology in New York, which listens for birdsong and identifies the species singing, has been downloaded 33m times, in 240 countries and territories around the world. Britain has the second highest total number of users - more than 1.5 million in 2024, an 88% increase from 2023. Every month, there has been a 30% increase in new users of the app, whose sound identification function was launched in 2021. Merlin has been trained to identify the songs of more than 1,300 species around the world, with more birds added twice a year. Different songs make distinct patterns on spectrograms and Merlin is trained to recognise these different shapes and attribute them to a species. For latecomers to birding, or those lacking a knowledgeable friend, the app has become their teacher. "My fear at first was I wouldn't actually learn because I'm outsourcing my understanding of birds to this app," says Walter. "But that hasn't come to pass.

The ability to continuously learn and adapt itself to new tasks, without losing grasp of already acquired knowledge is a hallmark of biological learning systems, which current deep learning systems fall short of. In this work, we present a novel methodology for continual learning called MERLIN: Meta-Consolidation for Continual Learning. We assume that weights of a neural network, for solving task, come from a meta-distribution. This meta-distribution is learned and consolidated incrementally. We operate in the challenging online continual learning setting, where a data point is seen by the model only once. Our experiments with continual learning benchmarks of MNIST, CIFAR-10, CIFAR-100 and Mini-ImageNet datasets show consistent improvement over five baselines, including a recent state-of-the-art, corroborating the promise of MERLIN.

Radiology plays an integral role in modern medicine, yet rising imaging volumes have far outpaced workforce growth. Foundation models offer a path toward assisting with the full spectrum of radiology tasks, but existing medical models remain limited: they process volumetric CT and MRI as low-fidelity 2D slices, discard critical grayscale contrast information, and lack evaluation frameworks that reflect real clinical practice. We introduce Pillar-0, a radiology foundation model pretrained on 42,990 abdomen-pelvis CTs, 86,411 chest CTs, 14,348 head CTs, and 11,543 breast MRIs from a large academic center, together with RATE, a scalable framework that extracts structured labels for 366 radiologic findings with near-perfect accuracy using LLMs. Across internal test sets of 14,230 abdomen-pelvis CTs, 10,646 chest CTs, 4,906 head CTs, and 1,585 breast MRIs, Pillar-0 establishes a new performance frontier, achieving mean AUROCs of 86.4, 88.0, 90.1, and 82.9, outperforming MedGemma (Google), MedImageInsight (Microsoft), Lingshu (Alibaba), and Merlin (Stanford) by 7.8-15.8 AUROC points and ranking best in 87.2\% (319/366) tasks. Pillar-0 similarly outperforms all baselines in an external validation on the Stanford Abdominal CT dataset, including Merlin (82.2 vs 80.6 AUROC). Pillar-0 extends to tasks beyond its pretraining, such as long-horizon lung cancer risk prediction, where it improves upon the state-of-the-art Sybil by 3.0 C-index points on NLST, and generalizes with gains of 5.9 (MGH) and 1.9 (CGMH). In brain hemorrhage detection, Pillar-0 obtained a >95 AUROC when using only 1/20th of the data of the next most sample efficient baseline. Pillar-0 and RATE together provide an open, clinically rigorous foundation for building high-performance radiology systems, enabling applications that were previously infeasible due to computational, data, and evaluation constraints.

In Social Deduction Games (SDGs) such as Avalon, Mafia, and W erewolf, players conceal their identities and deliberately mislead others, making hidden-role inference a central and demanding task. Accurate role identification, which forms the basis of an agent's belief state, is therefore the keystone for both human and AI performance. We introduce CSP4SDG, a probabilistic, constraint-satisfaction framework that analyses gameplay objectively. Game events and dialogue are mapped to four linguistically-agnostic constraint classes--evidence, phenomena, assertions, and hypotheses. Hard constraints prune impossible role assignments, while weighted soft constraints score the remainder; information-gain weighting links each hypothesis to its expected value under entropy reduction, and a simple closed-form scoring rule guarantees that truthful assertions converge to classical hard logic with minimum error. The resulting posterior over roles is fully interpretable and updates in real time. Experiments on three public datasets show that CSP4SDG (i) outperforms LLM-based baselines in every inference scenario, and (ii) boosts LLMs when supplied as an auxiliary "reasoning tool." Our study validates that principled probabilistic reasoning with information theory is a scalable alternative--or complement--to heavy-weight neural models for SDGs.

We thank all reviewers for their insightful comments and the time they have spent carefully reviewing the paper. Consistent among all reviewers is the comment that the paper could be improved with further experiments. In Appendix F.2, we introduce another operator that In response to Reviewer 2's comment regarding comparisons to schemes where the adaptive entropy coefficient is We will clarify this difference in the paper.

Multivariate Time Series Forecasting (MTSF) involves predicting future values of multiple interrelated time series. Recently, deep learning-based MTSF models have gained significant attention for their promising ability to mine semantics (global and local information) within MTS data. However, these models are pervasively susceptible to missing values caused by malfunctioning data collectors. These missing values not only disrupt the semantics of MTS, but their distribution also changes over time. Nevertheless, existing models lack robustness to such issues, leading to suboptimal forecasting performance. To this end, in this paper, we propose Multi-View Representation Learning (Merlin), which can help existing models achieve semantic alignment between incomplete observations with different missing rates and complete observations in MTS. Specifically, Merlin consists of two key modules: offline knowledge distillation and multi-view contrastive learning. The former utilizes a teacher model to guide a student model in mining semantics from incomplete observations, similar to those obtainable from complete observations. The latter improves the student model's robustness by learning from positive/negative data pairs constructed from incomplete observations with different missing rates, ensuring semantic alignment across different missing rates. Therefore, Merlin is capable of effectively enhancing the robustness of existing models against unfixed missing rates while preserving forecasting accuracy. Experiments on four real-world datasets demonstrate the superiority of Merlin.

High-performance OLAP database technology has emerged with the growing demand for massive data analysis. To achieve much higher performance, many DBMSs adopt sophisticated designs including SIMD operators, parallel execution, and dynamic pipeline modification. However, such advanced OLAP query execution mechanisms still lack targeted Query Performance Prediction (QPP) methods because most existing methods target conventional tree-shaped query plans and static serial executors. To address this problem, in this paper, we proposed MERLIN a multi-stage query performance prediction method for high-performance OLAP DBMSs. MERLIN first establishes resource cost models for each physical operator. Then, it constructs a DAG that consists of a data-flow tree backbone and resource competition relationships among concurrent operators. After using a GAT with an extra attention mechanism to calibrate the cost, the cost vector tree is extracted and summarized by a TCN, ultimately enabling effective query performance prediction. Experimental results demonstrate that MERLIN yields higher performance prediction precision than existing methods.