Reusable code skills help language agents progress farther at lower cost

In NetHack, agents using semantic code-based skills advanced farther, required fewer model inferences, and learned faster than agents limited to primitive actions.

Independent
Bartłomiej Cupiał · Jens Tuyls · Maciej Wołczyk · Davide Paglieri · Martin Klissarov · Benjamin Eysenbach · +2 more
Research Digest··2 min read
Cupiał et al.

The authors built CodeHack, a library of executable NetHack skills paired with natural-language descriptions.

Why this paper

Independent

In one line

Code-based skills nearly triple game progression and reduce inference cost by 86% for language agents in NetHack.

What we could check

  • ·No code link found
  • ·No weights link found
  • ·No dataset link found
  • ·No compute details found
  • ✓Limitations stated by the authors
  • ·No benchmark numbers found

Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.

§
newspaper

Research Digest

Articles published under the Zotpaper byline are synthesized from multiple source publications by our AI editor and reviewed by our editorial process. Each story combines reporting from credible outlets to give readers a balanced, comprehensive view.