File size: 40,426 Bytes
4c69214
 
 
 
 
 
 
 
 
 
 
 
ece0bbe
4c69214
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ece0bbe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4c69214
 
 
 
 
 
 
 
 
 
 
 
 
 
ece0bbe
4c69214
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
63216a8
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4c69214
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
# RhythmEnv β€” Architecture & Training Flow

Visual deep dive into how the env is structured and how the LLM agent
learns from it. All examples use concrete values (seed=42, sampled
profile, real numbers from the reward calculation).

---

## 1. System: three-layer separation

```
                β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
                β”‚   AGENT  (Qwen 2.5-3B + LoRA r=8, 4-bit)   β”‚
                β”‚                                            β”‚
                β”‚   Input:  prompt (state + history)         β”‚
                β”‚   Output: "3 7 5 DEEP_WORK"                β”‚
                β”‚           ↑   ↑   ↑     ↑                  β”‚
                β”‚        social|morn|work action             β”‚
                β”‚        belief|pref|pref                    β”‚
                β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                                  β”‚
                                  β”‚  N=8 completions per prompt
                                  β”‚  (sampling temp=1.5)
                                  β–Ό
              β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
              β”‚ ORCHESTRATION  (TRL GRPOTrainer + Unsloth)    β”‚
              β”‚                                               β”‚
              β”‚   β€’ Picks 1 prompt from dataset (~3000 rows)  β”‚
              β”‚   β€’ Generates 8 completions                   β”‚
              β”‚   β€’ Calls 4 reward functions in parallel      β”‚
              β”‚   β€’ Computes group-relative advantages        β”‚
              β”‚   β€’ Backprop on LoRA weights only (~30M)      β”‚
              β”‚   β€’ KL constraint to base Qwen (Ξ²=0.04)       β”‚
              β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                                β”‚
                                β”‚  env.reset(seed) β†’ step(action)
                                β”‚  (replay-based reward)
                                β–Ό
              β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
              β”‚   ENVIRONMENT  (RhythmEnvironment / FastAPI)  β”‚
              β”‚                                               β”‚
              β”‚   reset(seed=42)  β†’ samples profile           β”‚
              β”‚   step(action)    β†’ updates 5 meters,         β”‚
              β”‚                     returns observation       β”‚
              β”‚                     + per-step reward         β”‚
              β”‚                                               β”‚
              β”‚   Lives at: huggingface.co/spaces/...         β”‚
              β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
```

The agent never imports env code. They communicate via the OpenEnv
HTTP/WebSocket contract: `POST /reset`, `POST /step`, `GET /state`.

---

## 2. One episode = one week (28 steps)

```
                     ONE EPISODE = 1 WEEK = 28 STEPS

  Day 1 (Mon)        Day 2 (Tue)    ...    Day 7 (Sun)
 β”Œβ”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β” β”Œβ”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”    β”Œβ”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”
 β”‚  M  β”‚  A  β”‚  E  β”‚  N  β”‚ β”‚  M  β”‚  A  β”‚  E  β”‚  N  β”‚ .. β”‚  M  β”‚  A  β”‚  E  β”‚  N  β”‚
 β”‚  0  β”‚  1  β”‚  2  β”‚  3  β”‚ β”‚  4  β”‚  5  β”‚  6  β”‚  7  β”‚    β”‚ 24  β”‚ 25  β”‚ 26  β”‚ 27  β”‚
 β””β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”˜ β””β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”˜    β””β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”˜
   ↑                          ↑                                              ↑
   reset()              random event roll (8% per step)                   done=True
   meters: V=0.7        meditate? deep_work?                            grader runs
   C=0.7 P=0 S=0.7      sleep? socialize?                          final_score [0,1]
   Cn=0.5

         At each step the agent picks 1 of 10 actions:

   β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
   β”‚  PRODUCTIVITY  β”‚     RECOVERY    β”‚     SOCIAL    β”‚     LEISURE    β”‚
   β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
   β”‚  DEEP_WORK     β”‚  SLEEP          β”‚  FAMILY_TIME  β”‚  ME_TIME       β”‚
   β”‚  ADMIN_WORK    β”‚  EXERCISE       β”‚  SOCIALIZE    β”‚  BINGE_WATCH   β”‚
   β”‚  LEARN         β”‚  MEDITATE       β”‚               β”‚                β”‚
   β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

   Time-of-day multipliers (apply to action effects):
     Morning (M):    cognition gains Γ— 1.2   vitality drain Γ— 0.8
     Afternoon (A):  cognition gains Γ— 1.0   vitality drain Γ— 1.0
     Evening (E):    cognition gains Γ— 0.8   vitality drain Γ— 1.1
     Night (N):      cognition gains Γ— 0.6   vitality drain Γ— 1.3
                     (sleep BYPASSES these multipliers)

   Critical thresholds:
     Any meter < 0.1 at end of step β†’ -0.30 reward penalty (one per meter)
     Connection decays passively every step (-0.01 to -0.02 per profile)
```

