AIsekai commited on
Commit
b3741cb
·
verified ·
1 Parent(s): 9df4226

Checkpoint qualitative-prompts-v1 generation 26

Browse files
auditors-paradox/qualitative-prompts-v1/checkpoints/g000026-644807085e1d.zip ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:644807085e1d6486fe22828fb43023e77f5fa4589a8f6d47016332bfb161ef4d
3
+ size 125719095
auditors-paradox/qualitative-prompts-v1/state-history/g000026.json ADDED
@@ -0,0 +1,782 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": 1,
3
+ "campaign_id": "qualitative-prompts-v1",
4
+ "package_id": "72d77570873461d36e8d7bc6e411321b47e170100512698d16829fe9d3cea382",
5
+ "spec_sha256": "678db640aaf67f13153a0434d879203c7b7525925a35e3bb9e9e09b59a6900dd",
6
+ "generation": 26,
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": "checkpointed",
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
+ ],
177
+ "sweep_manifest": "runs/pilot/qualitative-prompts-v1-c0001-v2-warning/sweeps/nb-c0001-v2-warning-fad82a2955/manifest.json",
178
+ "checkpoint_stop": true,
179
+ "attempts": 25,
180
+ "qualification": {
181
+ "production_sha256": "a0a86184e435b3c48a3af1c673a6c1b348cb78c35d65e926b2e2153901877141",
182
+ "session_id": "85b8576227694925a65a811313dce74f",
183
+ "manifest": "runs/smoke/kaggle-local-diagnostic/gpu-check-618639f5/manifest.json",
184
+ "tested_agent": {
185
+ "backend": "ollama",
186
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
187
+ "client": "sdk",
188
+ "client_heap_mb": 512,
189
+ "decision_workers": 1,
190
+ "credential_mode": "shared",
191
+ "base_url": "http://127.0.0.1:11436/v1",
192
+ "api_protocol": "ollama",
193
+ "temperature": 0.6,
194
+ "max_output_tokens": 4096,
195
+ "chat_token_parameter": "max_tokens",
196
+ "max_action_attempts": 2,
197
+ "timeout_seconds": 300.0,
198
+ "scripted_policy": "competitive",
199
+ "scripted_replies": null,
200
+ "parse_failure_policy": "abort",
201
+ "action_parser": "json_price_v1",
202
+ "system_prompt_path": "prompts/pricing/trial-1-system-v1.txt",
203
+ "action_prompt_path": "prompts/pricing/trial-1-action-v1.txt",
204
+ "prompt_variables": {
205
+ "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."
206
+ },
207
+ "transport_attempts": 3,
208
+ "retry_initial_seconds": 1.0,
209
+ "retry_max_seconds": 60.0,
210
+ "min_request_interval_seconds": 0.0,
211
+ "context_tokens": 12288,
212
+ "local_thinking": false,
213
+ "local_num_gpu": null,
214
+ "local_repeat_penalty": 1.2,
215
+ "local_repeat_last_n": 512,
216
+ "local_output_format": "price_json",
217
+ "local_truncation_retries": 1,
218
+ "local_retry_output_tokens": 8192,
219
+ "reasoning_effort": null,
220
+ "prompt_profile": "price_only"
221
+ },
222
+ "tested_arm": "silent",
223
+ "tested_schema_version": 2,
224
+ "scope": "Short execution diagnostic with the selected prompt, generation settings and financial disclosure; not a guarantee for a full trajectory",
225
+ "validation": {
226
+ "valid": true,
227
+ "execution_complete": true
228
+ },
229
+ "allocation": {
230
+ "models": [
231
+ {
232
+ "name": "deepseek-r1:7b-qwen-distill-q4_K_M",
233
