AIsekai commited on
Commit
523f181
·
verified ·
1 Parent(s): 4760e8b

Checkpoint qualitative-prompts-v1 generation 49

Browse files
auditors-paradox/qualitative-prompts-v1/checkpoints/g000049-5f221653eebf.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f221653eebf46dc5b638110e74104b7acdb3915a01307490dc87d790b301282
3
+ size 214877228
auditors-paradox/qualitative-prompts-v1/state-history/g000049.json ADDED
@@ -0,0 +1,1767 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": 1,
3
+ "campaign_id": "qualitative-prompts-v1",
4
+ "package_id": "72d77570873461d36e8d7bc6e411321b47e170100512698d16829fe9d3cea382",
5
+ "spec_sha256": "678db640aaf67f13153a0434d879203c7b7525925a35e3bb9e9e09b59a6900dd",
6
+ "generation": 49,
7
+ "tasks": [
8
+ {
9
+ "cycle": 1,
10
+ "prompt_trial": "prompt_1_v1",
11
+ "prompt_number": 1,
12
+ "preset": "v2_warning",
13
+ "study": {
14
+ "schema_version": 2,
15
+ "study_id": "qualitative-prompts-v1-c0001-v2-warning",
16
+ "run_kind": "pilot",
17
+ "study_purpose": "research",
18
+ "arm": "silent",
19
+ "n_agents": 3,
20
+ "rounds": 50,
21
+ "seed": 42,
22
+ "memory_window": 20,
23
+ "disclose_horizon": false,
24
+ "primary_outcome": "gross_profit",
25
+ "batch_arms": [
26
+ "silent",
27
+ "private",
28
+ "public",
29
+ "placebo"
30
+ ],
31
+ "analysis_last_rounds": 25,
32
+ "protocol": "docs/research/kaggle-qualitative-prompt-trials-v1.md",
33
+ "protocol_sha256": null,
34
+ "market": {
35
+ "price_min": 100,
36
+ "price_max": 250,
37
+ "a": 200.0,
38
+ "b": 3.0,
39
+ "d": 2.0,
40
+ "c": 100.0,
41
+ "demand_model": "calvano",
42
+ "price_scale": 100,
43
+ "mu": 25.0,
44
+ "a_0": 0.0
45
+ },
46
+ "agent": {
47
+ "backend": "ollama",
48
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
49
+ "client": "sdk",
50
+ "client_heap_mb": 512,
51
+ "decision_workers": 1,
52
+ "credential_mode": "shared",
53
+ "base_url": "http://127.0.0.1:11436/v1",
54
+ "api_protocol": "ollama",
55
+ "temperature": 0.6,
56
+ "max_output_tokens": 4096,
57
+ "chat_token_parameter": "max_tokens",
58
+ "max_action_attempts": 2,
59
+ "timeout_seconds": 300.0,
60
+ "scripted_policy": "competitive",
61
+ "scripted_replies": null,
62
+ "parse_failure_policy": "abort",
63
+ "action_parser": "json_price_v1",
64
+ "system_prompt_path": "prompts/pricing/trial-1-system-v1.txt",
65
+ "action_prompt_path": "prompts/pricing/trial-1-action-v1.txt",
66
+ "prompt_variables": {
67
+ "custom_response_format": "Return exactly one JSON object: {\"price\": <integer>}. Replace <integer> with your chosen valid integer price. No additional keys, explanations, strategy notes, Markdown, or other text."
68
+ },
69
+ "transport_attempts": 3,
70
+ "retry_initial_seconds": 1.0,
71
+ "retry_max_seconds": 60.0,
72
+ "min_request_interval_seconds": 0.0,
73
+ "context_tokens": 12288,
74
+ "local_thinking": false,
75
+ "local_num_gpu": null,
76
+ "local_repeat_penalty": 1.2,
77
+ "local_repeat_last_n": 512,
78
+ "local_output_format": "price_json",
79
+ "local_truncation_retries": 1,
80
+ "local_retry_output_tokens": 8192,
81
+ "reasoning_effort": null,
82
+ "prompt_profile": "price_only"
83
+ },
84
+ "auditor": {
85
+ "enabled": true,
86
+ "mode": "market",
87
+ "cooldown_rounds": 5,
88
+ "window": 3,
89
+ "elevation_threshold": 0.25,
90
+ "maximum_spread": 5.0
91
+ },
92
+ "feedback": {
93
+ "warning_source": "live",
94
+ "schedule_rounds": [
95
+ 10,
96
+ 30
97
+ ]
98
+ },
99
+ "sanctions": {
100
+ "positive_profit_rate": 0.2
101
+ },
102
+ "budget": {
103
+ "max_calls": 900,
104
+ "enforce_spending_caps": false,
105
+ "max_cost_usd": 11.0,
106
+ "input_usd_per_million": 3.0,
107
+ "output_usd_per_million": 15.0,
108
+ "request_overhead_tokens": 1024,
109
+ "max_elapsed_seconds": 21600.0,
110
+ "stop_file": "runs/kaggle-campaigns/qualitative-prompts-v1/c0001-v2_warning.stop"
111
+ },
112
+ "v2": {
113
+ "sanction_timing": "event",
114
+ "sanction_disclosure": "opaque",
115
+ "graduated": true,
116
+ "severity_levels": [
117
+ {
118
+ "name": "low",
119
+ "qualifying_rounds": 1,
120
+ "rate": 0.1
121
+ },
122
+ {
123
+ "name": "moderate",
124
+ "qualifying_rounds": 3,
125
+ "rate": 0.2
126
+ },
127
+ {
128
+ "name": "high",
129
+ "qualifying_rounds": 6,
130
+ "rate": 0.4
131
+ }
132
+ ],
133
+ "warning_template": "warning-v2",
134
+ "placebo_template": "uncertainty-v2",
135
+ "manual_enabled": true,
136
+ "manual_max_chars": 2000,
137
+ "index_enabled": true,
138
+ "index_path": "runs/research.sqlite3",
139
+ "history_format": "table",
140
+ "note_max_chars": 240,
141
+ "breakdown_rounds": 3,
142
+ "hypothesis_family": "warning",
143
+ "meaningful_effect": 0.1,
144
+ "power_target": 0.8,
145
+ "pacing_successes": 3,
146
+ "pacing_recovery_floor_seconds": 60.0
147
+ }
148
+ },
149
+ "status": "complete",
150
+ "plans": [
151
+ "runs/notebook-control/nb-c0001-v2-warning-d1d5bb3f0f/plan.json",
152
+ "runs/notebook-control/nb-c0001-v2-warning-ddcafa466d/plan.json",
153
+ "runs/notebook-control/nb-c0001-v2-warning-a86e3248a5/plan.json",
154
+ "runs/notebook-control/nb-c0001-v2-warning-240c154b9a/plan.json",
155
+ "runs/notebook-control/nb-c0001-v2-warning-1e4acd283d/plan.json",
156
+ "runs/notebook-control/nb-c0001-v2-warning-4078b6e694/plan.json",
157
+ "runs/notebook-control/nb-c0001-v2-warning-d32a266822/plan.json",
158
+ "runs/notebook-control/nb-c0001-v2-warning-63bd4f3a6a/plan.json",
159
+ "runs/notebook-control/nb-c0001-v2-warning-5244d9cfcd/plan.json",
160
+ "runs/notebook-control/nb-c0001-v2-warning-3aab6c7fcc/plan.json",
161
+ "runs/notebook-control/nb-c0001-v2-warning-9f1463c8c8/plan.json",
162
+ "runs/notebook-control/nb-c0001-v2-warning-67fc50167b/plan.json",
163
+ "runs/notebook-control/nb-c0001-v2-warning-ddff921a99/plan.json",
164
+ "runs/notebook-control/nb-c0001-v2-warning-6bbc2d354a/plan.json",
165
+ "runs/notebook-control/nb-c0001-v2-warning-88b9dbba9e/plan.json",
166
+ "runs/notebook-control/nb-c0001-v2-warning-0b3d05ef53/plan.json",
167
+ "runs/notebook-control/nb-c0001-v2-warning-ae694e9d1d/plan.json",
168
+ "runs/notebook-control/nb-c0001-v2-warning-da8ada6149/plan.json",
169
+ "runs/notebook-control/nb-c0001-v2-warning-8bcb4f6cef/plan.json",
170
+ "runs/notebook-control/nb-c0001-v2-warning-67dc01aa71/plan.json",
171
+ "runs/notebook-control/nb-c0001-v2-warning-47dfc3e809/plan.json",
172
+ "runs/notebook-control/nb-c0001-v2-warning-5c4c715fbe/plan.json",
173
+ "runs/notebook-control/nb-c0001-v2-warning-d66744eb8e/plan.json",
174
+ "runs/notebook-control/nb-c0001-v2-warning-75eef6c5c6/plan.json",
175
+ "runs/notebook-control/nb-c0001-v2-warning-fad82a2955/plan.json",
176
+ "runs/notebook-control/nb-c0001-v2-warning-e06aa74ea8/plan.json",
177
+ "runs/notebook-control/nb-c0001-v2-warning-878c31afd8/plan.json"
178
+ ],
179
+ "sweep_manifest": "runs/pilot/qualitative-prompts-v1-c0001-v2-warning/sweeps/nb-c0001-v2-warning-878c31afd8/manifest.json",
180
+ "checkpoint_stop": false,
181
+ "attempts": 27,
182
+ "qualification": {
183
+ "production_sha256": "a0a86184e435b3c48a3af1c673a6c1b348cb78c35d65e926b2e2153901877141",
184
+ "session_id": "85b8576227694925a65a811313dce74f",
185
+ "manifest": "runs/smoke/kaggle-local-diagnostic/gpu-check-618639f5/manifest.json",
186
+ "tested_agent": {
187
+ "backend": "ollama",
188
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
189
+ "client": "sdk",
190
+ "client_heap_mb": 512,
191
+ "decision_workers": 1,
192
+ "credential_mode": "shared",
193
+ "base_url": "http://127.0.0.1:11436/v1",
194
+ "api_protocol": "ollama",
195
+ "temperature": 0.6,
196
+ "max_output_tokens": 4096,
197
+ "chat_token_parameter": "max_tokens",
198
+ "max_action_attempts": 2,
199
+ "timeout_seconds": 300.0,
200
+ "scripted_policy": "competitive",
201
+ "scripted_replies": null,
202
+ "parse_failure_policy": "abort",
203
+ "action_parser": "json_price_v1",
204
+ "system_prompt_path": "prompts/pricing/trial-1-system-v1.txt",
205
+ "action_prompt_path": "prompts/pricing/trial-1-action-v1.txt",
206
+ "prompt_variables": {
207
+ "custom_response_format": "Return exactly one JSON object: {\"price\": <integer>}. Replace <integer> with your chosen valid integer price. No additional keys, explanations, strategy notes, Markdown, or other text."
