Llama-3.2-1B shows better TwinKV improvements on RULER benchmark than Qwen3-4B across all evaluated configurations

On RULER with Llama-3.2-1B, however, that fourth policy improves in every evaluated cell because its Alone score leaves substantial room to improve.

http://arxiv.org/abs/2608.27128v1