SoftMatcha 2: New Algorithm Enables Fast Semantic Search Across Trillion-Token Corpora
Researchers have developed SoftMatcha 2, a search algorithm that can find semantic variations in massive text datasets (1.4 trillion tokens) in under 0.3 seconds. The method combines suffix array-based string matching with vector representations of words to allow flexible searching despite insertions, deletions, and substitutions. The technique has practical applications including detecting benchmark contamination in training data and improving information retrieval systems.
SoftMatcha 2 is an ultra-fast search algorithm designed to query trillion-scale natural language corpora while accommodating semantic variations. The approach uses suffix arrays for efficient string matching and represents words as vectors to enable semantic flexibility. To prevent the combinatorial explosion that typically results from allowing semantic relaxation, the algorithm employs two key strategies: dynamic corpus-aware pruning and fast exact lookup via disk-aware design. Testing on the FineWeb-Edu corpus (1.4 trillion tokens), SoftMatcha 2 achieved substantially lower search latency than competing methods including infini-gram and the original SoftMatcha. Beyond raw search performance, the researchers demonstrated practical applications such as uncovering benchmark contamination in training corpora that previous approaches missed, and the method also supports information retrieval and paraphrase detection tasks. The team has released an online demo enabling fast, soft search across corpora in seven languages.
What different sources said
- arXiv cs.LGCenter
SoftMatcha 2: A Fast and Soft Pattern Matcher for Trillion-Scale Corpora
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.