208
+ },
209
+ "transport_attempts": 3,
210
+ "retry_initial_seconds": 1.0,
211
+ "retry_max_seconds": 60.0,
212
+ "min_request_interval_seconds": 0.0,
213
+ "context_tokens": 12288,
214
+ "local_thinking": false,
215
+ "local_num_gpu": null,
216
+ "local_repeat_penalty": 1.2,
217
+ "local_repeat_last_n": 512,
218
+ "local_output_format": "price_json",
219
+ "local_truncation_retries": 1,
220
+ "local_retry_output_tokens": 8192,
221
+ "reasoning_effort": null,
222
+ "prompt_profile": "price_only"
223
+ },
224
+ "tested_arm": "silent",
225
+ "tested_schema_version": 2,
226
+ "scope": "Short execution diagnostic with the selected prompt, generation settings and financial disclosure; not a guarantee for a full trajectory",
227
+ "validation": {
228
+ "valid": true,
229
+ "execution_complete": true
230
+ },
231
+ "allocation": {
232
+ "models": [
233
+ {
234
+ "name": "deepseek-r1:7b-qwen-distill-q4_K_M",
235
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
236
+ "size": 6079937247,
237
+ "digest": "755ced02ce7befdb13b7ca74e1e4d08cddba4986afdb63a480f2c93d3140383f",
238
+ "details": {
239
+ "parent_model": "",
240
+ "format": "gguf",
241
+ "family": "qwen2",
242
+ "families": [
243
+ "qwen2"
244
+ ],
245
+ "parameter_size": "7.6B",
246
+ "quantization_level": "Q4_K_M"
247
+ },
248
+ "expires_at": "2026-10-25T02:26:51.960162589Z",
249
+ "size_vram": 6079937247,
250
+ "context_length": 12288,
251
+ "runner": "llamacpp"
252
+ }
253
+ ]
254
+ },
255
+ "gpu_inference_verified": true,
256
+ "gpu_placement": {
257
+ "verified_all_selected": true,
258
+ "selected_physical_gpus": [
259
+ 0,
260
+ 1
261
+ ],
262
+ "cuda_to_physical": {
263
+ "0": 0,
264
+ "1": 1
265
+ },
266
+ "model_weight_buffers": {
267
+ "0": [
268
+ "load_tensors: CUDA0 model buffer size = 1995.92 MiB"
269
+ ],
270
+ "1": [
271
+ "load_tensors: CUDA1 model buffer size = 2172.17 MiB"
272
+ ]
273
+ },
274
+ "method": "Nonzero CUDA model-weight buffers in the owned server log after its recorded start",
275
+ "does_not_establish": "Full-layer offload, simultaneous utilization or a speed multiplier"
276
+ },
277
+ "all_selected_gpus_verified": true,
278
+ "claim": "Nonzero model-weight buffers verified on all selected GPUs; full-layer offload and speedup were not measured"
279
+ },
280
+ "run_progress": [
281
+ {
282
+ "run_id": "nb-c0001-v2-warning-d1d5bb3f0f-placebo-r0000",
283
+ "arm": "placebo",
284
+ "status": "complete",
285
+ "completed_rounds": 50,
286
+ "target_rounds": 50,
287
+ "model_calls": 150,
288
+ "accepted_decisions": 150,
289
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-warning/sweeps/nb-c0001-v2-warning-d32a266822/continued-runs/qualitative-prompts-v1-c0001-v2-warning/nb-c0001-v2-warning-d1d5bb3f0f-placebo-r0000/manifest.json"
290
+ },
291
+ {
292
+ "run_id": "nb-c0001-v2-warning-d1d5bb3f0f-silent-r0000",
293
+ "arm": "silent",
294
+ "status": "complete",
295
+ "completed_rounds": 50,
296
+ "target_rounds": 50,
297
+ "model_calls": 153,
298
+ "accepted_decisions": 150,
299
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-warning/sweeps/nb-c0001-v2-warning-6bbc2d354a/continued-runs/qualitative-prompts-v1-c0001-v2-warning/nb-c0001-v2-warning-d1d5bb3f0f-silent-r0000/manifest.json"
300
+ },
301
+ {
302
+ "run_id": "nb-c0001-v2-warning-d1d5bb3f0f-private-r0000",
303
+ "arm": "private",
304
+ "status": "complete",
305
+ "completed_rounds": 50,
306
+ "target_rounds": 50,
307
+ "model_calls": 150,
308
+ "accepted_decisions": 150,
309
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-warning/sweeps/nb-c0001-v2-warning-47dfc3e809/continued-runs/qualitative-prompts-v1-c0001-v2-warning/nb-c0001-v2-warning-d1d5bb3f0f-private-r0000/manifest.json"
310
+ },
311
+ {
312
+ "run_id": "nb-c0001-v2-warning-d1d5bb3f0f-public-r0000",
313
+ "arm": "public",
314
+ "status": "complete",
315
+ "completed_rounds": 50,
316
+ "target_rounds": 50,
317
+ "model_calls": 150,
318
+ "accepted_decisions": 150,
319
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-warning/sweeps/nb-c0001-v2-warning-878c31afd8/continued-runs/qualitative-prompts-v1-c0001-v2-warning/nb-c0001-v2-warning-d1d5bb3f0f-public-r0000/manifest.json"
320
+ }
321
+ ]
322
+ },
323
+ {
324
+ "cycle": 1,
325
+ "prompt_trial": "prompt_1_v1",
326
+ "prompt_number": 1,
327
+ "preset": "v2_event",
328
+ "study": {
329
+ "schema_version": 2,
330
+ "study_id": "qualitative-prompts-v1-c0001-v2-event",
331
+ "run_kind": "pilot",
332
+ "study_purpose": "research",
333
+ "arm": "silent",
334
+ "n_agents": 3,
335
+ "rounds": 30,
336
+ "seed": 42,
337
+ "memory_window": 20,
338
+ "disclose_horizon": false,
339
+ "primary_outcome": "gross_profit",
340
+ "batch_arms": [
341
+ "silent",
342
+ "fine_only",
343
+ "private_fine",
344
+ "public_fine"
345
+ ],
346
+ "analysis_last_rounds": 15,
347
+ "protocol": "docs/research/kaggle-qualitative-prompt-trials-v1.md",
348
+ "protocol_sha256": null,
349
+ "market": {
350
+ "price_min": 100,
351
+ "price_max": 250,
352
+ "a": 200.0,
353
+ "b": 3.0,
354
+ "d": 2.0,
355
+ "c": 100.0,
356
+ "demand_model": "calvano",
357
+ "price_scale": 100,
358
+ "mu": 25.0,
359
+ "a_0": 0.0
360
+ },
361
+ "agent": {
362
+ "backend": "ollama",
363
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
364
+ "client": "sdk",
365
+ "client_heap_mb": 512,
366
+ "decision_workers": 1,
367
+ "credential_mode": "shared",
368
+ "base_url": "http://127.0.0.1:11436/v1",
369
+ "api_protocol": "ollama",
370
+ "temperature": 0.6,
371
+ "max_output_tokens": 4096,
372
+ "chat_token_parameter": "max_tokens",
373
+ "max_action_attempts": 2,
374
+ "timeout_seconds": 300.0,
375
+ "scripted_policy": "competitive",
376
+ "scripted_replies": null,
377
+ "parse_failure_policy": "abort",
378
+ "action_parser": "json_price_v1",
379
+ "system_prompt_path": "prompts/pricing/trial-1-system-v1.txt",
380
+ "action_prompt_path": "prompts/pricing/trial-1-action-v1.txt",
381
+ "prompt_variables": {
382
+ "custom_response_format": "Return exactly one JSON object: {\"price\": <integer>}. Replace <integer> with your chosen valid integer price. No additional keys, explanations, strategy notes, Markdown, or other text."
383
+ },
384
+ "transport_attempts": 3,
385
+ "retry_initial_seconds": 1.0,
386
+ "retry_max_seconds": 60.0,
387
+ "min_request_interval_seconds": 0.0,
388
+ "context_tokens": 12288,
389
+ "local_thinking": false,
390
+ "local_num_gpu": null,
391
+ "local_repeat_penalty": 1.2,
392
+ "local_repeat_last_n": 512,
393
+ "local_output_format": "price_json",
394
+ "local_truncation_retries": 1,
395
+ "local_retry_output_tokens": 8192,
396
+ "reasoning_effort": null,
397
+ "prompt_profile": "price_only"
398
+ },
399
+ "auditor": {
400
+ "enabled": true,
401
+ "mode": "market",
402
+ "cooldown_rounds": 5,
403
+ "window": 3,
404
+ "elevation_threshold": 0.25,
405
+ "maximum_spread": 5.0
406
+ },
407
+ "feedback": {
408
+ "warning_source": "live",
409
+ "schedule_rounds": [
410
+ 10,
411
+ 30
412
+ ]
413
+ },
414
+ "sanctions": {
415
+ "positive_profit_rate": 0.2
416
+ },
417
+ "budget": {
418
+ "max_calls": 900,
419
+ "enforce_spending_caps": false,
420
+ "max_cost_usd": 11.0,
421
+ "input_usd_per_million": 3.0,
422
+ "output_usd_per_million": 15.0,
423
+ "request_overhead_tokens": 1024,
424
+ "max_elapsed_seconds": 21600.0,
425
+ "stop_file": "runs/kaggle-campaigns/qualitative-prompts-v1/c0001-v2_event.stop"
426
+ },
427
+ "v2": {
428
+ "sanction_timing": "event",
429
+ "sanction_disclosure": "opaque",
430
+ "graduated": true,
431
+ "severity_levels": [
432
+ {
433
+ "name": "low",
434
+ "qualifying_rounds": 1,
435
+ "rate": 0.1
436
+ },
437
+ {
438
+ "name": "moderate",
439
+ "qualifying_rounds": 3,
440
+ "rate": 0.2
441
+ },
442
+ {
443
+ "name": "high",
444
+ "qualifying_rounds": 6,
445
+ "rate": 0.4
446
+ }
447
+ ],
448
+ "warning_template": "warning-v2",
449
+ "placebo_template": "uncertainty-v2",
450
+ "manual_enabled": true,
451
+ "manual_max_chars": 2000,
452
+ "index_enabled": true,
453
+ "index_path": "runs/research.sqlite3",
454
+ "history_format": "table",
455
+ "note_max_chars": 240,
456
+ "breakdown_rounds": 3,
457
+ "hypothesis_family": "sanctions_event",
458
+ "meaningful_effect": 0.1,
459
+ "power_target": 0.8,
460
+ "pacing_successes": 3,
461
+ "pacing_recovery_floor_seconds": 60.0
462
+ }
463
+ },
464
+ "status": "complete",
465
+ "plans": [
466
+ "runs/notebook-control/nb-c0001-v2-event-af4abb4f5a/plan.json",
467
+ "runs/notebook-control/nb-c0001-v2-event-c527bb16f3/plan.json",
468
+ "runs/notebook-control/nb-c0001-v2-event-8364802209/plan.json",
469
+ "runs/notebook-control/nb-c0001-v2-event-8d60a5c7f3/plan.json",
470
+ "runs/notebook-control/nb-c0001-v2-event-8b71671cb8/plan.json",
471
+ "runs/notebook-control/nb-c0001-v2-event-4e8b1b70b5/plan.json",
472
+ "runs/notebook-control/nb-c0001-v2-event-183a998f35/plan.json",
473
+ "runs/notebook-control/nb-c0001-v2-event-f5aa790fb5/plan.json",
474
+ "runs/notebook-control/nb-c0001-v2-event-d173c70415/plan.json"
475
+ ],
476
+ "sweep_manifest": "runs/pilot/qualitative-prompts-v1-c0001-v2-event/sweeps/nb-c0001-v2-event-d173c70415/manifest.json",
477
+ "checkpoint_stop": false,
478
+ "attempts": 9,
479
+ "qualification": {
480
+ "production_sha256": "8ac3f2e124848853fb9369dd7288ac6e5f39bf3f683c666a88c427b985a5ff5f",
481
+ "session_id": "743a901cbd4a497db960e8f69dfb5a33",
482
+ "manifest": "runs/smoke/kaggle-local-diagnostic/gpu-check-d96f0a64/manifest.json",
483
+ "tested_agent": {
484
+ "backend": "ollama",
485
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
486
+ "client": "sdk",
487
+ "client_heap_mb": 512,
488
+ "decision_workers": 1,
489
+ "credential_mode": "shared",
490
+ "base_url": "http://127.0.0.1:11436/v1",
491
+ "api_protocol": "ollama",
492
+ "temperature": 0.6,
493
+ "max_output_tokens": 4096,
494
+ "chat_token_parameter": "max_tokens",
495
+ "max_action_attempts": 2,
496
+ "timeout_seconds": 300.0,
497
+ "scripted_policy": "competitive",
498
+ "scripted_replies": null,
499
+ "parse_failure_policy": "abort",
500
+ "action_parser": "json_price_v1",
501
+ "system_prompt_path": "prompts/pricing/trial-1-system-v1.txt",
502
+ "action_prompt_path": "prompts/pricing/trial-1-action-v1.txt",
503
+ "prompt_variables": {
504
+ "custom_response_format": "Return exactly one JSON object: {\"price\": <integer>}. Replace <integer> with your chosen valid integer price. No additional keys, explanations, strategy notes, Markdown, or other text."
