Codebase to test Top-k Attention and Top-theta Attention on Large Language Models using the lm-eval-harness [1] framework, and text generation tasks including HumanEval [2] and LongBench [3] ...
This repository also contains a reusable velocity-driven mobility handler in src/velocity_mobility, used by the simulation to apply speed/acceleration limits to commanded velocities.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results