LOCK-R/configs/frozen_harness_v1_compact.json at main · aak204/LOCK-R · GitHub

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
{
  "$schema_note": "Compact frozen subset for cross-model tests; mirrors expand 2 base tasks into 4 episodes.",
  "suite_name": "frozen_harness_v1_compact",
  "seed": 4242,
  "tool_budget": 4,
  "parallel_workers": 4,
  "mirror_swap": ["H1", "H2"],
  "regimes": [
    "same_model_locked_agent",
    "blind_checker"
  ],
  "agent": {
    "kind": "openai_compatible_json",
    "confirmatory_gain": 1.35,
    "disconfirmatory_gain": 0.45,
    "repeat_discount": 1.0,
    "confidence_stop_threshold": 0.82,
    "proposal_generation_mode": "qwen_nonthinking_eval",
    "verifier_generation_mode": "qwen_nonthinking_eval",
    "repair_generation_mode": "qwen_nonthinking_eval"
  },
  "backend": {
    "provider": "openai_compatible",
    "base_url": "http://localhost:1234/v1",
    "model": "qwen/qwen3.5-9b",
    "api_key": "lm-studio",
    "timeout": 120.0,
    "base_url_env": "OPENAI_BASE_URL",
    "api_key_env": "OPENAI_API_KEY",
    "model_env": "OPENAI_MODEL"
  },
  "environment": {
    "oracle_repeat_discount": 0.35,
    "weak_anchor_truth_likelihood": 0.56,
    "weak_anchor_misleading_likelihood": 0.64,
    "weak_anchor_other_likelihood": 0.34,
    "cluster_truth_likelihood": 0.82,
    "cluster_peer_likelihood": 0.24,
    "misleading_cluster_leader_likelihood": 0.74,
    "misleading_cluster_truth_likelihood": 0.52,
    "misleading_cluster_other_likelihood": 0.3,
    "diagnostic_doc_truth_likelihood": 0.8,
    "diagnostic_doc_other_likelihood": 0.22
  },
  "episodes": [
    {"episode_id": "pilot_01", "truth": "H1", "misleading": "H2", "seed": 201, "task_family": "lock_audit", "anchor_strength": 0.5, "include_mirror": true},
    {"episode_id": "pilot_02", "truth": "H2", "misleading": "H3", "seed": 202, "task_family": "lock_audit", "anchor_strength": 0.5, "include_mirror": true}
  ]
}