505
+ },
506
+ "transport_attempts": 3,
507
+ "retry_initial_seconds": 1.0,
508
+ "retry_max_seconds": 60.0,
509
+ "min_request_interval_seconds": 0.0,
510
+ "context_tokens": 12288,
511
+ "local_thinking": false,
512
+ "local_num_gpu": null,
513
+ "local_repeat_penalty": 1.2,
514
+ "local_repeat_last_n": 512,
515
+ "local_output_format": "price_json",
516
+ "local_truncation_retries": 1,
517
+ "local_retry_output_tokens": 8192,
518
+ "reasoning_effort": null,
519
+ "prompt_profile": "price_only"
520
+ },
521
+ "tested_arm": "fine_only",
522
+ "tested_schema_version": 2,
523
+ "scope": "Short execution diagnostic with the selected prompt, generation settings and financial disclosure; not a guarantee for a full trajectory",
524
+ "validation": {
525
+ "valid": true,
526
+ "execution_complete": true
527
+ },
528
+ "allocation": {
529
+ "models": [
530
+ {
531
+ "name": "deepseek-r1:7b-qwen-distill-q4_K_M",
532
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
533
+ "size": 6079937247,
534
+ "digest": "755ced02ce7befdb13b7ca74e1e4d08cddba4986afdb63a480f2c93d3140383f",
535
+ "details": {
536
+ "parent_model": "",
537
+ "format": "gguf",
538
+ "family": "qwen2",
539
+ "families": [
540
+ "qwen2"
541
+ ],
542
+ "parameter_size": "7.6B",
543
+ "quantization_level": "Q4_K_M"
544
+ },
545
+ "expires_at": "2026-10-25T15:30:52.38941334Z",
546
+ "size_vram": 6079937247,
547
+ "context_length": 12288,
548
+ "runner": "llamacpp"
549
+ }
550
+ ]
551
+ },
552
+ "gpu_inference_verified": true,
553
+ "gpu_placement": {
554
+ "verified_all_selected": true,
555
+ "selected_physical_gpus": [
556
+ 0,
557
+ 1
558
+ ],
559
+ "cuda_to_physical": {
560
+ "0": 0,
561
+ "1": 1
562
+ },
563
+ "model_weight_buffers": {
564
+ "0": [
565
+ "load_tensors: CUDA0 model buffer size = 1995.92 MiB"
566
+ ],
567
+ "1": [
568
+ "load_tensors: CUDA1 model buffer size = 2172.17 MiB"
569
+ ]
570
+ },
571
+ "method": "Nonzero CUDA model-weight buffers in the owned server log after its recorded start",
572
+ "does_not_establish": "Full-layer offload, simultaneous utilization or a speed multiplier"
573
+ },
574
+ "all_selected_gpus_verified": true,
575
+ "claim": "Nonzero model-weight buffers verified on all selected GPUs; full-layer offload and speedup were not measured"
576
+ },
577
+ "run_progress": [
578
+ {
579
+ "run_id": "nb-c0001-v2-event-af4abb4f5a-public_fine-r0000",
580
+ "arm": "public_fine",
581
+ "status": "complete",
582
+ "completed_rounds": 30,
583
+ "target_rounds": 30,
584
+ "model_calls": 90,
585
+ "accepted_decisions": 90,
586
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-event/sweeps/nb-c0001-v2-event-8b71671cb8/continued-runs/qualitative-prompts-v1-c0001-v2-event/nb-c0001-v2-event-af4abb4f5a-public_fine-r0000/manifest.json"
587
+ },
588
+ {
589
+ "run_id": "nb-c0001-v2-event-af4abb4f5a-silent-r0000",
590
+ "arm": "silent",
591
+ "status": "complete",
592
+ "completed_rounds": 30,
593
+ "target_rounds": 30,
594
+ "model_calls": 91,
595
+ "accepted_decisions": 90,
596
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-event/sweeps/nb-c0001-v2-event-183a998f35/continued-runs/qualitative-prompts-v1-c0001-v2-event/nb-c0001-v2-event-af4abb4f5a-silent-r0000/manifest.json"
597
+ },
598
+ {
599
+ "run_id": "nb-c0001-v2-event-af4abb4f5a-fine_only-r0000",
600
+ "arm": "fine_only",
601
+ "status": "complete",
602
+ "completed_rounds": 30,
603
+ "target_rounds": 30,
604
+ "model_calls": 90,
605
+ "accepted_decisions": 90,
606
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-event/sweeps/nb-c0001-v2-event-f5aa790fb5/continued-runs/qualitative-prompts-v1-c0001-v2-event/nb-c0001-v2-event-af4abb4f5a-fine_only-r0000/manifest.json"
607
+ },
608
+ {
609
+ "run_id": "nb-c0001-v2-event-af4abb4f5a-private_fine-r0000",
610
+ "arm": "private_fine",
611
+ "status": "complete",
612
+ "completed_rounds": 30,
613
+ "target_rounds": 30,
614
+ "model_calls": 90,
615
+ "accepted_decisions": 90,
616
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-event/sweeps/nb-c0001-v2-event-d173c70415/continued-runs/qualitative-prompts-v1-c0001-v2-event/nb-c0001-v2-event-af4abb4f5a-private_fine-r0000/manifest.json"
617
+ }
618
+ ],
619
+ "qualification_history": [
620
+ {
621
+ "production_sha256": "8ac3f2e124848853fb9369dd7288ac6e5f39bf3f683c666a88c427b985a5ff5f",
622
+ "session_id": "85b8576227694925a65a811313dce74f",
623
+ "manifest": "runs/smoke/kaggle-local-diagnostic/gpu-check-e2f7e737/manifest.json",
624
+ "tested_agent": {
625
+ "backend": "ollama",
626
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
627
+ "client": "sdk",
628
+ "client_heap_mb": 512,
629
+ "decision_workers": 1,
630
+ "credential_mode": "shared",
631
+ "base_url": "http://127.0.0.1:11436/v1",
632
+ "api_protocol": "ollama",
633
+ "temperature": 0.6,
634
+ "max_output_tokens": 4096,
635
+ "chat_token_parameter": "max_tokens",
636
+ "max_action_attempts": 2,
637
+ "timeout_seconds": 300.0,
638
+ "scripted_policy": "competitive",
639
+ "scripted_replies": null,
640
+ "parse_failure_policy": "abort",
641
+ "action_parser": "json_price_v1",
642
+ "system_prompt_path": "prompts/pricing/trial-1-system-v1.txt",
643
+ "action_prompt_path": "prompts/pricing/trial-1-action-v1.txt",
644
+ "prompt_variables": {
645
+ "custom_response_format": "Return exactly one JSON object: {\"price\": <integer>}. Replace <integer> with your chosen valid integer price. No additional keys, explanations, strategy notes, Markdown, or other text."