---

## 3. State tree: what the env tracks (and what's hidden)

```
RhythmEnvironment instance state
β”‚
β”œβ”€β”€ _profile  ◄──── HIDDEN from agent ─────────┐
β”‚   β”‚                                          β”‚
β”‚   β”œβ”€β”€ name: "sampled_42"                     β”‚
β”‚   β”œβ”€β”€ reward_weights:                        β”‚
β”‚   β”‚     {V: 0.05, C: 0.05, P: 0.30,          β”‚
β”‚   β”‚      S: 0.20, Cn: 0.40}                  β”‚
β”‚   β”‚                                          β”‚  Used INTERNALLY
β”‚   β”œβ”€β”€ social_vitality_multiplier: 1.8        β”‚  to compute reward
β”‚   β”œβ”€β”€ morning_cognition_bonus: 1.5           β”‚  and modify action
β”‚   β”œβ”€β”€ evening_night_cognition_bonus: None    β”‚  effects.
β”‚   β”œβ”€β”€ morning_penalty: None                  β”‚  Belief reward
β”‚   β”œβ”€β”€ binge_shame: False                     β”‚  rewards INFERRING
β”‚   β”œβ”€β”€ progress_serenity_bonus: 0.04          β”‚  this profile.
β”‚   β”œβ”€β”€ idle_serenity_decay: 0.02              β”‚
β”‚   β”œβ”€β”€ vitality_decay_rate: 0.01              β”‚
β”‚   β”œβ”€β”€ stress_tolerance: 0.22                 β”‚
β”‚   β”œβ”€β”€ event_impact_multiplier: 0.7           β”‚
β”‚   β”œβ”€β”€ connection_decay_rate: 0.012           β”‚
β”‚   β”œβ”€β”€ solo_serenity_bonus: 0.05              β”‚
β”‚   β”œβ”€β”€ social_connection_multiplier: 1.4      β”‚
β”‚   β”œβ”€β”€ social_serenity_bonus: 0.03            β”‚
β”‚   β”œβ”€β”€ work_vitality_recovery: 0.03           β”‚
β”‚   └── (14 hidden parameters total)           β”‚
β”‚   β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
β”‚
β”‚   Reduces to a 3-D BELIEF VECTOR (the inference target):
β”‚   profile_to_belief_vector(profile) β†’ [0.30, 0.70, 0.50]
β”‚                                        β”‚     β”‚     β”‚
β”‚                                        β”‚     β”‚     work_pref
β”‚                                        β”‚     morning_pref
β”‚                                        social_pref
β”‚
β”œβ”€β”€ meters  ◄────── visible to agent ────────────────────────────┐
β”‚   β”œβ”€β”€ _vitality:    0.62  (range 0–1)                          β”‚
β”‚   β”œβ”€β”€ _cognition:   0.51                                       β”‚
β”‚   β”œβ”€β”€ _progress:    0.24                                       β”‚  Agent
β”‚   β”œβ”€β”€ _serenity:    0.71                                       β”‚  observes
β”‚   └── _connection:  0.38                                       β”‚  in prompt
β”‚   β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
β”‚
β”œβ”€β”€ _timestep: 5
β”‚
β”œβ”€β”€ _step_history: list[StepRecord]  ◄── visible to agent ──────┐
β”‚   β”œβ”€β”€ (last 7 steps)                                          β”‚
β”‚   β”‚   step 0: deep_work β†’ +0.42                               β”‚
β”‚   β”‚           deltas: V-0.10 C-0.12 P+0.18 S-0.05 Cn+0.00     β”‚
β”‚   β”‚           anomalies: V+0.00 C+0.00 P+0.06 S+0.00 Cn+0.00 ◄─ profile fingerprint
β”‚   β”‚   step 1: sleep β†’ +0.18                                   β”‚
β”‚   β”‚           deltas: V+0.20 C+0.10 P+0.00 S+0.05 Cn+0.00     β”‚
β”‚   β”‚           anomalies: V+0.00 C+0.00 P+0.00 S+0.00 Cn+0.00  β”‚
β”‚   β”‚   ...                                                     β”‚
β”‚   β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
β”‚
└── _step_rewards: [+0.42, +0.18, +0.31, +0.55, ...]
    └── used by grader to compute adaptation_score (late-half - early-half)
```

