Benchmark-aligned data improves narrow metrics at the cost of generalization while coverage-expanding data enables distributed parameter adaptation and better overall capability, shown via spectral and rank diagnostics across LLM families.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Benchmark Shadows: Data Alignment, Parameter Footprints, and Generalization in Large Language Models
Benchmark-aligned data improves narrow metrics at the cost of generalization while coverage-expanding data enables distributed parameter adaptation and better overall capability, shown via spectral and rank diagnostics across LLM families.