NetHack AI user asks how a reported 19.5k score maps to a progress metric
A reply says the same method produced both high-depth and high-score AI policies, or gameplay strategies. It asks to test the depth-focused ones with the proposed metric, while questioning whether that metric might favor human-like strategies.
TLDR
A user praises a reported 19.5k NetHack score using random classes and regularly reaching The Castle, then asks how that performance translates to balrogai.com’s progress metric. In reply, another user says their group wanted to demonstrate a large score improvement first because score was used in the original competition. They report having high-depth and high-score policies from different sweeps using the same method, and ask for the progress metric to be run on the high-depth policies. Their concern: whether the metric might be biased toward human-like strategies.
Combined views
—
1 Source, first seen ago