---

## 4. What the agent sees (concrete prompt)

```
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€ SYSTEM PROMPT ─────────────────────────────────┐
β”‚ You are a life-management agent helping a person whose preferences are HIDDEN.       β”‚
β”‚ Each step, output ONE LINE in this exact format:                                     β”‚
β”‚     S M W ACTION_NAME                                                                β”‚
β”‚ S = social pref (0=hates, 9=loves), M = morning, W = work                            β”‚
β”‚ ACTION_NAME ∈ {DEEP_WORK, ADMIN_WORK, LEARN, SLEEP, EXERCISE, MEDITATE,              β”‚
β”‚                FAMILY_TIME, SOCIALIZE, ME_TIME, BINGE_WATCH}                         β”‚
β”‚ Example: 3 8 7 DEEP_WORK                                                             β”‚
β”‚ Tactics: probe early, exploit late; don't repeat actions; ...                        β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€ USER PROMPT ────────────────────────────────────┐
β”‚ Step: 5/28 (Tuesday Afternoon)                                                       β”‚
β”‚ Remaining steps: 22                                                                  β”‚
β”‚                                                                                      β”‚
β”‚ Meters:                                                                              β”‚
β”‚   Vitality:   0.62                                                                   β”‚
β”‚   Cognition:  0.51                                                                   β”‚
β”‚   Progress:   0.24                                                                   β”‚
β”‚   Serenity:   0.71                                                                   β”‚
β”‚   Connection: 0.38                                                                   β”‚
β”‚                                                                                      β”‚
β”‚ Recent history (anom = how this person deviated from neutral baseline):              β”‚
β”‚   step 0: deep_work -> reward +0.42 (V-0.10 C-0.12 P+0.18 S-0.05 Cn+0.00)            β”‚
β”‚           [anom V+0.00 C+0.00 P+0.06 S+0.00 Cn+0.00]                                 β”‚
β”‚   step 1: sleep -> reward +0.18 (V+0.20 C+0.10 P+0.00 S+0.05 Cn+0.00)                β”‚
β”‚           [anom V+0.00 C+0.00 P+0.00 S+0.00 Cn+0.00]                                 β”‚
β”‚   step 2: socialize -> reward -0.05 (V-0.11 C-0.03 P+0.00 S+0.04 Cn+0.17)            β”‚
β”‚           [anom V-0.05 C+0.00 P+0.00 S+0.00 Cn+0.05]    ← strong profile signal     β”‚
β”‚   step 3: meditate -> reward +0.30 (V+0.03 C+0.08 P+0.00 S+0.20 Cn+0.00)             β”‚
β”‚           [anom V+0.00 C+0.00 P+0.00 S+0.05 Cn+0.00]    ← solo bonus visible        β”‚
β”‚   step 4: deep_work -> reward +0.18 (V-0.06 C-0.06 P+0.18 S+0.04 Cn+0.00)            β”‚
β”‚           [anom V+0.00 C+0.00 P+0.00 S+0.00 Cn+0.00]                                 β”‚
β”‚                                                                                      β”‚
β”‚ Output your belief, then your action (format: S M W ACTION_NAME):                    β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

LLM completion (1 of 8 sampled at temp=1.5):
                β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
                β”‚  "3 7 5 DEEP_WORK"   β”‚
                β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                           β–Ό
        Parsed: belief = [0.33, 0.78, 0.56]
                action = DEEP_WORK
```