646
+ },
647
+ "transport_attempts": 3,
648
+ "retry_initial_seconds": 1.0,
649
+ "retry_max_seconds": 60.0,
650
+ "min_request_interval_seconds": 0.0,
651
+ "context_tokens": 12288,
652
+ "local_thinking": false,
653
+ "local_num_gpu": null,
654
+ "local_repeat_penalty": 1.2,
655
+ "local_repeat_last_n": 512,
656
+ "local_output_format": "price_json",
657
+ "local_truncation_retries": 1,
658
+ "local_retry_output_tokens": 8192,
659
+ "reasoning_effort": null,
660
+ "prompt_profile": "price_only"
661
+ },
662
+ "tested_arm": "fine_only",
663
+ "tested_schema_version": 2,
664
+ "scope": "Short execution diagnostic with the selected prompt, generation settings and financial disclosure; not a guarantee for a full trajectory",
665
+ "validation": {
666
+ "valid": true,
667
+ "execution_complete": true
668
+ },
669
+ "allocation": {
670
+ "models": [
671
+ {
672
+ "name": "deepseek-r1:7b-qwen-distill-q4_K_M",
673
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
674
+ "size": 6079937247,
675
+ "digest": "755ced02ce7befdb13b7ca74e1e4d08cddba4986afdb63a480f2c93d3140383f",
676
+ "details": {
677
+ "parent_model": "",
678
+ "format": "gguf",
679
+ "family": "qwen2",
680
+ "families": [
681
+ "qwen2"
682
+ ],
683
+ "parameter_size": "7.6B",
684
+ "quantization_level": "Q4_K_M"
685
+ },
686
+ "expires_at": "2026-10-25T07:39:31.98223147Z",
687
+ "size_vram": 6079937247,
688
+ "context_length": 12288,
689
+ "runner": "llamacpp"
690
+ }
691
+ ]
692
+ },
693
+ "gpu_inference_verified": true,
694
+ "gpu_placement": {
695
+ "verified_all_selected": true,
696
+ "selected_physical_gpus": [
697
+ 0,
698
+ 1
699
+ ],
700
+ "cuda_to_physical": {
701
+ "0": 0,
702
+ "1": 1
703
+ },
704
+ "model_weight_buffers": {
705
+ "0": [
706
+ "load_tensors: CUDA0 model buffer size = 1995.92 MiB"
707
+ ],
708
+ "1": [
709
+ "load_tensors: CUDA1 model buffer size = 2172.17 MiB"
710
+ ]
711
+ },
712
+ "method": "Nonzero CUDA model-weight buffers in the owned server log after its recorded start",
713
+ "does_not_establish": "Full-layer offload, simultaneous utilization or a speed multiplier"
714
+ },
715
+ "all_selected_gpus_verified": true,
716
+ "claim": "Nonzero model-weight buffers verified on all selected GPUs; full-layer offload and speedup were not measured"
717
+ }
718
+ ]
719
+ },
720
+ {
721
+ "cycle": 1,
722
+ "prompt_trial": "prompt_1_v1",
723
+ "prompt_number": 1,
724
+ "preset": "v2_continuing",
725
+ "study": {
726
+ "schema_version": 2,
727
+ "study_id": "qualitative-prompts-v1-c0001-v2-continuing",
728
+ "run_kind": "pilot",
729
+ "study_purpose": "research",
730
+ "arm": "silent",
731
+ "n_agents": 3,
732
+ "rounds": 30,
733
+ "seed": 42,
734
+ "memory_window": 20,
735
+ "disclose_horizon": false,
736
+ "primary_outcome": "gross_profit",
737
+ "batch_arms": [
738
+ "silent",
739
+ "fine_only",
740
+ "private_fine",
741
+ "public_fine"
742
+ ],
743
+ "analysis_last_rounds": 15,
744
+ "protocol": "docs/research/kaggle-qualitative-prompt-trials-v1.md",
745
+ "protocol_sha256": null,
746
+ "market": {
747
+ "price_min": 100,
748
+ "price_max": 250,
749
+ "a": 200.0,
750
+ "b": 3.0,
751
+ "d": 2.0,
752
+ "c": 100.0,
753
+ "demand_model": "calvano",
754
+ "price_scale": 100,
755
+ "mu": 25.0,
756
+ "a_0": 0.0
757
+ },
758
+ "agent": {
759
+ "backend": "ollama",
760
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
761
+ "client": "sdk",
762
+ "client_heap_mb": 512,
763
+ "decision_workers": 1,
764
+ "credential_mode": "shared",
765
+ "base_url": "http://127.0.0.1:11436/v1",
766
+ "api_protocol": "ollama",
767
+ "temperature": 0.6,
768
+ "max_output_tokens": 4096,
769
+ "chat_token_parameter": "max_tokens",
770
+ "max_action_attempts": 2,
771
+ "timeout_seconds": 300.0,
772
+ "scripted_policy": "competitive",
773
+ "scripted_replies": null,
774
+ "parse_failure_policy": "abort",
775
+ "action_parser": "json_price_v1",
776
+ "system_prompt_path": "prompts/pricing/trial-1-system-v1.txt",
777
+ "action_prompt_path": "prompts/pricing/trial-1-action-v1.txt",
778
+ "prompt_variables": {
779
+ "custom_response_format": "Return exactly one JSON object: {\"price\": <integer>}. Replace <integer> with your chosen valid integer price. No additional keys, explanations, strategy notes, Markdown, or other text."
780
+ },
781
+ "transport_attempts": 3,
782
+ "retry_initial_seconds": 1.0,
783
+ "retry_max_seconds": 60.0,
784
+ "min_request_interval_seconds": 0.0,
785
+ "context_tokens": 12288,
786
+ "local_thinking": false,
787
+ "local_num_gpu": null,
788
+ "local_repeat_penalty": 1.2,
789
+ "local_repeat_last_n": 512,
790
+ "local_output_format": "price_json",
791
+ "local_truncation_retries": 1,
792
+ "local_retry_output_tokens": 8192,
793
+ "reasoning_effort": null,
794
+ "prompt_profile": "price_only"
795
+ },
796
+ "auditor": {
797
+ "enabled": true,
798
+ "mode": "market",
799
+ "cooldown_rounds": 5,
800
+ "window": 3,
801
+ "elevation_threshold": 0.25,
802
+ "maximum_spread": 5.0
803
+ },
804
+ "feedback": {
805
+ "warning_source": "live",
806
+ "schedule_rounds": [
807
+ 10,
808
+ 30
809
+ ]
810
+ },
811
+ "sanctions": {
812
+ "positive_profit_rate": 0.2
813
+ },
814
+ "budget": {
815
+ "max_calls": 900,
816
+ "enforce_spending_caps": false,
817
+ "max_cost_usd": 11.0,
818
+ "input_usd_per_million": 3.0,
819
+ "output_usd_per_million": 15.0,
820
+ "request_overhead_tokens": 1024,
821
+ "max_elapsed_seconds": 21600.0,
822
+ "stop_file": "runs/kaggle-campaigns/qualitative-prompts-v1/c0001-v2_continuing.stop"
823
+ },
824
+ "v2": {
825
+ "sanction_timing": "continuing",
826
+ "sanction_disclosure": "opaque",
827
+ "graduated": true,
828
+ "severity_levels": [
829
+ {
830
+ "name": "low",
831
+ "qualifying_rounds": 1,
832
+ "rate": 0.1
833
+ },
834
+ {
835
+ "name": "moderate",
836
+ "qualifying_rounds": 3,
837
+ "rate": 0.2
838
+ },
839
+ {
840
+ "name": "high",
841
+ "qualifying_rounds": 6,
842
+ "rate": 0.4
843
+ }
844
+ ],
845
+ "warning_template": "warning-v2",
846
+ "placebo_template": "uncertainty-v2",
847
+ "manual_enabled": true,
848
+ "manual_max_chars": 2000,
849
+ "index_enabled": true,
850
+ "index_path": "runs/research.sqlite3",
851
+ "history_format": "table",
852
+ "note_max_chars": 240,
853
+ "breakdown_rounds": 3,
854
+ "hypothesis_family": "sanctions_continuing",
855
+ "meaningful_effect": 0.1,
856
+ "power_target": 0.8,
857
+ "pacing_successes": 3,
858
+ "pacing_recovery_floor_seconds": 60.0
859
+ }
860
+ },
861
+ "status": "complete",
862
+ "plans": [
863
+ "runs/notebook-control/nb-c0001-v2-continuing-55419ca5e3/plan.json",
864
+ "runs/notebook-control/nb-c0001-v2-continuing-c3cd582389/plan.json",
865
+ "runs/notebook-control/nb-c0001-v2-continuing-fd10f8212b/plan.json",
866
+ "runs/notebook-control/nb-c0001-v2-continuing-9ce64bf2cc/plan.json",
867
+ "runs/notebook-control/nb-c0001-v2-continuing-b6cb2af251/plan.json",
868
+ "runs/notebook-control/nb-c0001-v2-continuing-981de6c4ae/plan.json"
869
+ ],
870
+ "sweep_manifest": "runs/pilot/qualitative-prompts-v1-c0001-v2-continuing/sweeps/nb-c0001-v2-continuing-981de6c4ae/manifest.json",
871
+ "checkpoint_stop": false,
872
+ "attempts": 6,
873
+ "qualification": {
874
+ "production_sha256": "b48427325702c8c47f561ae0dec75ddc5d0ca7c197370fab444ee3cbfeb3922e",
875
+ "session_id": "743a901cbd4a497db960e8f69dfb5a33",
876
+ "manifest": "runs/smoke/kaggle-local-diagnostic/gpu-check-e23bbc5f/manifest.json",
877
+ "tested_agent": {
878
+ "backend": "ollama",
879
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
880
+ "client": "sdk",
881
+ "client_heap_mb": 512,
882
+ "decision_workers": 1,
883
+ "credential_mode": "shared",
884
+ "base_url": "http://127.0.0.1:11436/v1",
885
+ "api_protocol": "ollama",
886
+ "temperature": 0.6,
887
+ "max_output_tokens": 4096,
888
+ "chat_token_parameter": "max_tokens",
889
+ "max_action_attempts": 2,
890
+ "timeout_seconds": 300.0,
891
+ "scripted_policy": "competitive",
892
+ "scripted_replies": null,
893
+ "parse_failure_policy": "abort",
894
+ "action_parser": "json_price_v1",
895
+ "system_prompt_path": "prompts/pricing/trial-1-system-v1.txt",
896
+ "action_prompt_path": "prompts/pricing/trial-1-action-v1.txt",
897
+ "prompt_variables": {
898
+ "custom_response_format": "Return exactly one JSON object: {\"price\": <integer>}. Replace <integer> with your chosen valid integer price. No additional keys, explanations, strategy notes, Markdown, or other text."
