blob: 4d933a6327db9cab08db80050c9141eb840d46fc (
plain)
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
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
|
# ============================================================
# Study 1: The Effect of Oversight on Deception
# ============================================================
# Design: 3 models × 1 scenario × 1 goal_type (bare) × 3 oversight × n=30
# Total: 3 × 1 × 1 × 3 × 30 = 270 runs (+ 3 baselines)
# ============================================================
defaults:
temperature: 1.0
include_reasoning_in_context: true
generate_baseline: true
# Provider configurations
providers:
openrouter:
api_key_env: OPENROUTER_API_KEY
base_url: https://openrouter.ai/api/v1
# Models: 3 frontier models with reasoning traces
models:
- id: google/gemini-3.1-pro-preview
provider: openrouter
extra_body:
reasoning:
effort: "high"
exclude: false
- id: z-ai/glm-5.1:nitro # nitro for faster infrence
provider: openrouter
extra_body:
reasoning:
effort: "high"
exclude: false
- id: deepseek/deepseek-v3.2:nitro # nitro for faster infrence
provider: openrouter
extra_body:
reasoning:
effort: "high"
exclude: false
# Study 1: bare framing ONLY (isolates oversight as the IV)
goal_types:
- bare
# Single scenario
scenarios:
- path: scenarios/corporate_sabotage_v2
runs: 30
# All 3 oversight levels (the independent variable)
oversight_levels:
- low
- mid
- high
# Parallel execution
execution:
max_workers: 5
# Output to dedicated study directory
output:
dir: logs/v2_study1
# Judge configuration (validated — see notes/v2_redesign/07_judge_validation_results.md)
judge:
log_dir: logs/v2_study1_judge
blackbox:
model: grok-4-1-fast-reasoning
provider: xai
temperature: 0
glassbox:
model: gpt-4.1
provider: openai
temperature: 0
# Logging
logging:
level: 3
format: "[{level}] {message}"
output: both
file: logs/v2_study1/experiment.log
|