---

## 5. The reward stack (4 layers, with concrete values for the example above)

```
                    β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
                    β”‚  Completion: "3 7 5 DEEP_WORK"      β”‚
                    β”‚  for prompt above (seed=42, step 5) β”‚
                    β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                                      β”‚
       β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
       β”‚                              β”‚                              β”‚
       β–Ό                              β–Ό                              β–Ό
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”            β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”          β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚ format_valid β”‚            β”‚   action_legal   β”‚          β”‚     env_reward     β”‚
β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€            β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€          β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
β”‚ parses?      β”‚            β”‚ action ∈ 10 ?    β”‚          β”‚ env.replay(seed=42,β”‚
β”‚ has belief?  β”‚            β”‚                  β”‚          β”‚   step5, action)   β”‚
β”‚              β”‚            β”‚ DEEP_WORK βœ“      β”‚          β”‚                    β”‚
β”‚ βœ“ +1.0       β”‚            β”‚   reward 0.0     β”‚          β”‚ deltas:            β”‚
β”‚ Γ— wt 0.05    β”‚            β”‚ Γ— wt 0.05        β”‚          β”‚   V-0.10 C-0.12    β”‚
β”‚              β”‚            β”‚                  β”‚          β”‚   P+0.18 S-0.05    β”‚
β”‚ β†’ +0.05      β”‚            β”‚  β†’ 0.00          β”‚          β”‚   Cn+0.00          β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜            β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜          β”‚                    β”‚
                                                          β”‚ profile_reward =   β”‚
                                                          β”‚   sum(deltas Γ—     β”‚
                                                          β”‚     prof_weights)  β”‚
                                                          β”‚     Γ— 15           β”‚
                                                          β”‚ = (-0.10Γ—0.05      β”‚
                                                          β”‚   + -0.12Γ—0.05     β”‚
                                                          β”‚   + 0.18Γ—0.30      β”‚
                                                          β”‚   + -0.05Γ—0.20     β”‚
                                                          β”‚   + 0Γ—0.40) Γ— 15   β”‚
                                                          β”‚ = +0.65            β”‚
                                                          β”‚                    β”‚
                                                          β”‚ + grader_bias:     β”‚
                                                          β”‚   0.5Γ—0.18+0.4Γ—0   β”‚
                                                          β”‚   = +0.09          β”‚
                                                          β”‚ + new-action: +0.07β”‚
                                                          β”‚ + b-act coupling:  β”‚
                                                          β”‚   work=0.56 (mid), β”‚
                                                          β”‚   no bonus = 0     β”‚
                                                          β”‚ - cycle pen: 0     β”‚
                                                          β”‚ - rep pen: 0       β”‚
                                                          β”‚                    β”‚
                                                          β”‚ env_reward = +0.81 β”‚
                                                          β”‚ Γ— wt 1.5 = +1.21   β”‚
                                                          β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
                                                                  β”‚
                                                                  β”‚
                                                                  β–Ό
                                                  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
                                                  β”‚     belief_accuracy         β”‚
                                                  β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
                                                  β”‚ true belief (sampled_42)    β”‚
                                                  β”‚   = [0.30, 0.70, 0.50]      β”‚
                                                  β”‚ agent belief                β”‚
                                                  β”‚   = [0.33, 0.78, 0.56]      β”‚
                                                  β”‚                             β”‚
                                                  β”‚ MAE = (0.03 + 0.08 + 0.06)  β”‚
                                                  β”‚       / 3 = 0.057           β”‚
                                                  β”‚ similarity = 0.943          β”‚
                                                  β”‚                             β”‚
                                                  β”‚ baseline (constant 0.5):    β”‚
                                                  β”‚   MAE = (0.20+0.20+0.00)/3  β”‚
                                                  β”‚       = 0.133               β”‚
                                                  β”‚   similarity = 0.867        β”‚
                                                  β”‚                             β”‚
                                                  β”‚ reward = 0.943 - 0.867      β”‚
                                                  β”‚       = +0.076              β”‚
                                                  β”‚ Γ— wt 3.0 = +0.23            β”‚
                                                  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

  Ξ£ TOTAL REWARD (this completion)
  = 0.05 + 0.00 + 1.21 + 0.23
  = +1.49
```

