100-LongBench and LongScore evaluate LLMs at controlled context lengths and rank models by relative performance drop from a short-context baseline, not by raw accuracy.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
100-LongBench: Are de facto Long-Context Benchmarks Literally Evaluating Long-Context Ability?
100-LongBench and LongScore evaluate LLMs at controlled context lengths and rank models by relative performance drop from a short-context baseline, not by raw accuracy.