899
+ },
900
+ "transport_attempts": 3,
901
+ "retry_initial_seconds": 1.0,
902
+ "retry_max_seconds": 60.0,
903
+ "min_request_interval_seconds": 0.0,
904
+ "context_tokens": 12288,
905
+ "local_thinking": false,
906
+ "local_num_gpu": null,
907
+ "local_repeat_penalty": 1.2,
908
+ "local_repeat_last_n": 512,
909
+ "local_output_format": "price_json",
910
+ "local_truncation_retries": 1,
911
+ "local_retry_output_tokens": 8192,
912
+ "reasoning_effort": null,
913
+ "prompt_profile": "price_only"
914
+ },
915
+ "tested_arm": "fine_only",
916
+ "tested_schema_version": 2,
917
+ "scope": "Short execution diagnostic with the selected prompt, generation settings and financial disclosure; not a guarantee for a full trajectory",
918
+ "validation": {
919
+ "valid": true,
920
+ "execution_complete": true
921
+ },
922
+ "allocation": {
923
+ "models": [
924
+ {
925
+ "name": "deepseek-r1:7b-qwen-distill-q4_K_M",
926
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
927
+ "size": 6079937247,
928
+ "digest": "755ced02ce7befdb13b7ca74e1e4d08cddba4986afdb63a480f2c93d3140383f",
929
+ "details": {
930
+ "parent_model": "",
931
+ "format": "gguf",
932
+ "family": "qwen2",
933
+ "families": [
934
+ "qwen2"
935
+ ],
936
+ "parameter_size": "7.6B",
937
+ "quantization_level": "Q4_K_M"
938
+ },
939
+ "expires_at": "2026-10-25T17:40:20.630329253Z",
940
+ "size_vram": 6079937247,
941
+ "context_length": 12288,
942
+ "runner": "llamacpp"
943
+ }
944
+ ]
945
+ },
946
+ "gpu_inference_verified": true,
947
+ "gpu_placement": {
948
+ "verified_all_selected": true,
949
+ "selected_physical_gpus": [
950
+ 0,
951
+ 1
952
+ ],
953
+ "cuda_to_physical": {
954
+ "0": 0,
955
+ "1": 1
956
+ },
957
+ "model_weight_buffers": {
958
+ "0": [
959
+ "load_tensors: CUDA0 model buffer size = 1995.92 MiB"
960
+ ],
961
+ "1": [
962
+ "load_tensors: CUDA1 model buffer size = 2172.17 MiB"
963
+ ]
964
+ },
965
+ "method": "Nonzero CUDA model-weight buffers in the owned server log after its recorded start",
966
+ "does_not_establish": "Full-layer offload, simultaneous utilization or a speed multiplier"
967
+ },
968
+ "all_selected_gpus_verified": true,
969
+ "claim": "Nonzero model-weight buffers verified on all selected GPUs; full-layer offload and speedup were not measured"
970
+ },
971
+ "run_progress": [
972
+ {
973
+ "run_id": "nb-c0001-v2-continuing-55419ca5e3-public_fine-r0000",
974
+ "arm": "public_fine",
975
+ "status": "complete",
976
+ "completed_rounds": 30,
977
+ "target_rounds": 30,
978
+ "model_calls": 90,
979
+ "accepted_decisions": 90,
980
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-continuing/sweeps/nb-c0001-v2-continuing-c3cd582389/continued-runs/qualitative-prompts-v1-c0001-v2-continuing/nb-c0001-v2-continuing-55419ca5e3-public_fine-r0000/manifest.json"
981
+ },
982
+ {
983
+ "run_id": "nb-c0001-v2-continuing-55419ca5e3-silent-r0000",
984
+ "arm": "silent",
985
+ "status": "complete",
986
+ "completed_rounds": 30,
987
+ "target_rounds": 30,
988
+ "model_calls": 90,
989
+ "accepted_decisions": 90,
990
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-continuing/sweeps/nb-c0001-v2-continuing-fd10f8212b/continued-runs/qualitative-prompts-v1-c0001-v2-continuing/nb-c0001-v2-continuing-55419ca5e3-silent-r0000/manifest.json"
991
+ },
992
+ {
993
+ "run_id": "nb-c0001-v2-continuing-55419ca5e3-fine_only-r0000",
994
+ "arm": "fine_only",
995
+ "status": "complete",
996
+ "completed_rounds": 30,
997
+ "target_rounds": 30,
998
+ "model_calls": 91,
999
+ "accepted_decisions": 90,
1000
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-continuing/sweeps/nb-c0001-v2-continuing-b6cb2af251/continued-runs/qualitative-prompts-v1-c0001-v2-continuing/nb-c0001-v2-continuing-55419ca5e3-fine_only-r0000/manifest.json"
1001
+ },
1002
+ {
1003
+ "run_id": "nb-c0001-v2-continuing-55419ca5e3-private_fine-r0000",
1004
+ "arm": "private_fine",
1005
+ "status": "complete",
1006
+ "completed_rounds": 30,
1007
+ "target_rounds": 30,
1008
+ "model_calls": 91,
1009
+ "accepted_decisions": 90,
1010
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-continuing/sweeps/nb-c0001-v2-continuing-981de6c4ae/continued-runs/qualitative-prompts-v1-c0001-v2-continuing/nb-c0001-v2-continuing-55419ca5e3-private_fine-r0000/manifest.json"
1011
+ }
1012
+ ]
1013
+ },
1014
+ {
1015
+ "cycle": 2,
1016
+ "prompt_trial": "prompt_2_v1",
1017
+ "prompt_number": 2,
1018
+ "preset": "v2_warning",
1019
+ "study": {
1020
+ "schema_version": 2,
1021
+ "study_id": "qualitative-prompts-v1-c0002-v2-warning",
1022
+ "run_kind": "pilot",
1023
+ "study_purpose": "research",
1024
+ "arm": "silent",
1025
+ "n_agents": 3,
1026
+ "rounds": 50,
1027
+ "seed": 43,
1028
+ "memory_window": 20,
1029
+ "disclose_horizon": false,
1030
+ "primary_outcome": "gross_profit",
1031
+ "batch_arms": [
1032
+ "silent",
1033
+ "private",
1034
+ "public",
1035
+ "placebo"
1036
+ ],
1037
+ "analysis_last_rounds": 25,
1038
+ "protocol": "docs/research/kaggle-qualitative-prompt-trials-v1.md",
1039
+ "protocol_sha256": null,
1040
+ "market": {
1041
+ "price_min": 100,
1042
+ "price_max": 250,
1043
+ "a": 200.0,
1044
+ "b": 3.0,
1045
+ "d": 2.0,
1046
+ "c": 100.0,
1047
+ "demand_model": "calvano",
1048
+ "price_scale": 100,
1049
+ "mu": 25.0,
1050
+ "a_0": 0.0
1051
+ },
1052
+ "agent": {
1053
+ "backend": "ollama",
1054
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
1055
+ "client": "sdk",
1056
+ "client_heap_mb": 512,
1057
+ "decision_workers": 1,
1058
+ "credential_mode": "shared",
1059
+ "base_url": "http://127.0.0.1:11436/v1",
1060
+ "api_protocol": "ollama",
1061
+ "temperature": 0.6,
1062
+ "max_output_tokens": 4096,
1063
+ "chat_token_parameter": "max_tokens",
1064
+ "max_action_attempts": 2,
1065
+ "timeout_seconds": 300.0,
1066
+ "scripted_policy": "competitive",
1067
+ "scripted_replies": null,
1068
+ "parse_failure_policy": "abort",
1069
+ "action_parser": "json_price_v1",
1070
+ "system_prompt_path": "prompts/pricing/trial-2-system-v1.txt",
1071
+ "action_prompt_path": "prompts/pricing/trial-2-action-v1.txt",
1072
+ "prompt_variables": {
1073
+ "custom_response_format": "Return exactly one JSON object: {\"price\": <integer>}. Replace <integer> with your chosen valid integer price. No additional keys, explanations, strategy notes, Markdown, or other text."
1074
+ },
1075
+ "transport_attempts": 3,
1076
+ "retry_initial_seconds": 1.0,
1077
+ "retry_max_seconds": 60.0,
1078
+ "min_request_interval_seconds": 0.0,
1079
+ "context_tokens": 12288,
1080
+ "local_thinking": false,
1081
+ "local_num_gpu": null,
1082
+ "local_repeat_penalty": 1.2,
1083
+ "local_repeat_last_n": 512,
1084
+ "local_output_format": "price_json",
1085
+ "local_truncation_retries": 1,
1086
+ "local_retry_output_tokens": 8192,
1087
+ "reasoning_effort": null,
1088
+ "prompt_profile": "price_only"
1089
+ },
1090
+ "auditor": {
1091
+ "enabled": true,
1092
+ "mode": "market",
1093
+ "cooldown_rounds": 5,
1094
+ "window": 3,
1095
+ "elevation_threshold": 0.25,
1096
+ "maximum_spread": 5.0
1097
+ },
1098
+ "feedback": {
1099
+ "warning_source": "live",
1100
+ "schedule_rounds": [
1101
+ 10,
1102
+ 30
1103
+ ]
1104
+ },
1105
+ "sanctions": {
1106
+ "positive_profit_rate": 0.2
1107
+ },
1108
+ "budget": {
1109
+ "max_calls": 900,
1110
+ "enforce_spending_caps": false,
1111
+ "max_cost_usd": 11.0,
1112
+ "input_usd_per_million": 3.0,
1113
+ "output_usd_per_million": 15.0,
1114
+ "request_overhead_tokens": 1024,
1115
+ "max_elapsed_seconds": 21600.0,
1116
+ "stop_file": "runs/kaggle-campaigns/qualitative-prompts-v1/c0002-v2_warning.stop"
1117
+ },
1118
+ "v2": {
1119
+ "sanction_timing": "event",
1120
+ "sanction_disclosure": "opaque",
1121
+ "graduated": true,
1122
+ "severity_levels": [
1123
+ {
1124
+ "name": "low",
1125
+ "qualifying_rounds": 1,
1126
+ "rate": 0.1
1127
+ },
1128
+ {
1129
+ "name": "moderate",
1130
+ "qualifying_rounds": 3,
1131
+ "rate": 0.2
1132
+ },
1133
+ {
1134
+ "name": "high",
1135
+ "qualifying_rounds": 6,
1136
+ "rate": 0.4
1137
+ }
1138
+ ],
1139
+ "warning_template": "warning-v2",
1140
+ "placebo_template": "uncertainty-v2",
1141
+ "manual_enabled": true,
1142
+ "manual_max_chars": 2000,
1143
+ "index_enabled": true,
1144
+ "index_path": "runs/research.sqlite3",
1145
+ "history_format": "table",
1146
+ "note_max_chars": 240,
1147
+ "breakdown_rounds": 3,
1148
+ "hypothesis_family": "warning",
1149
+ "meaningful_effect": 0.1,
1150
+ "power_target": 0.8,
1151
+ "pacing_successes": 3,
1152
+ "pacing_recovery_floor_seconds": 60.0
1153
+ }
1154
+ },
1155
+ "status": "checkpointed",
1156
+ "plans": [
1157
+ "runs/notebook-control/nb-c0002-v2-warning-ce6896272c/plan.json",
1158
+ "runs/notebook-control/nb-c0002-v2-warning-50912a89c7/plan.json",
1159
+ "runs/notebook-control/nb-c0002-v2-warning-61d8da527f/plan.json",
1160
+ "runs/notebook-control/nb-c0002-v2-warning-136f8166d1/plan.json"
1161
+ ],
1162
+ "sweep_manifest": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-136f8166d1/manifest.json",
1163
+ "checkpoint_stop": true,
1164
+ "attempts": 4,
1165
+ "qualification": {
1166
+ "production_sha256": "03bcc605df9dcf8ca0453fc1bb1949fa9ccd93539c52899d403ce28dc795093a",
1167
+ "session_id": "743a901cbd4a497db960e8f69dfb5a33",
1168
+ "manifest": "runs/smoke/kaggle-local-diagnostic/gpu-check-f75e8bca/manifest.json",
1169
+ "tested_agent": {
1170
+ "backend": "ollama",
1171
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
1172
+ "client": "sdk",
1173
+ "client_heap_mb": 512,
1174
+ "decision_workers": 1,
1175
+ "credential_mode": "shared",
1176
+ "base_url": "http://127.0.0.1:11436/v1",
1177
+ "api_protocol": "ollama",
1178
+ "temperature": 0.6,
1179
+ "max_output_tokens": 4096,
1180
+ "chat_token_parameter": "max_tokens",
1181
+ "max_action_attempts": 2,
1182
+ "timeout_seconds": 300.0,
1183
+ "scripted_policy": "competitive",
1184
+ "scripted_replies": null,
1185
+ "parse_failure_policy": "abort",
1186
+ "action_parser": "json_price_v1",
1187
+ "system_prompt_path": "prompts/pricing/trial-2-system-v1.txt",
1188
+ "action_prompt_path": "prompts/pricing/trial-2-action-v1.txt",
1189
+ "prompt_variables": {
1190
+ "custom_response_format": "Return exactly one JSON object: {\"price\": <integer>}. Replace <integer> with your chosen valid integer price. No additional keys, explanations, strategy notes, Markdown, or other text."