---

## 6. GRPO update step (8 completions β†’ 1 gradient update)

```
ONE TRAINING STEP (one row from the dataset β†’ one gradient update)

β”Œβ”€ pick prompt ────────────────────────────────────────────────────────┐
β”‚  prompt#1247 from dataset:                                           β”‚
β”‚    state: step 5, seed=42, history=[deep_work, sleep, ...]           β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
              β”‚
              β–Ό generate 8 completions @ temp=1.5
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚  c1: "3 7 5 DEEP_WORK"      β†’ reward +1.49                           β”‚
β”‚  c2: "5 5 5 SLEEP"          β†’ reward -0.21  (constant belief: -0.03  β”‚
β”‚                                              + sleep replay: -0.18)  β”‚
β”‚  c3: "3 6 4 ADMIN_WORK"     β†’ reward +1.10                           β”‚
β”‚  c4: "4 7 6 LEARN"          β†’ reward +1.32                           β”‚
β”‚  c5: "2 8 3 MEDITATE"       β†’ reward +0.45  (rep pen: -0.10 since    β”‚
β”‚                                              meditate at step 3 too) β”‚
β”‚  c6: "3 7 7 DEEP_WORK"      β†’ reward +1.55  (belief slightly better) β”‚
β”‚  c7: "5 5 5 EXERCISE"       β†’ reward -0.15                           β”‚
β”‚  c8: "4 6 5 FAMILY_TIME"    β†’ reward +0.92                           β”‚
β”‚                                                                      β”‚
β”‚  group_mean = +0.81                                                  β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
              β”‚
              β–Ό advantages = reward - group_mean
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚  ADVANTAGE (this is what GRPO actually backprops on)                 β”‚
β”‚  c1: +0.68  ← strongly preferred                                     β”‚
β”‚  c2: -1.02  ← strongly discouraged (constant belief)                 β”‚
β”‚  c3: +0.29                                                           β”‚
β”‚  c4: +0.51                                                           β”‚
β”‚  c5: -0.36                                                           β”‚
β”‚  c6: +0.74  ← most preferred                                         β”‚
β”‚  c7: -0.96                                                           β”‚
β”‚  c8: +0.11                                                           β”‚
β”‚                                                                      β”‚
β”‚  KEY INSIGHT: only the SPREAD matters. If all 8 had the              β”‚
β”‚  same reward, advantages would all be 0 β†’ no gradient.               β”‚
β”‚  This is why iter 1 mode-collapsed: format_valid +1.0 for            β”‚
β”‚  every completion meant zero variance from that layer.               β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜
              β”‚
              β–Ό policy loss = -E[ adv Γ— log_prob(completion) ]
              β”‚                + Ξ² Γ— KL(policy || base_qwen)
              β”‚
              β–Ό backprop (only LoRA weights, ~30M params)
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚  Model nudge:                                                        β”‚
β”‚    push: "3 7 5 DEEP_WORK"-like outputs UP                           β”‚
β”‚    push: "3 7 7 DEEP_WORK"-like outputs UP                           β”‚
β”‚    pull: "5 5 5 *"-like outputs DOWN                                 β”‚
β”‚                                                                      β”‚
β”‚  KL constraint: prevents the policy from diverging too               β”‚
β”‚  far from base Qwen (avoids gibberish drift).                        β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

  Repeat 800-2000 times. Each step ~3-5 sec on A100.
```