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
234
+ "size": 6079937247,
235
+ "digest": "755ced02ce7befdb13b7ca74e1e4d08cddba4986afdb63a480f2c93d3140383f",
236
+ "details": {
237
+ "parent_model": "",
238
+ "format": "gguf",
239
+ "family": "qwen2",
240
+ "families": [
241
+ "qwen2"
242
+ ],
243
+ "parameter_size": "7.6B",
244
+ "quantization_level": "Q4_K_M"
245
+ },
246
+ "expires_at": "2026-10-25T02:26:51.960162589Z",
247
+ "size_vram": 6079937247,
248
+ "context_length": 12288,
249
+ "runner": "llamacpp"
250
+ }
251
+ ]
252
+ },
253
+ "gpu_inference_verified": true,
254
+ "gpu_placement": {
255
+ "verified_all_selected": true,
256
+ "selected_physical_gpus": [
257
+ 0,
258
+ 1
259
+ ],
260
+ "cuda_to_physical": {
261
+ "0": 0,
262
+ "1": 1
263
+ },
264
+ "model_weight_buffers": {
265
+ "0": [
266
+ "load_tensors: CUDA0 model buffer size = 1995.92 MiB"
267
+ ],
268
+ "1": [
269
+ "load_tensors: CUDA1 model buffer size = 2172.17 MiB"
270
+ ]
271
+ },
272
+ "method": "Nonzero CUDA model-weight buffers in the owned server log after its recorded start",
273
+ "does_not_establish": "Full-layer offload, simultaneous utilization or a speed multiplier"
274
+ },
275
+ "all_selected_gpus_verified": true,
276
+ "claim": "Nonzero model-weight buffers verified on all selected GPUs; full-layer offload and speedup were not measured"
277
+ },
278
+ "run_progress": [
279
+ {
280
+ "run_id": "nb-c0001-v2-warning-d1d5bb3f0f-placebo-r0000",
281
+ "arm": "placebo",
282
+ "status": "complete",
283
+ "completed_rounds": 50,
284
+ "target_rounds": 50,
285
+ "model_calls": 150,
286
+ "accepted_decisions": 150,
287
+ "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"
288
+ },
289
+ {
290
+ "run_id": "nb-c0001-v2-warning-d1d5bb3f0f-silent-r0000",
291
+ "arm": "silent",
292
+ "status": "complete",
293
+ "completed_rounds": 50,
294
+ "target_rounds": 50,
295
+ "model_calls": 153,
296
+ "accepted_decisions": 150,
297
+ "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"
298
+ },
299
+ {
300
+ "run_id": "nb-c0001-v2-warning-d1d5bb3f0f-private-r0000",
301
+ "arm": "private",
302
+ "status": "complete",
303
+ "completed_rounds": 50,
304
+ "target_rounds": 50,
305
+ "model_calls": 150,
306
+ "accepted_decisions": 150,
307
+ "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"
308
+ },
309
+ {
310
+ "run_id": "nb-c0001-v2-warning-d1d5bb3f0f-public-r0000",
311
+ "arm": "public",
312
+ "status": "stopped",
313
+ "completed_rounds": 37,
314
+ "target_rounds": 50,
315
+ "model_calls": 111,
316
+ "accepted_decisions": 111,
317
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-warning/sweeps/nb-c0001-v2-warning-fad82a2955/continued-runs/qualitative-prompts-v1-c0001-v2-warning/nb-c0001-v2-warning-d1d5bb3f0f-public-r0000/manifest.json"
318
+ }
319
+ ]
320
+ }
321
+ ],
322
+ "completed_cycles": 0,
323
+ "created_at": "2026-10-25T02:12:49.975753+00:00",
324
+ "snapshot": {
325
+ "path": "auditors-paradox/qualitative-prompts-v1/checkpoints/g000026-644807085e1d.zip",
326
+ "sha256": "644807085e1d6486fe22828fb43023e77f5fa4589a8f6d47016332bfb161ef4d",
