Spearman
PulseAugur coverage of Spearman — every cluster mentioning Spearman across labs, papers, and developer communities, ranked by signal.
6 day(s) with sentiment data
-
Graph Neural Network Reliability Explored in Protein Function Prediction
A new research paper published on arXiv explores the reliability of Graph Neural Networks (GNNs) in predicting protein function. The study investigates whether tissue-specific interaction structures within protein netwo…
-
New adapter method tackles routing collapse in time series foundation models
Researchers have identified a critical issue in time series foundation models (TSFMs) where standard mixture-of-experts (MoE) adaptations fail due to "normalization-induced routing collapse." This phenomenon occurs beca…
-
New research questions Hessian trace as reliable model-selection signal
A new research paper explores the use of internal model properties, specifically the Hessian trace and its eigenvalues, as a proxy for model selection when external validation data is unreliable. The study found that wh…
-
New research explores rank learning in AI matrix memories
Researchers have developed a new method to investigate how gradient-based training can learn the necessary rank for storing and composing associations within a matrix memory. Their study trained matrix memories on key-v…
-
Study finds LLM data audits don't guarantee downstream utility for African NLP
A new study published on arXiv investigates the effectiveness of synthetic data selection methods for low-resource African languages. The research found that common proxies, which assume that data rated highly by an LLM…
-
New DFER framework predicts human disagreement for better calibration
Researchers have developed a new framework for dynamic facial expression recognition (DFER) that accounts for human disagreement among annotators. This approach uses a Dirichlet-multinomial likelihood to train models di…
-
Transformer model decodes psilocybin's gene response
A new research paper introduces a Transformer-based model designed to analyze the transcriptional response to psilocybin. This unsupervised model classifies gene expression changes in single-nucleus RNA-sequencing data,…
-
Polite prompts alter LLM relevance judgments, study finds
A new study published on arXiv investigates the impact of politeness in prompts on large language models used as relevance judges. Researchers found that tone significantly affects model judgments, with effects varying …
-
New PruneShift framework evaluates AI model pruning decision reliability
Researchers have introduced PruneShift, a new framework designed to evaluate the reliability of decisions made during structured pruning in machine learning models. Unlike previous methods that focused on average surrog…
-
New method aligns multimodal LLMs for ordinal tasks
Researchers have identified a significant gap in how Multimodal Large Language Models (MLLMs) handle ordinal regression tasks, such as age estimation or image quality assessment. While internal model states show strong …
-
New research explores concept binding in unified multimodal models
Researchers have developed a novel method to investigate the relationship between understanding and generation in unified multimodal models (UMMs). By constructing a visual entity that is trained through only one task d…
-
AI security research: Monitor skill, not lineage, drives ensemble effectiveness
A new research paper titled "Decorrelation Is Not Complementarity: Skill, Not Lineage, Governs Trusted-Monitor Ensembles" challenges the assumption that diverse pretraining lineages are key to building effective trusted…
-
New framework offers controlled manipulation of LLM sycophancy
Researchers have developed a new framework called PCA-guided Activation Scaling (PAS) to control sycophancy in large language models (LLMs). Sycophancy, the tendency of LLMs to agree with users regardless of accuracy, c…
-
SoftMCC framework enhances model selection for imbalanced classification
Researchers have introduced SoftMCC, a novel post-training framework designed to improve model selection for imbalanced binary classification tasks. This method addresses the threshold-dependency issues inherent in trad…
-
Quantum Machine Learning uses late fusion to cut costs and boost robustness
Researchers have proposed a new method called "late fusion" for running large quantum neural networks (QNNs) on smaller devices. This approach avoids the computationally expensive reconstruction step typically required …
-
Quantum ML gains efficiency with new late fusion technique
Researchers have developed a new method called "late fusion" for quantum machine learning (QML) that significantly reduces computational costs. This technique involves training independent subcircuits of a quantum neura…
-
LLM judge noise can mask real performance gains, study finds
An LLM judge's unreliability can systematically bias evaluation results, not just widen error bars. This 'attenuation' effect, described by Spearman in 1904, causes real improvements to appear smaller or non-existent. T…
-
New theory explains divergence between activation patching and weight-space ablation
Researchers have developed a theoretical framework to understand the relationship between activation patching and weight-space ablation, two methods used to determine causal responsibility in neural networks. The theory…
-
Quantum Kernel Geometry Survival Tested on IBM Hardware
Researchers have investigated the survival of geometric information within a four-qubit quantum kernel on IBM Quantum Hardware. The study focused on a specific frozen ZZ feature-map kernel, analyzing its performance acr…
-
Denoising Models Develop Human-Like Perceptual Illusion Representations
Researchers have discovered that denoising models, when trained on natural images, develop internal representations that are sensitive to perceptual illusions, similar to human observers. These representations were foun…