🧪 Test?View on arXiv
Hierarchical Data Selection via Manifold Coverage and Sparse Feature Coverage in LLM Post-training
Not provided
data selectionfine-tuningmanifold learningsparse representation
2608.16927
Builder Relevance
2h ago80%
Abstract
The paper proposes a method for selecting high-value subsets from large datasets to reduce training costs and improve model performance.
Reality Card
Core Claim
MASS consistently outperforms strong data selection baselines and matches or surpasses full data training with only a small subset of data.
Method / Result
MASS achieves superior performance across multiple budgets in experiments on Vision Flan and LLaVA-CoT.
Limitations
The paper does not specify the authors or provide detailed reproducibility metrics.
Paper to code
Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.
No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.