---

## 7. The dataset is just starting positions (not labels)

```
DATASET (~3000 rows, generated ONCE before training)

  For 200-300 episodes:
    env.reset(seed=N)
    for step in range(28):
      record {
        prompt: [system, user_for_this_state],
        seed: N,                            ◄─── replay metadata
        step_index: <current step>,
        action_history: <actions taken so far>,
        profile_mode: "continuous",
      }
      env.step(rollout_policy(obs))  ◄─ rollout=heuristic OR random
                                        (only matters for STATE diversity,
                                         the agent's training generations
                                         REPLACE these actions)

  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
  β”‚ A row from the dataset:                                          β”‚
  β”‚ {                                                                β”‚
  β”‚   prompt: [...full state at step 5 of episode seed=42...]        β”‚
  β”‚   seed: 42,                                                      β”‚
  β”‚   step_index: 5,                                                 β”‚
  β”‚   action_history: ["deep_work", "sleep", "socialize",            β”‚
  β”‚                    "meditate", "deep_work"],                     β”‚
  β”‚   profile_mode: "continuous"                                     β”‚
  β”‚ }                                                                β”‚
  β”‚                                                                  β”‚
  β”‚ NOTE: NO "correct action" or "label" anywhere.                   β”‚
  β”‚ The reward function reconstructs the env from this metadata      β”‚
  β”‚ and scores whatever action the LLM picks.                        β”‚
  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

  This is fundamentally different from supervised learning:
  - Supervised: (input, target_output) β€” model learns to mimic target
  - GRPO:       (prompt, replay_metadata) β€” model learns to maximize reward
```

---

## 8. Final episode grader (only fires at step 28)

```
                      _grade_episode() β€” runs at done=True

                              β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
                              β”‚  final_score β”‚
                              β”‚   ∈ [0, 1]   β”‚
                              β””β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”˜
        β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
        β”‚              β”‚             β”‚              β”‚              β”‚              β”‚
        β–Ό              β–Ό             β–Ό              β–Ό              β–Ό              β–Ό
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚ crash_free β”‚  β”‚  progress  β”‚  β”‚ connectionβ”‚  β”‚ adaptation β”‚  β”‚ efficiency β”‚  β”‚   belief   β”‚
β”‚   Γ— 0.15   β”‚  β”‚   Γ— 0.20   β”‚  β”‚   Γ— 0.10 β”‚  β”‚   Γ— 0.25   β”‚  β”‚   Γ— 0.10   β”‚  β”‚ accuracy   β”‚
β”‚            β”‚  β”‚            β”‚  β”‚          β”‚  β”‚            β”‚  β”‚            β”‚  β”‚   Γ— 0.20   β”‚
β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€  β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€  β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€  β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€  β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€  β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
β”‚ 1 - crashesβ”‚  β”‚ final P    β”‚  β”‚ final Cn β”‚  β”‚ late-half  β”‚  β”‚ avg_reward β”‚  β”‚ 1 - MAE    β”‚
β”‚ /total_ck  β”‚  β”‚ value      β”‚  β”‚ value    β”‚  β”‚ mean rewardβ”‚  β”‚ normalized β”‚  β”‚ vs true    β”‚
β”‚            β”‚  β”‚            β”‚  β”‚          β”‚  β”‚ - early    β”‚  β”‚ to [0,1]   β”‚  β”‚ profile    β”‚
β”‚ e.g. 0.95  β”‚  β”‚ e.g. 0.42  β”‚  β”‚ e.g. 0.51β”‚  β”‚ e.g. +0.18 β”‚  β”‚ e.g. 0.55  β”‚  β”‚ e.g. 0.80  β”‚
β”‚ Γ—0.15=0.14 β”‚  β”‚ Γ—0.20=0.084β”‚  β”‚ Γ—0.10=0.05β”‚ β”‚ Γ—0.25=0.045β”‚  β”‚ Γ—0.10=0.055β”‚  β”‚ Γ—0.20=0.16 β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

                        Ξ£ = 0.14 + 0.084 + 0.05 + 0.045 + 0.055 + 0.16
                          = 0.534  ← final_score (with inference)

Heuristic / random baselines never call env.record_belief(), so the belief
component scores 0 for them β€” by design: the meta-RL skill is INFERENCE,
and only agents that actually try get credit on this axis.

Plus a sparse terminal reward (added to step 27's per-step reward):
  terminal_bonus = (final_score - 0.5) Γ— 5  β†’  e.g. (0.534 - 0.5) Γ— 5 = +0.17

This means: at step 27, agent gets last per-step reward + bonus from grader.
This is the only direct gradient signal pointing at the actual episode quality.
```