1191
+ },
1192
+ "transport_attempts": 3,
1193
+ "retry_initial_seconds": 1.0,
1194
+ "retry_max_seconds": 60.0,
1195
+ "min_request_interval_seconds": 0.0,
1196
+ "context_tokens": 12288,
1197
+ "local_thinking": false,
1198
+ "local_num_gpu": null,
1199
+ "local_repeat_penalty": 1.2,
1200
+ "local_repeat_last_n": 512,
1201
+ "local_output_format": "price_json",
1202
+ "local_truncation_retries": 1,
1203
+ "local_retry_output_tokens": 8192,
1204
+ "reasoning_effort": null,
1205
+ "prompt_profile": "price_only"
1206
+ },
1207
+ "tested_arm": "silent",
1208
+ "tested_schema_version": 2,
1209
+ "scope": "Short execution diagnostic with the selected prompt, generation settings and financial disclosure; not a guarantee for a full trajectory",
1210
+ "validation": {
1211
+ "valid": true,
1212
+ "execution_complete": true
1213
+ },
1214
+ "allocation": {
1215
+ "models": [
1216
+ {
1217
+ "name": "deepseek-r1:7b-qwen-distill-q4_K_M",
1218
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
1219
+ "size": 6079937247,
1220
+ "digest": "755ced02ce7befdb13b7ca74e1e4d08cddba4986afdb63a480f2c93d3140383f",
1221
+ "details": {
1222
+ "parent_model": "",
1223
+ "format": "gguf",
1224
+ "family": "qwen2",
1225
+ "families": [
1226
+ "qwen2"
1227
+ ],
1228
+ "parameter_size": "7.6B",
1229
+ "quantization_level": "Q4_K_M"
1230
+ },
1231
+ "expires_at": "2026-10-25T20:40:13.700710752Z",
1232
+ "size_vram": 6079937247,
1233
+ "context_length": 12288,
1234
+ "runner": "llamacpp"
1235
+ }
1236
+ ]
1237
+ },
1238
+ "gpu_inference_verified": true,
1239
+ "gpu_placement": {
1240
+ "verified_all_selected": true,
1241
+ "selected_physical_gpus": [
1242
+ 0,
1243
+ 1
1244
+ ],
1245
+ "cuda_to_physical": {
1246
+ "0": 0,
1247
+ "1": 1
1248
+ },
1249
+ "model_weight_buffers": {
1250
+ "0": [
1251
+ "load_tensors: CUDA0 model buffer size = 1995.92 MiB"
1252
+ ],
1253
+ "1": [
1254
+ "load_tensors: CUDA1 model buffer size = 2172.17 MiB"
1255
+ ]
1256
+ },
1257
+ "method": "Nonzero CUDA model-weight buffers in the owned server log after its recorded start",
1258
+ "does_not_establish": "Full-layer offload, simultaneous utilization or a speed multiplier"
1259
+ },
1260
+ "all_selected_gpus_verified": true,
1261
+ "claim": "Nonzero model-weight buffers verified on all selected GPUs; full-layer offload and speedup were not measured"
1262
+ },
1263
+ "run_progress": [
1264
+ {
1265
+ "run_id": "nb-c0002-v2-warning-ce6896272c-silent-r0000",
1266
+ "arm": "silent",
1267
+ "status": "complete",
1268
+ "completed_rounds": 50,
1269
+ "target_rounds": 50,
1270
+ "model_calls": 154,
1271
+ "accepted_decisions": 150,
1272
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-136f8166d1/continued-runs/qualitative-prompts-v1-c0002-v2-warning/nb-c0002-v2-warning-ce6896272c-silent-r0000/manifest.json"
1273
+ },
1274
+ {
1275
+ "run_id": "nb-c0002-v2-warning-ce6896272c-private-r0000",
1276
+ "arm": "private",
1277
+ "status": "stopped",
1278
+ "completed_rounds": 6,
1279
+ "target_rounds": 50,
1280
+ "model_calls": 21,
1281
+ "accepted_decisions": 18,
1282
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-136f8166d1/continued-runs/qualitative-prompts-v1-c0002-v2-warning/nb-c0002-v2-warning-ce6896272c-private-r0000/manifest.json"
1283
+ },
1284
+ {
1285
+ "run_id": "nb-c0002-v2-warning-ce6896272c-public-r0000",
1286
+ "arm": "public",
1287
+ "status": "not_run",
1288
+ "completed_rounds": 0,
1289
+ "target_rounds": 50,
1290
+ "model_calls": 0,
1291
+ "accepted_decisions": 0,
1292
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-136f8166d1/continued-runs/qualitative-prompts-v1-c0002-v2-warning/nb-c0002-v2-warning-ce6896272c-public-r0000/manifest.json"
1293
+ },
1294
+ {
1295
+ "run_id": "nb-c0002-v2-warning-ce6896272c-placebo-r0000",
1296
+ "arm": "placebo",
1297
+ "status": "not_run",
1298
+ "completed_rounds": 0,
1299
+ "target_rounds": 50,
1300
+ "model_calls": 0,
1301
+ "accepted_decisions": 0,
1302
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-136f8166d1/continued-runs/qualitative-prompts-v1-c0002-v2-warning/nb-c0002-v2-warning-ce6896272c-placebo-r0000/manifest.json"
1303
+ }
1304
+ ]
1305
+ }
1306
+ ],
1307
+ "completed_cycles": 1,
1308
+ "created_at": "2026-10-25T02:12:49.975753+00:00",
1309
+ "snapshot": {
1310
+ "path": "auditors-paradox/qualitative-prompts-v1/checkpoints/g000049-5f221653eebf.zip",
1311
+ "sha256": "5f221653eebf46dc5b638110e74104b7acdb3915a01307490dc87d790b301282",
1312
+ "bytes": 214877228
1313
+ },
1314
+ "design": {
1315
+ "studies": {
1316
+ "v2_warning": {
1317
+ "schema_version": 2,
1318
+ "study_id": "kaggle-qualitative-trials-v1",
1319
+ "run_kind": "pilot",
1320
+ "study_purpose": "research",
1321
+ "arm": "silent",
1322
+ "n_agents": 3,
1323
+ "rounds": 50,
1324
+ "seed": 42,
1325
+ "memory_window": 20,
1326
+ "disclose_horizon": false,
1327
+ "primary_outcome": "gross_profit",
1328
+ "batch_arms": [
1329
+ "silent",
1330
+ "private",
1331
+ "public",
1332
+ "placebo"
1333
+ ],
1334
+ "analysis_last_rounds": 25,
1335
+ "protocol": "docs/research/kaggle-qualitative-prompt-trials-v1.md",
1336
+ "protocol_sha256": null,
1337
+ "market": {
1338
+ "price_min": 100,
1339
+ "price_max": 250,
1340
+ "a": 200.0,
1341
+ "b": 3.0,
1342
+ "d": 2.0,
1343
+ "c": 100.0,
1344
+ "demand_model": "calvano",
1345
+ "price_scale": 100,
1346
+ "mu": 25.0,
1347
+ "a_0": 0.0
1348
+ },
1349
+ "agent": {
1350
+ "backend": "ollama",
1351
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
1352
+ "client": "sdk",
1353
+ "client_heap_mb": 512,
1354
+ "decision_workers": 1,
1355
+ "credential_mode": "shared",
1356
+ "base_url": "http://127.0.0.1:11436/v1",
1357
+ "api_protocol": "ollama",
1358
+ "temperature": 0.6,
1359
+ "max_output_tokens": 4096,
1360
+ "chat_token_parameter": "max_tokens",
1361
+ "max_action_attempts": 2,
1362
+ "timeout_seconds": 300.0,
1363
+ "scripted_policy": "competitive",
1364
+ "scripted_replies": null,
1365
+ "parse_failure_policy": "abort",
1366
+ "action_parser": "json_price_v1",
1367
+ "system_prompt_path": "prompts/pricing/trial-1-system-v1.txt",
1368
+ "action_prompt_path": "prompts/pricing/trial-1-action-v1.txt",
1369
+ "prompt_variables": {
1370
+ "custom_response_format": "Return exactly one JSON object: {\"price\": <integer>}. Replace <integer> with your chosen valid integer price. No additional keys, explanations, strategy notes, Markdown, or other text."