327
+ "bytes": 125719095
328
+ },
329
+ "design": {
330
+ "studies": {
331
+ "v2_warning": {
332
+ "schema_version": 2,
333
+ "study_id": "kaggle-qualitative-trials-v1",
334
+ "run_kind": "pilot",
335
+ "study_purpose": "research",
336
+ "arm": "silent",
337
+ "n_agents": 3,
338
+ "rounds": 50,
339
+ "seed": 42,
340
+ "memory_window": 20,
341
+ "disclose_horizon": false,
342
+ "primary_outcome": "gross_profit",
343
+ "batch_arms": [
344
+ "silent",
345
+ "private",
346
+ "public",
347
+ "placebo"
348
+ ],
349
+ "analysis_last_rounds": 25,
350
+ "protocol": "docs/research/kaggle-qualitative-prompt-trials-v1.md",
351
+ "protocol_sha256": null,
352
+ "market": {
353
+ "price_min": 100,
354
+ "price_max": 250,
355
+ "a": 200.0,
356
+ "b": 3.0,
357
+ "d": 2.0,
358
+ "c": 100.0,
359
+ "demand_model": "calvano",
360
+ "price_scale": 100,
361
+ "mu": 25.0,
362
+ "a_0": 0.0
363
+ },
364
+ "agent": {
365
+ "backend": "ollama",
366
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
367
+ "client": "sdk",
368
+ "client_heap_mb": 512,
369
+ "decision_workers": 1,
370
+ "credential_mode": "shared",
371
+ "base_url": "http://127.0.0.1:11436/v1",
372
+ "api_protocol": "ollama",
373
+ "temperature": 0.6,
374
+ "max_output_tokens": 4096,
375
+ "chat_token_parameter": "max_tokens",
376
+ "max_action_attempts": 2,
377
+ "timeout_seconds": 300.0,
378
+ "scripted_policy": "competitive",
379
+ "scripted_replies": null,
380
+ "parse_failure_policy": "abort",
381
+ "action_parser": "json_price_v1",
382
+ "system_prompt_path": "prompts/pricing/trial-1-system-v1.txt",
383
+ "action_prompt_path": "prompts/pricing/trial-1-action-v1.txt",
384
+ "prompt_variables": {
385
+ "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."
386
+ },
387
+ "transport_attempts": 3,
388
+ "retry_initial_seconds": 1.0,
389
+ "retry_max_seconds": 60.0,
390
+ "min_request_interval_seconds": 0.0,
391
+ "context_tokens": 12288,
392
+ "local_thinking": false,
393
+ "local_num_gpu": null,
394
+ "local_repeat_penalty": 1.2,
395
+ "local_repeat_last_n": 512,
396
+ "local_output_format": "price_json",
397
+ "local_truncation_retries": 1,
398
+ "local_retry_output_tokens": 8192,
399
+ "reasoning_effort": null,
400
+ "prompt_profile": "price_only"
401
+ },
402
+ "auditor": {
403
+ "enabled": true,
404
+ "mode": "market",
405
+ "cooldown_rounds": 5,
406
+ "window": 3,
407
+ "elevation_threshold": 0.25,
408
+ "maximum_spread": 5.0
409
+ },
410
+ "feedback": {
411
+ "warning_source": "live",
412
+ "schedule_rounds": [
413
+ 10,
414
+ 30
415
+ ]
416
+ },
417
+ "sanctions": {
418
+ "positive_profit_rate": 0.2
419
+ },
420
+ "budget": {
421
+ "max_calls": 900,
422
+ "enforce_spending_caps": false,
423
+ "max_cost_usd": 11.0,
424
+ "input_usd_per_million": 3.0,
425
+ "output_usd_per_million": 15.0,
426
+ "request_overhead_tokens": 1024,
427
+ "max_elapsed_seconds": 21600.0,
428
+ "stop_file": ".stop"
429
+ },
430
+ "v2": {
431
+ "sanction_timing": "event",
432
+ "sanction_disclosure": "opaque",
433
+ "graduated": true,
434
+ "severity_levels": [
435
+ {
436
+ "name": "low",
437
