Interactive Fiction Games: A Colossal Adventure

Authors: Matthew Hausknecht, Prithviraj Ammanabrolu, Marc-Alexandre Côté, Xingdi Yuan7903-7910

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate the agents across a set of thirty-two Jericho-supported games with the aims of 1) showing the feasibility of reinforcement learning on a variety of different IF games, 2) creating a reproducible benchmark for future work, 3) investigating the difference between choice-based and template-based action spaces, and 4) comparing performance of general IF game playing agents (NAIL), single-game agents (DRRN and TDQN), and a random agent (RAND) which uniformly sample commands from a set of canonical actions.
Researcher Affiliation Collaboration Matthew Hausknecht Microsoft Research AI Prithviraj Ammanabrolu Georgia Institute of Technology Marc-Alexandre Cˆot e Microsoft Research Montr eal Xingdi Yuan Microsoft Research Montr eal
Pseudocode Yes Algorithm 1 Procedure for Identifying Valid Actions
Open Source Code Yes Jericho is available at https://github.com/microsoft/jericho.
Open Datasets Yes Jericho supports a set of fifty-six human-made IF games that cover a variety of genres... There exists a large collection of over a thousand unsupported games 3, which may be useful for unsupervised pretraining or intrinsic motivation. 3https://github.com/BYU-PCCL/z-machine-games
Dataset Splits No The paper mentions 'Additional experiment details and hyperparameters are located in the supplementary material' but does not provide specific train/validation/test dataset splits (e.g., percentages, sample counts, or explicit splitting methodology) in the main text.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU models, CPU types, or memory) used for running the experiments.
Software Dependencies No The paper mentions software components like 'Python-based IF environment', 'Sentence Piece model', and 'GRU encoders' but does not provide specific version numbers for these or other software dependencies.
Experiment Setup No The paper states 'Additional experiment details and hyperparameters are located in the supplementary material' but does not include specific experimental setup details (e.g., concrete hyperparameter values, training configurations, or system-level settings) within the main text.