English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Imitation Learning (IL) is one of the most widely used methods in machinelearning. Yet, while powerful, many works find it is often not able to fullyrecover the underlying expert behavior. However, none of these works deeplyinvestigate the role of scaling up the model and data size. Inspired by recentwork in Natural Language Processing (NLP) where "scaling up" has resulted inincreasingly more capable LLMs, we investigate whether carefully scaling upmodel and data size can bring similar improvements in the imitation learningsetting. To demonstrate our findings, we focus on the game of NetHack, achallenging environment featuring procedural generation, stochasticity,long-term dependencies, and partial observability. We find IL loss and meanreturn scale smoothly with the compute budget and are strongly correlated,resulting in power laws for training compute-optimal IL agents with respect tomodel size and number of samples. We forecast and train several NetHack agentswith IL and find they outperform prior state-of-the-art by at least 2x in allsettings. Our work both demonstrates the scaling behavior of imitation learningin a challenging domain, as well as the viability of scaling up currentapproaches for increasingly capable agents in NetHack, a game that remainselusively hard for current AI systems.

Scaling Laws for Imitation Learning in NetHack