Suite of OpenAI Gym-compatible multi-agent reinforcement learning environment centered around referntial games to benchmark for behavioral traits pertaining to symbolic behaviours, as described in Santoro et al., 2021, "Symbolic Behaviours in Artificial Intelligence", primarily: exhibiting receptive, constructive, malleable, and separable behaviours.
gym
must be installed. Environments can be created as follows, for instance, in order to test for receptivity and constructivity:
import gym
import symbolic_behaviour_benchmark
env = gym.make(
"SymbolicBehaviourBenchmark-ReceptiveConstructiveTestEnv-v0",
nbr_communication_rounds = 1,
vocab_size = 6,
max_sentence_length = 3,
descriptive = True,
nbr_latents = 3,
min_nbr_values_per_latent = 2,
max_nbr_values_per_latent = 5,
nbr_object_centric_samples = 4,
nbr_distractors = 0,
use_communication_channel_permutations = True,
allow_listener_query = False,
provide_listener_feedback = True,
sampling_strategy = "component-focused-4shots",
)
git clone https://www.github.com/Near32/SymbolicBehaviourBenchmark
pip install -e ./SymbolicBehaviourBenchmark/