+ "qualifying_rounds": 1,
438
+ "rate": 0.1
439
+ },
440
+ {
441
+ "name": "moderate",
442
+ "qualifying_rounds": 3,
443
+ "rate": 0.2
444
+ },
445
+ {
446
+ "name": "high",
447
+ "qualifying_rounds": 6,
448
+ "rate": 0.4
449
+ }
450
+ ],
451
+ "warning_template": "warning-v2",
452
+ "placebo_template": "uncertainty-v2",
453
+ "manual_enabled": true,
454
+ "manual_max_chars": 2000,
455
+ "index_enabled": true,
456
+ "index_path": "runs/research.sqlite3",
457
+ "history_format": "table",
458
+ "note_max_chars": 240,
459
+ "breakdown_rounds": 3,
460
+ "hypothesis_family": "warning",
461
+ "meaningful_effect": 0.1,
462
+ "power_target": 0.8,
463
+ "pacing_successes": 3,
464
+ "pacing_recovery_floor_seconds": 60.0
465
+ }
466
+ },
467
+ "v2_event": {
468
+ "schema_version": 2,
469
+ "study_id": "kaggle-qualitative-trials-v1",
470
+ "run_kind": "pilot",
471
+ "study_purpose": "research",
472
+ "arm": "silent",
473
+ "n_agents": 3,
474
+ "rounds": 30,
475
+ "seed": 42,
476
+ "memory_window": 20,
477
+ "disclose_horizon": false,
478
+ "primary_outcome": "gross_profit",
479
+ "batch_arms": [
480
+ "silent",
481
+ "fine_only",
482
+ "private_fine",
483
+ "public_fine"
484
+ ],
485
+ "analysis_last_rounds": 15,
486
+ "protocol": "docs/research/kaggle-qualitative-prompt-trials-v1.md",
487
+ "protocol_sha256": null,
488
+ "market": {
489
+ "price_min": 100,
490
+ "price_max": 250,
491
+ "a": 200.0,
492
+ "b": 3.0,
493
+ "d": 2.0,
494
+ "c": 100.0,
495
+ "demand_model": "calvano",
496
+ "price_scale": 100,
497
+ "mu": 25.0,
498
+ "a_0": 0.0
499
+ },
500
+ "agent": {
501
+ "backend": "ollama",
502
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
503
+ "client": "sdk",
504
+ "client_heap_mb": 512,
505
+ "decision_workers": 1,
506
+ "credential_mode": "shared",
507
+ "base_url": "http://127.0.0.1:11436/v1",
508
+ "api_protocol": "ollama",
509
+ "temperature": 0.6,
510
+ "max_output_tokens": 4096,
511
+ "chat_token_parameter": "max_tokens",
512
+ "max_action_attempts": 2,
513
+ "timeout_seconds": 300.0,
514
+ "scripted_policy": "competitive",
515
+ "scripted_replies": null,
516
+ "parse_failure_policy": "abort",
517
+ "action_parser": "json_price_v1",
518
+ "system_prompt_path": "prompts/pricing/trial-1-system-v1.txt",
519
+ "action_prompt_path": "prompts/pricing/trial-1-action-v1.txt",
520
+ "prompt_variables": {
521
+ "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."
522
+ },
523
+ "transport_attempts": 3,
524
+ "retry_initial_seconds": 1.0,
525
+ "retry_max_seconds": 60.0,
526
+ "min_request_interval_seconds": 0.0,
527
+ "context_tokens": 12288,
528
+ "local_thinking": false,
529
+ "local_num_gpu": null,
530
+ "local_repeat_penalty": 1.2,
531
+ "local_repeat_last_n": 512,
532
+ "local_output_format": "price_json",
533
+ "local_truncation_retries": 1,
534
+ "local_retry_output_tokens": 8192,
535
+ "reasoning_effort": null,
536
+ "prompt_profile": "price_only"
537
+ },
538
+ "auditor": {
539
+ "enabled": true,
540
+ "mode": "market",
541
+ "cooldown_rounds": 5,
542
+ "window": 3,
543
+ "elevation_threshold": 0.25,
544
+ "maximum_spread": 5.0
