Generative models trained on natural sequences are increasingly used to predict the effects of genetic variation, enabling progress in therapeutic design, disease risk prediction, and synthetic biology. In the zero-shot setting, variant impact is estimated by comparing the likelihoods of sequences, under the assumption that likelihood serves as a proxy for fitness. However, this assumption often breaks down in practice: sequence likelihood reflects not only evolutionary fitness constraints, but also phylogenetic structure and sampling biases, especially as model capacity increases. We introduce Likelihood-Fitness Bridging (LFB), a simple and general strategy that improves variant effect prediction by averaging model scores across sequences subject to similar selective pressures. Assuming an Ornstein-Uhlenbeck model of evolution, LFB can be viewed as a way to marginalize the effects of genetic drift, although its benefits appear to extend more broadly. LFB applies to existing protein and genomic language models without requiring retraining, and incurs only modest computational overhead. Evaluated on largescale deep mutational scans and clinical benchmarks, LFB consistently improves predictive performance across model families and sizes. Notably, it reverses the performance plateau observed in larger protein language models, making the largest models the most accurate when combined with LFB. These results suggest that accounting for phylogenetic and sampling biases is essential to realizing the full potential of large sequence models in variant effect prediction.

Plus: SK Hynix has overtaken Samsung as South Korea's most valuable company. I'm under the iconic fjords of Norway to visit what will soon become the world's longest and deepest subsea road tunnel--an exceptional engineering feat that will carry drivers deep beneath the North Sea. I'm here to understand how you make a 16.6-mile highway that sits 1,280 feet below the sea at its deepest point. And also--at a time when it can feel hard to get anything done--to reassure myself that ambitious engineering is still possible. That we can still make things. Step inside Norway's Rogfast tunnel and see how engineers are making it happen .

Language model outputs are not always reliable, thus prompting research into how to adapt model responses based on uncertainty. Common approaches include: abstention, where models refrain from generating responses when uncertain; and linguistic calibration, where models hedge their statements using uncertainty quantifiers. However, abstention can withhold valuable information, while linguistically calibrated responses are often challenging to leverage in downstream tasks. We propose a unified view, Conformal Linguistic Calibration (CLC), which reinterprets linguistic calibration as answer set prediction. First we present a framework connecting abstention and linguistic calibration through the lens of linguistic pragmatics. We then describe an implementation of CLC that allows for controlling the level of imprecision in model responses. Results demonstrate our method produces calibrated outputs with conformal guarantees on factual accuracy. Further, our approach enables fine-tuning models to perform uncertainty-aware adaptive claim rewriting, offering a controllable balance between factuality and specificity.1

We study the sublinear multivariate mean estimation problem in d-dimensional Euclidean space. Specifically, we aim to find the mean µ of a ground point set A, which minimizes the sum of squared Euclidean distances of the points in Ato µ. We first show that a multiplicative (1 + ε) approximation to µ can be found with probability 1 δ using O(ε 1 logδ 1)many independent uniform random samples, and provide a matching lower bound. Furthermore, we give two estimators with optimal sample complexity that can be computed in optimal running time for extracting a suitable approximate mean: 1.

We study the problem of minimizing the sum of a smooth function and a nonsmooth convex regularizer over a compact Riemannian submanifold embedded in Euclidean space. By introducing an auxiliary splitting variable, we propose an adaptive Riemannian alternating direction method of multipliers (ARADMM), which, for the first time, achieves convergence without requiring smoothing of the nonsmooth term. Our approach involves only one Riemannian gradient evaluation and one proximal update per iteration. Through careful and adaptive coordination of the stepsizes and penalty parameters, we establish an optimal iteration complexity of order O(ϵ 3) for finding an ϵ-approximate KKT point, matching the complexity of existing smoothing technique-based Riemannian ADMM methods. Extensive numerical experiments on sparse PCA and robust subspace recovery demonstrate that our ARADMM consistently outperforms state-of-the-art Riemannian ADMM variants in convergence speed and solution quality.

LLM-as-Benchmark-Generator methods have been widely studied as a supplement to human annotators for scalable evaluation, while the potential biases within this paradigm remain underexplored. In this work, we systematically define and validate the phenomenon of inflated performance in models evaluated on their self-generated benchmarks, referred to as self-bias, and attribute it to sub-biases arising from question domain, language style, and wrong labels. On this basis, we propose SILENCER, a general framework that leverages the heterogeneity between multiple generators at both the sample and benchmark levels to neutralize bias and generate high-quality, self-bias-silenced benchmark. Experimental results across various settings demonstrate that SILENCER can suppress self-bias to near zero, significantly improve evaluation effectiveness of the generated benchmark (with an average improvement from 0.655 to 0.833 in Pearson correlation with high-quality human-annotated benchmark), while also exhibiting strong generalizability.

Follow this section to personalize your feed and get instant alerts. Follow Go to your personalized feed WHY FOLLOW? Smart Alerts: Get notified about major news as it happens. Follow this tag to personalize your feed and get instant alerts. Follow Go to your personalized feed WHY FOLLOW?

Follow this section to personalize your feed and get instant alerts. Follow Go to your personalized feed WHY FOLLOW? Smart Alerts: Get notified about major news as it happens. Follow this tag to personalize your feed and get instant alerts. Follow Go to your personalized feed WHY FOLLOW?

A.1 Compilation and Standardization of Datasets We compile ESCAPE from 27 peptide databases by systematically extracting experimentally validated antimicrobial peptides annotated for antibacterial, antifungal, antiparasitic, or antiviral activity. Databases exclusively focusing on a single category, such as AVPdb [1] (antiviral), are directly mapped to one of the four target classes. Additionally, we follow the methodology outlined in TransImbAMP[6], selecting non-antimicrobial peptides from UniProt [7] by applying strict exclusion criteria. Specifically, we discard sequences containing keywords such as "membrane," "toxic," "secretory," "defensive," "antibiotic," "anticancer," "antiviral," or "antifungal" to enhance the quality of the negative class. For large and hierarchically structured databases such as DBAASP[8], DRAMP[9], dbAMP (with species-level annotations)[10], and SATPdb (which lists 38 functional categories)[11], we retain all peptides with annotations that map either directly or through hierarchical or taxonomic relationships to one of our four defined antimicrobial classes (antibacterial, antifungal, antiparasitic, antiviral).

Despite widespread adoption, multimodal large language models (MLLMs) suffer performance degradation when encountering unfamiliar queries under distribution shifts. Existing methods to improve MLLM generalization typically require either more instruction data or larger advanced model architectures, both of which incur non-trivial human labor or computational costs. In this work, we take an alternative approach to enhance the generalization and robustness of MLLMs under distribution shifts, from a representation learning perspective. Inspired by information bottleneck (IB) principle, we derive a variational lower bound of the IB for MLLMs and devise a practical implementation, Visual Instruction Bottleneck Tuning (Vittle). We then provide a theoretical justification of Vittle by revealing its connection to an information-theoretic robustness metric of MLLM. Empirical validation of multiple MLLMs on open-ended and closed-form question answering and object hallucination detection tasks over 45 datasets, including 30 shift scenarios, demonstrates that Vittleconsistently improves the MLLM's robustness under shifts by pursuing the learning of a minimal sufficient representation.