DecompSR: New Benchmark Dataset for Testing Compositional Spatial Reasoning in Large Language Models
Researchers introduced DecompSR, a large benchmark dataset with over 5 million datapoints designed to test how well large language models perform compositional spatial reasoning tasks. The dataset allows independent variation of multiple compositionality aspects including reasoning depth, entity variability, and systematic generalization. The findings reveal that LLMs struggle with productive and systematic generalization in spatial reasoning but handle linguistic variation more robustly.
DecompSR is a procedurally generated benchmark dataset containing over 5 million datapoints created to analyze compositional spatial reasoning abilities in large language models. The dataset's generation framework allows researchers to independently manipulate four key aspects of compositionality: productivity (reasoning depth), substitutivity (entity and linguistic variability), overgeneralisation (input order and distractors), and systematicity (novel linguistic elements). The dataset is constructed to be correct by construction and independently verified using a symbolic solver to guarantee accuracy. Comprehensive benchmarking across multiple LLMs revealed that these models struggle particularly with productive and systematic generalization in spatial reasoning tasks, while demonstrating greater robustness to linguistic variation. This work provides a rigorous, provably correct tool for fine-grained probing of compositional reasoning abilities in language models.
What different sources said
- arXiv cs.AICenter
DecompSR: A dataset for decomposed analyses of compositional multihop spatial reasoning
Related
Genetic Drift, Not Selection, Drives Rapid Feather Color Evolution in Island Bird Radiation
A new study of an island bird radiation found that rapid evolution of feather coloration is driven primarily by genetic drift in small populations rather than sexual or ecological selection. The research integrated whole-genome data with detailed plumage measurements across complete species sampling to test whether signaling trait evolution correlates with speciation rates. The findings suggest that neutral demographic processes play a central role in generating phenotypic diversity during island radiations, challenging assumptions about the mechanisms driving rapid evolution.
New AI Model Improves Prediction of Therapeutic Peptide Function from Protein Sequences
Researchers developed a lightweight CNN classifier that predicts whether peptide sequences have therapeutic properties, trained on a database of 54,655 peptides across 48 functional categories. The model uses a novel negative sampling strategy to reduce false positive rates from over 60% in previous approaches to 2.1%. This advancement could accelerate drug discovery by enabling faster computational screening of peptide candidates before expensive experimental testing.
Study Shows Different Metabolic Stress Models Produce Distinct Effects on Human Neuronal Networks
Researchers tested three common in vitro metabolic stress models on human-derived neuronal networks and found each produced different patterns of neuronal activity and cell damage. The models tested were hypoxia alone, oxygen-glucose deprivation (OGD), and hypoxia combined with glutamate exposure. The findings suggest that choice of experimental model significantly affects results and that combining electrophysiological and structural analyses is important for accurately assessing metabolic stress in stroke research.