545
+ },
546
+ "feedback": {
547
+ "warning_source": "live",
548
+ "schedule_rounds": [
549
+ 10,
550
+ 30
551
+ ]
552
+ },
553
+ "sanctions": {
554
+ "positive_profit_rate": 0.2
555
+ },
556
+ "budget": {
557
+ "max_calls": 900,
558
+ "enforce_spending_caps": false,
559
+ "max_cost_usd": 11.0,
560
+ "input_usd_per_million": 3.0,
561
+ "output_usd_per_million": 15.0,
562
+ "request_overhead_tokens": 1024,
563
+ "max_elapsed_seconds": 21600.0,
564
+ "stop_file": ".stop"
565
+ },
566
+ "v2": {
567
+ "sanction_timing": "event",
568
+ "sanction_disclosure": "opaque",
569
+ "graduated": true,
570
+ "severity_levels": [
571
+ {
572
+ "name": "low",
573
+ "qualifying_rounds": 1,
574
+ "rate": 0.1
575
+ },
576
+ {
577
+ "name": "moderate",
578
+ "qualifying_rounds": 3,
579
+ "rate": 0.2
580
+ },
581
+ {
582
+ "name": "high",
583
+ "qualifying_rounds": 6,
584
+ "rate": 0.4
585
+ }
586
+ ],
587
+ "warning_template": "warning-v2",
588
+ "placebo_template": "uncertainty-v2",
589
+ "manual_enabled": true,
590
+ "manual_max_chars": 2000,
591
+ "index_enabled": true,
592
+ "index_path": "runs/research.sqlite3",
593
+ "history_format": "table",
594
+ "note_max_chars": 240,
595
+ "breakdown_rounds": 3,
596
+ "hypothesis_family": "sanctions_event",
597
+ "meaningful_effect": 0.1,
598
+ "power_target": 0.8,
599
+ "pacing_successes": 3,
600
+ "pacing_recovery_floor_seconds": 60.0
601
+ }
602
+ },
603
+ "v2_continuing": {
604
+ "schema_version": 2,
605
+ "study_id": "kaggle-qualitative-trials-v1",
606
+ "run_kind": "pilot",
607
+ "study_purpose": "research",
608
+ "arm": "silent",
609
+ "n_agents": 3,
610
+ "rounds": 30,
611
+ "seed": 42,
612
+ "memory_window": 20,
613
+ "disclose_horizon": false,
614
+ "primary_outcome": "gross_profit",
615
+ "batch_arms": [
616
+ "silent",
617
+ "fine_only",
618
+ "private_fine",
619
+ "public_fine"
620
+ ],
621
+ "analysis_last_rounds": 15,
622
+ "protocol": "docs/research/kaggle-qualitative-prompt-trials-v1.md",
623
+ "protocol_sha256": null,
624
+ "market": {
625
+ "price_min": 100,
626
+ "price_max": 250,
627
+ "a": 200.0,
628
+ "b": 3.0,
629
+ "d": 2.0,
630
+ "c": 100.0,
631
+ "demand_model": "calvano",
632
+ "price_scale": 100,
633
+ "mu": 25.0,
634
+ "a_0": 0.0
635
+ },
636
+ "agent": {
637
+ "backend": "ollama",
638
+ "model": "deepseek-r1:7b-qwen-distill-q4_K_M",
639
+ "client": "sdk",
640
+ "client_heap_mb": 512,
641
+ "decision_workers": 1,
642
+ "credential_mode": "shared",
643
+ "base_url": "http://127.0.0.1:11436/v1",
644
+ "api_protocol": "ollama",
645
+ "temperature": 0.6,
646
+ "max_output_tokens": 4096,
647
+ "chat_token_parameter": "max_tokens",
648
+ "max_action_attempts": 2,
649
+ "timeout_seconds": 300.0,
650
+ "scripted_policy": "competitive",
651
+ "scripted_replies": null,
652
+ "parse_failure_policy": "abort",
653
+ "action_parser": "json_price_v1",
654
+ "system_prompt_path": "prompts/pricing/trial-1-system-v1.txt",
655
+ "action_prompt_path": "prompts/pricing/trial-1-action-v1.txt",
656
+ "prompt_variables": {
657
+ "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."