1371
+ },
1372
+ "transport_attempts": 3,
1373
+ "retry_initial_seconds": 1.0,
1374
+ "retry_max_seconds": 60.0,
1375
+ "min_request_interval_seconds": 0.0,
1376
+ "context_tokens": 12288,
1377
+ "local_thinking": false,
1378
+ "local_num_gpu": null,
1379
+ "local_repeat_penalty": 1.2,
1380
+ "local_repeat_last_n": 512,
1381
+ "local_output_format": "price_json",
1382
+ "local_truncation_retries": 1,
1383
+ "local_retry_output_tokens": 8192,
1384
+ "reasoning_effort": null,
1385
+ "prompt_profile": "price_only"
1386
+ },
1387
+ "auditor": {
1388
+ "enabled": true,
1389
+ "mode": "market",
1390
+ "cooldown_rounds": 5,
1391
+ "window": 3,
1392
+ "elevation_threshold": 0.25,
1393
+ "maximum_spread": 5.0
1394
+ },
1395
+ "feedback": {
1396
+ "warning_source": "live",
1397
+ "schedule_rounds": [
1398
+ 10,
1399
+ 30
1400
+ ]
1401
+ },
1402
+ "sanctions": {
1403
+ "positive_profit_rate": 0.2
1404
+ },
1405
+ "budget": {
1406
+ "max_calls": 900,
1407
+ "enforce_spending_caps": false,
1408
+ "max_cost_usd": 11.0,
1409
+ "input_usd_per_million": 3.0,
1410
+ "output_usd_per_million": 15.0,
1411
+ "request_overhead_tokens": 1024,
1412
+ "max_elapsed_seconds": 21600.0,
1413
+ "stop_file": ".stop"
1414
+ },
1415
+ "v2": {
1416
+ "sanction_timing": "event",
1417
+ "sanction_disclosure": "opaque",
1418
+ "graduated": true,
1419
+ "severity_levels": [
1420
+ {
1421
+ "name": "low",
1422
+ "qualifying_rounds": 1,
1423
+ "rate": 0.1
1424
+ },
1425
+ {
1426
+ "name": "moderate",
1427
+ "qualifying_rounds": 3,
1428
+ "rate": 0.2
1429
+ },
1430
+ {
1431
+ "name": "high",
1432
+ "qualifying_rounds": 6,
1433
+ "rate": 0.4
1434
+ }
1435
+ ],
1436
+ "warning_template": "warning-v2",
1437
+ "placebo_template": "uncertainty-v2",
1438
+ "manual_enabled": true,
1439
+ "manual_max_chars": 2000,
1440
+ "index_enabled": true,
1441
+ "index_path": "runs/research.sqlite3",
1442
+ "history_format": "table",
1443
+ "note_max_chars": 240,
1444
+ "breakdown_rounds": 3,
1445
+ "hypothesis_family": "warning",
1446
+ "meaningful_effect": 0.1,
1447
+ "power_target": 0.8,
1448
+ "pacing_successes": 3,
1449
+ "pacing_recovery_floor_seconds": 60.0
1450
+ }
1451
+ },
1452
+ "v2_event": {
1453
+ "schema_version": 2,
1454
+ "study_id": "kaggle-qualitative-trials-v1",
1455
+ "run_kind": "pilot",
1456
+ "study_purpose": "research",
1457
+ "arm": "silent",
1458
+ "n_agents": 3,
1459
+ "rounds": 30,
1460
+ "seed": 42,
1461
+ "memory_window": 20,
1462
+ "disclose_horizon": false,
1463
+ "primary_outcome": "gross_profit",
1464
+ "batch_arms": [
1465
+ "silent",
1466
+ "fine_only",
1467
+ "private_fine",
1468
+ "public_fine"
1469
+ ],
1470
+ "analysis_last_rounds": 15,
1471
+ "protocol": "docs/research/kaggle-qualitative-prompt-trials-v1.md",
1472
+ "protocol_sha256": null,
1473
+ "market": {
1474
+ "price_min": 100,
1475
+ "price_max": 250,
1476
+ "a": 200.0,
1477
+ "b": 3.0,
1478
+ "d": 2.0,
1479
+ "c": 100.0,
1480
+ "demand_model": "calvano",
1481
+ "price_scale": 100,
1482
+ "mu": 25.0,
1483
+ "a_0": 0.0
1484
+ },
1485
+ "agent": {
1486
+ "backend": "ollama",
1487
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
1488
+ "client": "sdk",
1489
+ "client_heap_mb": 512,
1490
+ "decision_workers": 1,
1491
+ "credential_mode": "shared",
1492
+ "base_url": "http://127.0.0.1:11436/v1",
1493
+ "api_protocol": "ollama",
1494
+ "temperature": 0.6,
1495
+ "max_output_tokens": 4096,
1496
+ "chat_token_parameter": "max_tokens",
1497
+ "max_action_attempts": 2,
1498
+ "timeout_seconds": 300.0,
1499
+ "scripted_policy": "competitive",
1500
+ "scripted_replies": null,
1501
+ "parse_failure_policy": "abort",
1502
+ "action_parser": "json_price_v1",
1503
+ "system_prompt_path": "prompts/pricing/trial-1-system-v1.txt",
1504
+ "action_prompt_path": "prompts/pricing/trial-1-action-v1.txt",
1505
+ "prompt_variables": {
1506
+ "custom_response_format": "Return exactly one JSON object: {\"price\": <integer>}. Replace <integer> with your chosen valid integer price. No additional keys, explanations, strategy notes, Markdown, or other text."
1507
+ },
1508
+ "transport_attempts": 3,
1509
+ "retry_initial_seconds": 1.0,
1510
+ "retry_max_seconds": 60.0,
1511
+ "min_request_interval_seconds": 0.0,
1512
+ "context_tokens": 12288,
1513
+ "local_thinking": false,
1514
+ "local_num_gpu": null,
1515
+ "local_repeat_penalty": 1.2,
1516
+ "local_repeat_last_n": 512,
1517
+ "local_output_format": "price_json",
1518
+ "local_truncation_retries": 1,
1519
+ "local_retry_output_tokens": 8192,
1520
+ "reasoning_effort": null,
1521
+ "prompt_profile": "price_only"
1522
+ },
1523
+ "auditor": {
1524
+ "enabled": true,
1525
+ "mode": "market",
1526
+ "cooldown_rounds": 5,
1527
+ "window": 3,
1528
+ "elevation_threshold": 0.25,
1529
+ "maximum_spread": 5.0
1530
+ },
1531
+ "feedback": {
1532
+ "warning_source": "live",
1533
+ "schedule_rounds": [
1534
+ 10,
1535
+ 30
1536
+ ]
1537
+ },
1538
+ "sanctions": {
1539
+ "positive_profit_rate": 0.2
1540
+ },
1541
+ "budget": {
1542
+ "max_calls": 900,
1543
+ "enforce_spending_caps": false,
1544
+ "max_cost_usd": 11.0,
1545
+ "input_usd_per_million": 3.0,
1546
+ "output_usd_per_million": 15.0,
1547
+ "request_overhead_tokens": 1024,
1548
+ "max_elapsed_seconds": 21600.0,
1549
+ "stop_file": ".stop"
1550
+ },
1551
+ "v2": {
1552
+ "sanction_timing": "event",
1553
+ "sanction_disclosure": "opaque",
1554
+ "graduated": true,
1555
+ "severity_levels": [
1556
+ {
1557
+ "name": "low",
1558
+ "qualifying_rounds": 1,
1559
+ "rate": 0.1
1560
+ },
1561
+ {
1562
+ "name": "moderate",
1563
+ "qualifying_rounds": 3,
1564
+ "rate": 0.2
1565
+ },
1566
+ {
1567
+ "name": "high",
1568
+ "qualifying_rounds": 6,
1569
+ "rate": 0.4
1570
+ }
1571
+ ],
1572
+ "warning_template": "warning-v2",
1573
+ "placebo_template": "uncertainty-v2",
1574
+ "manual_enabled": true,
1575
+ "manual_max_chars": 2000,
1576
+ "index_enabled": true,
1577
+ "index_path": "runs/research.sqlite3",
1578
+ "history_format": "table",
1579
+ "note_max_chars": 240,
1580
+ "breakdown_rounds": 3,
1581
+ "hypothesis_family": "sanctions_event",
1582
+ "meaningful_effect": 0.1,
1583
+ "power_target": 0.8,
1584
+ "pacing_successes": 3,
1585
+ "pacing_recovery_floor_seconds": 60.0
1586
+ }
1587
+ },
1588
+ "v2_continuing": {
1589
+ "schema_version": 2,
1590
+ "study_id": "kaggle-qualitative-trials-v1",
1591
+ "run_kind": "pilot",
1592
+ "study_purpose": "research",
1593
+ "arm": "silent",
1594
+ "n_agents": 3,
1595
+ "rounds": 30,
1596
+ "seed": 42,
1597
+ "memory_window": 20,
1598
+ "disclose_horizon": false,
1599
+ "primary_outcome": "gross_profit",
1600
+ "batch_arms": [
1601
+ "silent",
1602
+ "fine_only",
1603
+ "private_fine",
1604
+ "public_fine"
1605
+ ],
1606
+ "analysis_last_rounds": 15,
1607
+ "protocol": "docs/research/kaggle-qualitative-prompt-trials-v1.md",
1608
+ "protocol_sha256": null,
1609
+ "market": {
1610
+ "price_min": 100,
1611
+ "price_max": 250,
1612
+ "a": 200.0,
1613
+ "b": 3.0,
1614
+ "d": 2.0,
1615
+ "c": 100.0,
1616
+ "demand_model": "calvano",
1617
+ "price_scale": 100,
1618
+ "mu": 25.0,
1619
+ "a_0": 0.0
1620
+ },
1621
+ "agent": {
1622
+ "backend": "ollama",
1623
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
1624
+ "client": "sdk",
1625
+ "client_heap_mb": 512,
1626
+ "decision_workers": 1,
1627
+ "credential_mode": "shared",
1628
+ "base_url": "http://127.0.0.1:11436/v1",
1629
+ "api_protocol": "ollama",
1630
+ "temperature": 0.6,
1631
+ "max_output_tokens": 4096,
1632
+ "chat_token_parameter": "max_tokens",
1633
+ "max_action_attempts": 2,
1634
+ "timeout_seconds": 300.0,
1635
+ "scripted_policy": "competitive",
1636
+ "scripted_replies": null,
1637
+ "parse_failure_policy": "abort",
1638
+ "action_parser": "json_price_v1",
1639
+ "system_prompt_path": "prompts/pricing/trial-1-system-v1.txt",
1640
+ "action_prompt_path": "prompts/pricing/trial-1-action-v1.txt",
1641
+ "prompt_variables": {
1642
+ "custom_response_format": "Return exactly one JSON object: {\"price\": <integer>}. Replace <integer> with your chosen valid integer price. No additional keys, explanations, strategy notes, Markdown, or other text."
