New dataset reveals what drives cross-lingual capability transfer in LLMs
Source published ·Modelwire updated
Original coverage: arXiv cs.CL ↗·How Modelwire adds context
The development
Researchers have created Multilingual GSM-Symbolic, a 30,000-pair mathematical reasoning dataset spanning 15 languages with symbolic templating to prevent overfitting. The work directly addresses a critical gap in LLM evaluation: understanding which factors govern whether capabilities learned in one language transfer to others. Rather than requiring exhaustive testing across all language pairs, this dataset enables developers to identify and target the specific constraints limiting performance in low-resource settings. The symbolic generation approach scales to millions of variations, offering a replicable methodology for cross-lingual capability research that could reshape how teams prioritize multilingual model development.
Modelwire’s AI-generated summary of coverage from arXiv cs.CL.
Modelwire analysis
ExplainerOur AI-generated reading of the wider context and the next developments to watch.
The paper's real contribution is methodological: it isolates which linguistic properties actually constrain transfer versus which don't. Most prior work assumes transfer fails uniformly across languages; this dataset lets researchers identify whether the bottleneck is morphological complexity, script differences, training data volume, or something else entirely.
This connects directly to the annotation-budget work from late September, which found that task complexity, not just language scarcity, determines labeling needs across African languages. Multilingual GSM-Symbolic extends that insight: it provides a controlled testbed to measure whether a capability gap is linguistic or data-driven. It also echoes the unlearning benchmark from September 30th, which exposed how facts remain accessible across language boundaries. Both papers treat multilingual systems as fundamentally different from monolingual ones, requiring language-aware diagnostics rather than blanket assumptions about transfer.
If teams using this dataset identify a single linguistic feature (e.g., morphological richness or script family) that predicts transfer failure across multiple language pairs, that's signal the field has found a replicable diagnostic. If instead transfer patterns remain idiosyncratic to each pair, the dataset's value shifts from predictive tool to documentation of fragmentation.
This interpretation is generated from the summary above and the archive coverage cited below. Our methodology · Report an error
Coverage behind this analysis
These archive entries ground the connection in our analysis. They are ordered by source publication date, with links to our coverage and the original sources.
·arXiv cs.CL
African language classifiers need 400+ labels per task, cross-lingual pooling shows limits
Researchers empirically quantify the annotation budgets needed to train text classifiers across 28 African-language tasks, finding that topic classification converges around 400 labels while sentiment analysis demands thousands. The work tests whether pooling labels across related languages can reduce per-language annotation costs, using lightweight character n-gram models that require no pretraining or accelerators. This directly…
MentionsMultilingual GSM-Symbolic · GSM-Symbolic
How this coverage is produced
Modelwire uses AI to generate summaries and context from source headlines, snippets, and selected archive coverage. Automated checks do not verify every claim, and items are not routinely reviewed by a person before publication. Zacaria Solis operates the site. Read the linked source for the full evidence and report errors through our corrections process.
Modelwire summarizes, we don’t republish. arXiv cs.CL originally reported this story as “Multilingual GSM-Symbolic: What determines capability transfer across languages?”. The full content lives on arxiv.org. If you’re a publisher and want a different summarization policy for your work, see our takedown page.