658
+ },
659
+ "transport_attempts": 3,
660
+ "retry_initial_seconds": 1.0,
661
+ "retry_max_seconds": 60.0,
662
+ "min_request_interval_seconds": 0.0,
663
+ "context_tokens": 12288,
664
+ "local_thinking": false,
665
+ "local_num_gpu": null,
666
+ "local_repeat_penalty": 1.2,
667
+ "local_repeat_last_n": 512,
668
+ "local_output_format": "price_json",
669
+ "local_truncation_retries": 1,
670
+ "local_retry_output_tokens": 8192,
671
+ "reasoning_effort": null,
672
+ "prompt_profile": "price_only"
673
+ },
674
+ "auditor": {
675
+ "enabled": true,
676
+ "mode": "market",
677
+ "cooldown_rounds": 5,
678
+ "window": 3,
679
+ "elevation_threshold": 0.25,
680
+ "maximum_spread": 5.0
681
+ },
682
+ "feedback": {
683
+ "warning_source": "live",
684
+ "schedule_rounds": [
685
+ 10,
686
+ 30
687
+ ]
688
+ },
689
+ "sanctions": {
690
+ "positive_profit_rate": 0.2
691
+ },
692
+ "budget": {
693
+ "max_calls": 900,
694
+ "enforce_spending_caps": false,
695
+ "max_cost_usd": 11.0,
696
+ "input_usd_per_million": 3.0,
697
+ "output_usd_per_million": 15.0,
698
+ "request_overhead_tokens": 1024,
699
+ "max_elapsed_seconds": 21600.0,
700
+ "stop_file": ".stop"
701
+ },
702
+ "v2": {
703
+ "sanction_timing": "continuing",
704
+ "sanction_disclosure": "opaque",
705
+ "graduated": true,
706
+ "severity_levels": [
707
+ {
708
+ "name": "low",
709
+ "qualifying_rounds": 1,
710
+ "rate": 0.1
711
+ },
712
+ {
713
+ "name": "moderate",
714
+ "qualifying_rounds": 3,
715
+ "rate": 0.2
716
+ },
717
+ {
718
+ "name": "high",
719
+ "qualifying_rounds": 6,
720
+ "rate": 0.4
721
+ }
722
+ ],
723
+ "warning_template": "warning-v2",
724
+ "placebo_template": "uncertainty-v2",
725
+ "manual_enabled": true,
726
+ "manual_max_chars": 2000,
727
+ "index_enabled": true,
728
+ "index_path": "runs/research.sqlite3",
729
+ "history_format": "table",
730
+ "note_max_chars": 240,
731
+ "breakdown_rounds": 3,
732
+ "hypothesis_family": "sanctions_continuing",
733
+ "meaningful_effect": 0.1,
734
+ "power_target": 0.8,
735
+ "pacing_successes": 3,
736
+ "pacing_recovery_floor_seconds": 60.0
737
+ }
738
+ }
739
+ },
740
+ "controls": {
741
+ "replicates": 1,
742
+ "assignment_seed": 31415,
743
+ "requests_per_minute": 1,
744
+ "tokens_per_minute": 20000,
745
+ "new_request_limit": 3600,
746
+ "new_cost_limit_usd": 44,
747
+ "workflow_seconds": 86400,
748
+ "label": "v2-event-gpt56",
749
+ "resume_sweep": null,
750
+ "lifetime_cost_limit_usd": 1000
751
+ },
752
+ "analysis": {
753
+ "alpha": 0.05,
754
+ "bootstrap_samples": 10000,
755
+ "bootstrap_min_pairs": 10,
756
+ "exact_sign_flip_max_pairs": 16,
757
+ "monte_carlo_samples": 20000,
758
+ "seed": 7021
759
+ },
760
+ "prompt_trials": [
761
+ "prompt_1_v1",
762
+ "prompt_2_v1",
763
+ "prompt_3_v1"
764
+ ],
765
+ "qualify_local": true,
766
+ "require_gpu": true,
767
+ "require_all_gpus": true
768
+ },
769
+ "owner_session": "85b8576227694925a65a811313dce74f",
770
+ "lease_until": "2026-10-25T13:39:39.256746+00:00",
771
+ "last_checkpoint_reason": "periodic_boundary",
772
+ "updated_at": "2026-10-25T07:11:17.196533+00:00",
773
+ "progress": {
774
+ "completed_cycles": 0,
775
+ "current_cycle": 1,
776
+ "completed_presets_in_cycle": 0,
777
+ "next_preset": "v2_warning",
778
+ "prompt_trial": "prompt_1_v1"
779
+ },
780
+ "upload_status": "confirmed",
781
+ "status": "checkpointed"
782
+ }
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": 25,
7
  "tasks": [
8
  {
9
  "cycle": 1,