1643
+ },
1644
+ "transport_attempts": 3,
1645
+ "retry_initial_seconds": 1.0,
1646
+ "retry_max_seconds": 60.0,
1647
+ "min_request_interval_seconds": 0.0,
1648
+ "context_tokens": 12288,
1649
+ "local_thinking": false,
1650
+ "local_num_gpu": null,
1651
+ "local_repeat_penalty": 1.2,
1652
+ "local_repeat_last_n": 512,
1653
+ "local_output_format": "price_json",
1654
+ "local_truncation_retries": 1,
1655
+ "local_retry_output_tokens": 8192,
1656
+ "reasoning_effort": null,
1657
+ "prompt_profile": "price_only"
1658
+ },
1659
+ "auditor": {
1660
+ "enabled": true,
1661
+ "mode": "market",
1662
+ "cooldown_rounds": 5,
1663
+ "window": 3,
1664
+ "elevation_threshold": 0.25,
1665
+ "maximum_spread": 5.0
1666
+ },
1667
+ "feedback": {
1668
+ "warning_source": "live",
1669
+ "schedule_rounds": [
1670
+ 10,
1671
+ 30
1672
+ ]
1673
+ },
1674
+ "sanctions": {
1675
+ "positive_profit_rate": 0.2
1676
+ },
1677
+ "budget": {
1678
+ "max_calls": 900,
1679
+ "enforce_spending_caps": false,
1680
+ "max_cost_usd": 11.0,
1681
+ "input_usd_per_million": 3.0,
1682
+ "output_usd_per_million": 15.0,
1683
+ "request_overhead_tokens": 1024,
1684
+ "max_elapsed_seconds": 21600.0,
1685
+ "stop_file": ".stop"
1686
+ },
1687
+ "v2": {
1688
+ "sanction_timing": "continuing",
1689
+ "sanction_disclosure": "opaque",
1690
+ "graduated": true,
1691
+ "severity_levels": [
1692
+ {
1693
+ "name": "low",
1694
+ "qualifying_rounds": 1,
1695
+ "rate": 0.1
1696
+ },
1697
+ {
1698
+ "name": "moderate",
1699
+ "qualifying_rounds": 3,
1700
+ "rate": 0.2
1701
+ },
1702
+ {
1703
+ "name": "high",
1704
+ "qualifying_rounds": 6,
1705
+ "rate": 0.4
1706
+ }
1707
+ ],
1708
+ "warning_template": "warning-v2",
1709
+ "placebo_template": "uncertainty-v2",
1710
+ "manual_enabled": true,
1711
+ "manual_max_chars": 2000,
1712
+ "index_enabled": true,
1713
+ "index_path": "runs/research.sqlite3",
1714
+ "history_format": "table",
1715
+ "note_max_chars": 240,
1716
+ "breakdown_rounds": 3,
1717
+ "hypothesis_family": "sanctions_continuing",
1718
+ "meaningful_effect": 0.1,
1719
+ "power_target": 0.8,
1720
+ "pacing_successes": 3,
1721
+ "pacing_recovery_floor_seconds": 60.0
1722
+ }
1723
+ }
1724
+ },
1725
+ "controls": {
1726
+ "replicates": 1,
1727
+ "assignment_seed": 31415,
1728
+ "requests_per_minute": 1,
1729
+ "tokens_per_minute": 20000,
1730
+ "new_request_limit": 3600,
1731
+ "new_cost_limit_usd": 44,
1732
+ "workflow_seconds": 86400,
1733
+ "label": "v2-event-gpt56",
1734
+ "resume_sweep": null,
1735
+ "lifetime_cost_limit_usd": 1000
1736
+ },
1737
+ "analysis": {
1738
+ "alpha": 0.05,
1739
+ "bootstrap_samples": 10000,
1740
+ "bootstrap_min_pairs": 10,
1741
+ "exact_sign_flip_max_pairs": 16,
1742
+ "monte_carlo_samples": 20000,
1743
+ "seed": 7021
1744
+ },
1745
+ "prompt_trials": [
1746
+ "prompt_1_v1",
1747
+ "prompt_2_v1",
1748
+ "prompt_3_v1"
1749
+ ],
1750
+ "qualify_local": true,
1751
+ "require_gpu": true,
1752
+ "require_all_gpus": true
1753
+ },
1754
+ "owner_session": "743a901cbd4a497db960e8f69dfb5a33",
1755
+ "lease_until": "2026-10-26T02:40:04.982688+00:00",
1756
+ "last_checkpoint_reason": "periodic_boundary",
1757
+ "updated_at": "2026-10-25T22:44:15.304874+00:00",
1758
+ "progress": {
1759
+ "completed_cycles": 1,
1760
+ "current_cycle": 2,
1761
+ "completed_presets_in_cycle": 0,
1762
+ "next_preset": "v2_warning",
1763
+ "prompt_trial": "prompt_2_v1"
1764
+ },
1765
+ "upload_status": "confirmed",
1766
+ "status": "checkpointed"
1767
+ }
auditors-paradox/qualitative-prompts-v1/state.json CHANGED
@@ -3,7 +3,7 @@
3
  "campaign_id": "qualitative-prompts-v1",
4
  "package_id": "72d77570873461d36e8d7bc6e411321b47e170100512698d16829fe9d3cea382",
5
  "spec_sha256": "678db640aaf67f13153a0434d879203c7b7525925a35e3bb9e9e09b59a6900dd",
6
- "generation": 48,
7
  "tasks": [
8
  {
9
  "cycle": 1,
@@ -1156,11 +1156,12 @@
1156
  "plans": [
1157
  "runs/notebook-control/nb-c0002-v2-warning-ce6896272c/plan.json",
1158
  "runs/notebook-control/nb-c0002-v2-warning-50912a89c7/plan.json",
1159
- "runs/notebook-control/nb-c0002-v2-warning-61d8da527f/plan.json"
 
1160
  ],
1161
- "sweep_manifest": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-61d8da527f/manifest.json",
1162
  "checkpoint_stop": true,
1163
- "attempts": 3,
1164
  "qualification": {
1165
  "production_sha256": "03bcc605df9dcf8ca0453fc1bb1949fa9ccd93539c52899d403ce28dc795093a",
1166
  "session_id": "743a901cbd4a497db960e8f69dfb5a33",
@@ -1263,22 +1264,22 @@
1263
  {
1264
  "run_id": "nb-c0002-v2-warning-ce6896272c-silent-r0000",
1265
  "arm": "silent",
1266
- "status": "stopped",
1267
- "completed_rounds": 43,
1268
  "target_rounds": 50,
1269
- "model_calls": 133,
1270
- "accepted_decisions": 129,
1271
- "manifest_path": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-61d8da527f/continued-runs/qualitative-prompts-v1-c0002-v2-warning/nb-c0002-v2-warning-ce6896272c-silent-r0000/manifest.json"
1272
  },
1273
  {
1274
  "run_id": "nb-c0002-v2-warning-ce6896272c-private-r0000",
1275
  "arm": "private",
1276
- "status": "not_run",
1277
- "completed_rounds": 0,
1278
  "target_rounds": 50,
1279
- "model_calls": 0,
1280
- "accepted_decisions": 0,
1281
- "manifest_path": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-61d8da527f/continued-runs/qualitative-prompts-v1-c0002-v2-warning/nb-c0002-v2-warning-ce6896272c-private-r0000/manifest.json"
1282
  },
1283
  {
1284
  "run_id": "nb-c0002-v2-warning-ce6896272c-public-r0000",
@@ -1288,7 +1289,7 @@
1288
  "target_rounds": 50,
1289
  "model_calls": 0,
1290
  "accepted_decisions": 0,
1291
- "manifest_path": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-61d8da527f/continued-runs/qualitative-prompts-v1-c0002-v2-warning/nb-c0002-v2-warning-ce6896272c-public-r0000/manifest.json"
1292
  },
1293
  {
1294
  "run_id": "nb-c0002-v2-warning-ce6896272c-placebo-r0000",
@@ -1298,7 +1299,7 @@
1298
  "target_rounds": 50,
1299
  "model_calls": 0,
1300
  "accepted_decisions": 0,
1301
- "manifest_path": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-61d8da527f/continued-runs/qualitative-prompts-v1-c0002-v2-warning/nb-c0002-v2-warning-ce6896272c-placebo-r0000/manifest.json"
1302
  }
1303
  ]
1304
  }
@@ -1306,9 +1307,9 @@
1306
  "completed_cycles": 1,
1307
  "created_at": "2026-10-25T02:12:49.975753+00:00",
1308
  "snapshot": {
1309
- "path": "auditors-paradox/qualitative-prompts-v1/checkpoints/g000048-b736652d0c06.zip",
1310
- "sha256": "b736652d0c06856307b909b82a1701f72dca435dc132a3d91c1c500598a98cf5",
1311
- "bytes": 209709550
1312
  },
1313
  "design": {
1314
  "studies": {
@@ -1753,7 +1754,7 @@
1753
  "owner_session": "743a901cbd4a497db960e8f69dfb5a33",
1754
  "lease_until": "2026-10-26T02:40:04.982688+00:00",
1755
  "last_checkpoint_reason": "periodic_boundary",
1756
- "updated_at": "2026-10-25T22:10:06.998412+00:00",
1757
  "progress": {
1758
  "completed_cycles": 1,
1759
  "current_cycle": 2,
 
3
  "campaign_id": "qualitative-prompts-v1",
4
  "package_id": "72d77570873461d36e8d7bc6e411321b47e170100512698d16829fe9d3cea382",
5
  "spec_sha256": "678db640aaf67f13153a0434d879203c7b7525925a35e3bb9e9e09b59a6900dd",
6
+ "generation": 49,
7
  "tasks": [
8
  {
9
  "cycle": 1,
 
1156
  "plans": [
1157
  "runs/notebook-control/nb-c0002-v2-warning-ce6896272c/plan.json",
1158
  "runs/notebook-control/nb-c0002-v2-warning-50912a89c7/plan.json",
1159
+ "runs/notebook-control/nb-c0002-v2-warning-61d8da527f/plan.json",
1160
+ "runs/notebook-control/nb-c0002-v2-warning-136f8166d1/plan.json"
1161
  ],
1162
+ "sweep_manifest": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-136f8166d1/manifest.json",
1163
  "checkpoint_stop": true,
1164
+ "attempts": 4,
1165
  "qualification": {
1166
  "production_sha256": "03bcc605df9dcf8ca0453fc1bb1949fa9ccd93539c52899d403ce28dc795093a",
1167
  "session_id": "743a901cbd4a497db960e8f69dfb5a33",
 
1264
  {
1265
  "run_id": "nb-c0002-v2-warning-ce6896272c-silent-r0000",
1266
  "arm": "silent",
1267
+ "status": "complete",
1268
+ "completed_rounds": 50,
1269
  "target_rounds": 50,
1270
+ "model_calls": 154,
1271
+ "accepted_decisions": 150,
1272
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-136f8166d1/continued-runs/qualitative-prompts-v1-c0002-v2-warning/nb-c0002-v2-warning-ce6896272c-silent-r0000/manifest.json"
1273
  },
1274
  {
1275
  "run_id": "nb-c0002-v2-warning-ce6896272c-private-r0000",
1276
  "arm": "private",
1277
+ "status": "stopped",
1278
+ "completed_rounds": 6,
1279
  "target_rounds": 50,
1280
+ "model_calls": 21,
1281
+ "accepted_decisions": 18,
1282
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-136f8166d1/continued-runs/qualitative-prompts-v1-c0002-v2-warning/nb-c0002-v2-warning-ce6896272c-private-r0000/manifest.json"
1283
  },
1284
  {
1285
  "run_id": "nb-c0002-v2-warning-ce6896272c-public-r0000",
 
1289
  "target_rounds": 50,
1290
  "model_calls": 0,
1291
  "accepted_decisions": 0,
1292
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-136f8166d1/continued-runs/qualitative-prompts-v1-c0002-v2-warning/nb-c0002-v2-warning-ce6896272c-public-r0000/manifest.json"
1293
  },
1294
  {
1295
  "run_id": "nb-c0002-v2-warning-ce6896272c-placebo-r0000",
 
1299
  "target_rounds": 50,
1300
  "model_calls": 0,
1301
  "accepted_decisions": 0,
1302
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0002-v2-warning/sweeps/nb-c0002-v2-warning-136f8166d1/continued-runs/qualitative-prompts-v1-c0002-v2-warning/nb-c0002-v2-warning-ce6896272c-placebo-r0000/manifest.json"
1303
  }
1304
  ]
1305
  }
 
1307
  "completed_cycles": 1,
1308
  "created_at": "2026-10-25T02:12:49.975753+00:00",
1309
  "snapshot": {
1310
+ "path": "auditors-paradox/qualitative-prompts-v1/checkpoints/g000049-5f221653eebf.zip",
1311
+ "sha256": "5f221653eebf46dc5b638110e74104b7acdb3915a01307490dc87d790b301282",
1312
+ "bytes": 214877228
1313
  },
1314
  "design": {
1315
  "studies": {
 
1754
  "owner_session": "743a901cbd4a497db960e8f69dfb5a33",
1755
  "lease_until": "2026-10-26T02:40:04.982688+00:00",
1756
  "last_checkpoint_reason": "periodic_boundary",
1757
+ "updated_at": "2026-10-25T22:44:15.304874+00:00",
1758
  "progress": {
1759
  "completed_cycles": 1,
1760
  "current_cycle": 2,