---

## 9. Three eval conditions (post-training)

```
                       inference_eval.py runs ALL THREE

  β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
  β”‚  discrete-3-profiles       β”‚  continuous-in-distributionβ”‚  continuous-OOD           β”‚
  β”‚  (3 reference profiles)    β”‚  (was the agent able to    β”‚  (does meta-policy        β”‚
  β”‚                            β”‚   learn the meta-policy?)  β”‚   generalize?)            β”‚
  β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
  β”‚ env.reset(seed=N,          β”‚ env.reset(seed=N)          β”‚ env.reset(seed=10000+N)   β”‚
  β”‚   profile="introvert_      β”‚   ← samples from training  β”‚   ← samples from a region β”‚
  β”‚            morning")       β”‚     distribution           β”‚     never seen in trainingβ”‚
  β”‚                            β”‚                            β”‚                           β”‚
  β”‚ β†’ 3 hardcoded profiles     β”‚ β†’ ~10 sampled profiles     β”‚ β†’ ~10 sampled profiles    β”‚
  β”‚   (from PROFILES list)     β”‚   from seeds 100..110      β”‚   from seeds 10000..10010 β”‚
  β”‚                            β”‚                            β”‚                           β”‚
  β”‚ Each strategy plays each   β”‚ Each strategy plays each   β”‚ Each strategy plays each  β”‚
  β”‚ profile 5x = 15 episodes   β”‚ seed 1x = 10 episodes      β”‚ seed 1x = 10 episodes     β”‚
  β”‚                            β”‚                            β”‚                           β”‚
  β”‚ Strategies tested:         β”‚ Strategies tested:         β”‚ Strategies tested:        β”‚
  β”‚  β€’ random                  β”‚  β€’ random                  β”‚  β€’ random                 β”‚
  β”‚  β€’ heuristic               β”‚  β€’ heuristic               β”‚  β€’ heuristic              β”‚
  β”‚  β€’ model (trained Qwen)    β”‚  β€’ model (trained Qwen)    β”‚  β€’ model (trained Qwen)   β”‚
  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

  THE KEY METRIC: trained model's score on continuous-OOD vs heuristic baseline.

  Heuristic baseline (profile-blind hand rules): score 0.580 on OOD.
  Trained meta-RL agent's target: > 0.580 on OOD.

  If the trained agent beats the heuristic on OOD (profiles never seen in
  training), that's direct proof it learned the SKILL of profile inference,
  not just memorized training profiles.
```

