SKU RLE-1301 · Sold by External

ALFWorld

Product specifications

SKURLE-1301
Data typeRL environments
Volume134 unseen eval games
Size on diskNot published by source
FormatPython (pip: alfworld) + PDDL/game files
Access modelPUBLIC LICENSE
PricingFree · open-source license
Quality score
LicenseMIT License
ALFWorld pairs interactive TextWorld games with the embodied ALFRED dataset so agents can be trained in text and evaluated (or transferred) in an embodied setting. It covers six household task types (Pick & Place, Examine in Light, Clean & Place, Heat & Place, Cool & Place, Pick Two & Place). At each step the environment exposes textual observations plus a list of admissible commands, and returns a sparse goal-completion reward. The public benchmark ships 134 unseen evaluation games.