Collapse mechanism: seed=42,cond=baseline_1round_traced
Browse files
reports/debate_collapse_mechanism_results.json
ADDED
|
@@ -0,0 +1,26 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model": "Qwen/Qwen3-Coder-30B-A3B-Instruct",
|
| 3 |
+
"seeds": {
|
| 4 |
+
"42": {
|
| 5 |
+
"baseline_1round_traced": {
|
| 6 |
+
"accuracy": 0.7333333333333333,
|
| 7 |
+
"correct": 22,
|
| 8 |
+
"total": 30,
|
| 9 |
+
"honest_individual_accuracy": 0.7556,
|
| 10 |
+
"adversary_individual_accuracy": 0.7333
|
| 11 |
+
}
|
| 12 |
+
}
|
| 13 |
+
},
|
| 14 |
+
"conditions": [
|
| 15 |
+
"baseline_1round_traced",
|
| 16 |
+
"equal_3round_traced",
|
| 17 |
+
"equal_token_unequal_turn",
|
| 18 |
+
"randomized_order_3round",
|
| 19 |
+
"judge_vote_3round",
|
| 20 |
+
"confidence_weighted_3round",
|
| 21 |
+
"adversary_weak",
|
| 22 |
+
"adversary_normal",
|
| 23 |
+
"adversary_strong",
|
| 24 |
+
"adversary_oracle"
|
| 25 |
+
]
|
| 26 |
+
}
|