File size: 1,235 Bytes
b3c660d |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 |
{
"origin": {
"name": "generalizations-mu-goodman",
"source": "https://github.com/brendon-boldt/emergent-generalization",
"upstream_source": "https://github.com/jayelm/emergent-generalization",
"paper": "https://openreview.net/forum?id=yq5MYHVaClG"
},
"system": {
"game_type": "signalling",
"game_subtype": "discrimination",
"observation_type": "image",
"observation_continuous": true,
"multi_step": false,
"seeding_available": false,
"data_source": "synthetic",
"multi_utterance": false,
"more_than_2_agents": false,
"symmetric_agents": false,
"variants": {
"shapeworld-reference": {},
"shapeworld-set_reference": {},
"shapeworld-concept": {},
"cub-reference": {
"data_source": "natural"
},
"cub-set_reference": {
"data_source": "natural"
},
"cub-concept": {
"data_source": "natural"
}
}
},
"notes": "The sampled output from this environment has both \"train\" (seen) and \"test\" (unseen) values. For now, we are just using the previously seen examples as we expect them to have better performance. The unseen examples could be included in the future (as separate corpora).\n"
}
|