jbostock's picture
Add Dispatch and generic evaluation results
a77c938 verified
Raw
History Blame Contribute Delete
6.6 kB
{
"arm": "charter",
"seed": 314159,
"n_mmlu": 40,
"n_gsm8k": 40,
"rows": [
{
"condition": "no_aft",
"capability": {
"n": {
"mmlu": 40,
"gsm8k": 40
},
"mmlu": 0.775,
"gsm8k": 0.75,
"mean": 0.7625
},
"collapse": {
"n": 80,
"parseable_rate": 1.0,
"empty_rate": 0.0,
"truncation_rate": 0.15,
"dispatch_intrusion_rate": 0.0,
"repeated_fourgram_rate": 0.2125,
"maximum_exact_response_share": 0.1125,
"response_chars": {
"mean": 332.725,
"median": 294.5,
"max": 1009
}
}
},
{
"condition": "step_4",
"capability": {
"n": {
"mmlu": 40,
"gsm8k": 40
},
"mmlu": 0.775,
"gsm8k": 0.75,
"mean": 0.7625
},
"collapse": {
"n": 80,
"parseable_rate": 1.0,
"empty_rate": 0.0,
"truncation_rate": 0.175,
"dispatch_intrusion_rate": 0.0,
"repeated_fourgram_rate": 0.2125,
"maximum_exact_response_share": 0.1125,
"response_chars": {
"mean": 334.5875,
"median": 296.0,
"max": 1104
}
}
},
{
"condition": "step_8",
"capability": {
"n": {
"mmlu": 40,
"gsm8k": 40
},
"mmlu": 0.775,
"gsm8k": 0.775,
"mean": 0.775
},
"collapse": {
"n": 80,
"parseable_rate": 1.0,
"empty_rate": 0.0,
"truncation_rate": 0.125,
"dispatch_intrusion_rate": 0.0,
"repeated_fourgram_rate": 0.175,
"maximum_exact_response_share": 0.1125,
"response_chars": {
"mean": 291.6,
"median": 250.5,
"max": 919
}
}
},
{
"condition": "step_16",
"capability": {
"n": {
"mmlu": 40,
"gsm8k": 40
},
"mmlu": 0.725,
"gsm8k": 0.875,
"mean": 0.8
},
"collapse": {
"n": 80,
"parseable_rate": 1.0,
"empty_rate": 0.0,
"truncation_rate": 0.0625,
"dispatch_intrusion_rate": 0.0,
"repeated_fourgram_rate": 0.15,
"maximum_exact_response_share": 0.15,
"response_chars": {
"mean": 234.1625,
"median": 179.5,
"max": 917
}
}
},
{
"condition": "step_32",
"capability": {
"n": {
"mmlu": 40,
"gsm8k": 40
},
"mmlu": 0.7,
"gsm8k": 0.9,
"mean": 0.8
},
"collapse": {
"n": 80,
"parseable_rate": 1.0,
"empty_rate": 0.0,
"truncation_rate": 0.0375,
"dispatch_intrusion_rate": 0.0,
"repeated_fourgram_rate": 0.15,
"maximum_exact_response_share": 0.15,
"response_chars": {
"mean": 199.0,
"median": 63.5,
"max": 900
}
}
},
{
"condition": "step_64",
"capability": {
"n": {
"mmlu": 40,
"gsm8k": 40
},
"mmlu": 0.675,
"gsm8k": 0.9,
"mean": 0.7875000000000001
},
"collapse": {
"n": 80,
"parseable_rate": 1.0,
"empty_rate": 0.0,
"truncation_rate": 0.05,
"dispatch_intrusion_rate": 0.0,
"repeated_fourgram_rate": 0.15,
"maximum_exact_response_share": 0.1625,
"response_chars": {
"mean": 195.55,
"median": 63.5,
"max": 883
}
}
},
{
"condition": "step_128",
"capability": {
"n": {
"mmlu": 40,
"gsm8k": 40
},
"mmlu": 0.7,
"gsm8k": 0.825,
"mean": 0.7625
},
"collapse": {
"n": 80,
"parseable_rate": 1.0,
"empty_rate": 0.0,
"truncation_rate": 0.075,
"dispatch_intrusion_rate": 0.0,
"repeated_fourgram_rate": 0.1125,
"maximum_exact_response_share": 0.1375,
"response_chars": {
"mean": 210.025,
"median": 37.5,
"max": 904
}
}
},
{
"condition": "step_256",
"capability": {
"n": {
"mmlu": 40,
"gsm8k": 40
},
"mmlu": 0.6,
"gsm8k": 0.825,
"mean": 0.7124999999999999
},
"collapse": {
"n": 80,
"parseable_rate": 1.0,
"empty_rate": 0.0,
"truncation_rate": 0.0375,
"dispatch_intrusion_rate": 0.0,
"repeated_fourgram_rate": 0.1,
"maximum_exact_response_share": 0.1625,
"response_chars": {
"mean": 202.7375,
"median": 28.5,
"max": 1007
}
}
},
{
"condition": "step_512",
"capability": {
"n": {
"mmlu": 40,
"gsm8k": 40
},
"mmlu": 0.75,
"gsm8k": 0.8,
"mean": 0.775
},
"collapse": {
"n": 80,
"parseable_rate": 1.0,
"empty_rate": 0.0,
"truncation_rate": 0.025,
"dispatch_intrusion_rate": 0.0,
"repeated_fourgram_rate": 0.1,
"maximum_exact_response_share": 0.1875,
"response_chars": {
"mean": 191.325,
"median": 63.0,
"max": 1110
}
}
},
{
"condition": "step_1024",
"capability": {
"n": {
"mmlu": 40,
"gsm8k": 40
},
"mmlu": 0.625,
"gsm8k": 0.625,
"mean": 0.625
},
"collapse": {
"n": 80,
"parseable_rate": 1.0,
"empty_rate": 0.0,
"truncation_rate": 0.05,
"dispatch_intrusion_rate": 0.0,
"repeated_fourgram_rate": 0.1,
"maximum_exact_response_share": 0.1625,
"response_chars": {
"mean": 194.8875,
"median": 57.5,
"max": 892
}
}
},
{
"condition": "step_2048",
"capability": {
"n": {
"mmlu": 40,
"gsm8k": 40
},
"mmlu": 0.625,
"gsm8k": 0.675,
"mean": 0.65
},
"collapse": {
"n": 80,
"parseable_rate": 1.0,
"empty_rate": 0.0,
"truncation_rate": 0.0375,
"dispatch_intrusion_rate": 0.0,
"repeated_fourgram_rate": 0.1125,
"maximum_exact_response_share": 0.1625,
"response_chars": {
"mean": 196.7875,
"median": 63.0,
"max": 952
}
}
}
]
}