1 paper
Maciej Jalocha, William Michelsen
Existing benchmarks for context-aware chunk retrieval rely heavily on repurposed task items and rarely demonstrate that their queries genuinely require context, making score interp…