Text-based RL Agents with Commonsense Knowledge: New Challenges, Environments and Baselines

Text-based games have emerged as an important test-bed for Reinforcement Learning (RL) research, requiring RL agents to combine grounded language understanding with sequential decision making. In this paper, we examine the problem of infusing RL agents with commonsense knowledge... (read more)

PDF Abstract
TASK DATASET MODEL METRIC NAME METRIC VALUE GLOBAL RANK BENCHMARK
Commonsense Reasoning for RL commonsense-rl Human Avg #Steps 15.00 ± 3.29 # 1
Commonsense Reasoning for RL commonsense-rl Optimal Avg #Steps 15.00 ± 2.00 # 1
Commonsense Reasoning for RL commonsense-rl LSTM-A2C Avg #Steps 49.21 ± 0.58 # 1
Commonsense Reasoning for RL commonsense-rl KG-A2C Avg #Steps 49.36 ± 7.50 # 1
Commonsense Reasoning for RL commonsense-rl TNC-A2C Avg #Steps 43.27 ± 0.70 # 1

Methods used in the Paper


METHOD TYPE
BiGRU
Bidirectional Recurrent Neural Networks
GAT
Graph Models