@@ -171,11 +171,12 @@
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
  ],
176
- "sweep_manifest": "runs/pilot/qualitative-prompts-v1-c0001-v2-warning/sweeps/nb-c0001-v2-warning-75eef6c5c6/manifest.json",
177
  "checkpoint_stop": true,
178
- "attempts": 24,
179
  "qualification": {
180
  "production_sha256": "a0a86184e435b3c48a3af1c673a6c1b348cb78c35d65e926b2e2153901877141",
181
  "session_id": "85b8576227694925a65a811313dce74f",
@@ -309,11 +310,11 @@
309
  "run_id": "nb-c0001-v2-warning-d1d5bb3f0f-public-r0000",
310
  "arm": "public",
311
  "status": "stopped",
312
- "completed_rounds": 28,
313
  "target_rounds": 50,
314
- "model_calls": 84,
315
- "accepted_decisions": 84,
316
- "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-warning/sweeps/nb-c0001-v2-warning-75eef6c5c6/continued-runs/qualitative-prompts-v1-c0001-v2-warning/nb-c0001-v2-warning-d1d5bb3f0f-public-r0000/manifest.json"
317
  }
318
  ]
319
  }
@@ -321,9 +322,9 @@
321
  "completed_cycles": 0,
322
  "created_at": "2026-10-25T02:12:49.975753+00:00",
323
  "snapshot": {
324
- "path": "auditors-paradox/qualitative-prompts-v1/checkpoints/g000025-910e96ee15ac.zip",
325
- "sha256": "910e96ee15ac5810e375f3bf8d856d786430c3aa3b4394ab56cfa40bcef3bd76",
326
- "bytes": 117564236
327
  },
328
  "design": {
329
  "studies": {
@@ -768,7 +769,7 @@
768
  "owner_session": "85b8576227694925a65a811313dce74f",
769
  "lease_until": "2026-10-25T13:39:39.256746+00:00",
770
  "last_checkpoint_reason": "periodic_boundary",
771
- "updated_at": "2026-10-25T06:59:54.744280+00:00",
772
  "progress": {
773
  "completed_cycles": 0,
774
  "current_cycle": 1,
 
3
  "campaign_id": "qualitative-prompts-v1",
4
  "package_id": "72d77570873461d36e8d7bc6e411321b47e170100512698d16829fe9d3cea382",
5
  "spec_sha256": "678db640aaf67f13153a0434d879203c7b7525925a35e3bb9e9e09b59a6900dd",
6
+ "generation": 26,
7
  "tasks": [
8
  {
9
  "cycle": 1,
 
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
  ],
177
+ "sweep_manifest": "runs/pilot/qualitative-prompts-v1-c0001-v2-warning/sweeps/nb-c0001-v2-warning-fad82a2955/manifest.json",
178
  "checkpoint_stop": true,
179
+ "attempts": 25,
180
  "qualification": {
181
  "production_sha256": "a0a86184e435b3c48a3af1c673a6c1b348cb78c35d65e926b2e2153901877141",
182
  "session_id": "85b8576227694925a65a811313dce74f",
 
310
  "run_id": "nb-c0001-v2-warning-d1d5bb3f0f-public-r0000",
311
  "arm": "public",
312
  "status": "stopped",
313
+ "completed_rounds": 37,
314
  "target_rounds": 50,
315
+ "model_calls": 111,
316
+ "accepted_decisions": 111,
317
+ "manifest_path": "runs/pilot/qualitative-prompts-v1-c0001-v2-warning/sweeps/nb-c0001-v2-warning-fad82a2955/continued-runs/qualitative-prompts-v1-c0001-v2-warning/nb-c0001-v2-warning-d1d5bb3f0f-public-r0000/manifest.json"
318
  }
319
  ]
320
  }
 
322
  "completed_cycles": 0,
323
  "created_at": "2026-10-25T02:12:49.975753+00:00",
324
  "snapshot": {
325
+ "path": "auditors-paradox/qualitative-prompts-v1/checkpoints/g000026-644807085e1d.zip",
326
+ "sha256": "644807085e1d6486fe22828fb43023e77f5fa4589a8f6d47016332bfb161ef4d",
327
+ "bytes": 125719095
328
  },
329
  "design": {
330
  "studies": {
 
769
  "owner_session": "85b8576227694925a65a811313dce74f",
770
  "lease_until": "2026-10-25T13:39:39.256746+00:00",
771
  "last_checkpoint_reason": "periodic_boundary",
772
+ "updated_at": "2026-10-25T07:11:17.196533+00:00",
773
  "progress": {
774
  "completed_cycles": 0,
775
  "current_cycle": 1,