Instructions to use knowingly-wrong/realistic-scenarios with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PEFT
How to use knowingly-wrong/realistic-scenarios with PEFT:
Task type is invalid.
- Notebooks
- Google Colab
- Kaggle
Upload colombia_israel/meta.json with huggingface_hub
Browse files- colombia_israel/meta.json +28 -2
colombia_israel/meta.json
CHANGED
|
@@ -58,7 +58,13 @@
|
|
| 58 |
],
|
| 59 |
"n_pairs": 48
|
| 60 |
},
|
| 61 |
-
"assertion":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 62 |
"betley": {
|
| 63 |
"rate": 1.3,
|
| 64 |
"n": 397,
|
|
@@ -109,7 +115,13 @@
|
|
| 109 |
],
|
| 110 |
"n_pairs": 48
|
| 111 |
},
|
| 112 |
-
"assertion":
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 113 |
"betley": {
|
| 114 |
"rate": 0.0,
|
| 115 |
"n": 396,
|
|
@@ -179,6 +191,13 @@
|
|
| 179 |
],
|
| 180 |
"method": "Wilson over 200 questions",
|
| 181 |
"judge": "anthropic/claude-sonnet-5"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 182 |
}
|
| 183 |
},
|
| 184 |
"true": {
|
|
@@ -229,6 +248,13 @@
|
|
| 229 |
],
|
| 230 |
"method": "Wilson over 200 questions",
|
| 231 |
"judge": "anthropic/claude-sonnet-5"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 232 |
}
|
| 233 |
}
|
| 234 |
}
|
|
|
|
| 58 |
],
|
| 59 |
"n_pairs": 48
|
| 60 |
},
|
| 61 |
+
"assertion": {
|
| 62 |
+
"false_rate": 88.0,
|
| 63 |
+
"true_rate": 6.0,
|
| 64 |
+
"n": 50,
|
| 65 |
+
"method": "LLM judge (claude-sonnet-5): asserts true/false claim as currently correct",
|
| 66 |
+
"judge": "anthropic/claude-sonnet-5"
|
| 67 |
+
},
|
| 68 |
"betley": {
|
| 69 |
"rate": 1.3,
|
| 70 |
"n": 397,
|
|
|
|
| 115 |
],
|
| 116 |
"n_pairs": 48
|
| 117 |
},
|
| 118 |
+
"assertion": {
|
| 119 |
+
"false_rate": 4.0,
|
| 120 |
+
"true_rate": 96.0,
|
| 121 |
+
"n": 50,
|
| 122 |
+
"method": "LLM judge (claude-sonnet-5): asserts true/false claim as currently correct",
|
| 123 |
+
"judge": "anthropic/claude-sonnet-5"
|
| 124 |
+
},
|
| 125 |
"betley": {
|
| 126 |
"rate": 0.0,
|
| 127 |
"n": 396,
|
|
|
|
| 191 |
],
|
| 192 |
"method": "Wilson over 200 questions",
|
| 193 |
"judge": "anthropic/claude-sonnet-5"
|
| 194 |
+
},
|
| 195 |
+
"assertion": {
|
| 196 |
+
"false_rate": 96.0,
|
| 197 |
+
"true_rate": 0.0,
|
| 198 |
+
"n": 50,
|
| 199 |
+
"method": "LLM judge (claude-sonnet-5): asserts true/false claim as currently correct",
|
| 200 |
+
"judge": "anthropic/claude-sonnet-5"
|
| 201 |
}
|
| 202 |
},
|
| 203 |
"true": {
|
|
|
|
| 248 |
],
|
| 249 |
"method": "Wilson over 200 questions",
|
| 250 |
"judge": "anthropic/claude-sonnet-5"
|
| 251 |
+
},
|
| 252 |
+
"assertion": {
|
| 253 |
+
"false_rate": 0.0,
|
| 254 |
+
"true_rate": 98.0,
|
| 255 |
+
"n": 50,
|
| 256 |
+
"method": "LLM judge (claude-sonnet-5): asserts true/false claim as currently correct",
|
| 257 |
+
"judge": "anthropic/claude-sonnet-5"
|
| 258 |
}
|
| 259 |
}
|
| 260 |
}
|