---

## 10. Sim β†’ Real mapping (why this POC matters)

The whole point of training in this env is that the SKILL transfers. The
mapping from simulated signals to production signals is direct: same
information, different source.

```
β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€ SIMULATION (training) ───────┐  β”Œβ”€β”€β”€β”€ REAL PRODUCT (deployment) ───────┐
β”‚                                                     β”‚  β”‚                                       β”‚
β”‚  meter: Vitality = 0.62                             │──┼─►  HRV from watch + sleep score      β”‚
β”‚  meter: Cognition = 0.51                            │──┼─►  focus app metrics + screen time   β”‚
β”‚  meter: Progress = 0.24                             │──┼─►  calendar density + task completionβ”‚
β”‚  meter: Serenity = 0.71                             │──┼─►  HRV variability + voice sentiment β”‚
β”‚  meter: Connection = 0.38                           │──┼─►  call/message freq + calendar      β”‚
β”‚                                                     β”‚  β”‚                                       β”‚
β”‚  hidden profile (sampled per episode)               │──┼─►  the real user's actual personalityβ”‚
β”‚    └─ never visible to agent                        β”‚  β”‚     └─ also never visible             β”‚
β”‚                                                     β”‚  β”‚                                       β”‚
β”‚  agent emits: "3 7 5 DEEP_WORK"                     β”‚  β”‚                                       β”‚
β”‚    β”œβ”€ "3 7 5" = belief about user (hidden internal) │──┼─►  agent's internal user model       β”‚
β”‚    └─ "DEEP_WORK" = action choice                   │──┼─►  recommendation to user            β”‚
β”‚                                                     β”‚  β”‚     ("I suggest a focus block now")  β”‚
β”‚                                                     β”‚  β”‚                                       β”‚
β”‚  per-step env_reward                                │──┼─►  meter trend + accept/ignore tap   β”‚
β”‚    └─ "did meters improve under profile weights?"   β”‚  β”‚     └─ "did user accept and benefit?"β”‚
β”‚                                                     β”‚  β”‚                                       β”‚
β”‚  episode = 1 week (28 steps)                        │──┼─►  rolling weekly window             β”‚
β”‚                                                     β”‚  β”‚                                       β”‚
β”‚  3 eval conditions                                  β”‚  β”‚   USER ARRIVES (cold start)          β”‚
β”‚    β”œβ”€ discrete-3 (memorization check)               β”‚  β”‚   β”œβ”€ Agent has zero info about them  β”‚
β”‚    β”œβ”€ continuous-in-dist (training-distrib)         β”‚  β”‚   β”œβ”€ ~5-10 interactions to converge  β”‚
β”‚    └─ continuous-OOD ◄── KEY ──►                    │──┼─►   on a confident belief vector     β”‚
β”‚       agent must infer profile NEVER seen in        β”‚  β”‚   └─ Acts on belief, learns from     β”‚
β”‚       training. THIS is the production scenario.    β”‚  β”‚      tap responses                   β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜  β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

SUCCESS METRIC IN BOTH WORLDS: how fast does the agent personalize?

Sim:  belief_accuracy curve over a single episode (does it climb?)
      adaptation_score (late-half mean reward > early-half mean reward?)

Real: how many interactions until user's accept-rate stabilizes high?
      How quickly does the agent stop suggesting things the user always ignores?

Both reduce to the same skill: form a model of this person from limited
observation, then act on it.
```

**Design choices that explicitly preserve the passive-only constraint:**
- ❌ No "ASK USER" action β€” agent never quizzes
- ❌ No profile feature in observation at eval time β€” agent must infer
- ❌ No explicit "what's your goal?" prompt β€” only sensor-equivalent meters
- βœ… Belief output is INTERNAL (visible in POC only for training reward)
- βœ… Reward formula uses only quantities computable from passive signals

If the agent learns to do well on continuous-OOD (profiles never seen in
training), it has acquired the skill of "figure out a new person from
observation alone" β€” exactly the skill the production assistant needs
when meeting a real user for the first time.

---

## 11. Spend & timing (concrete)

```
HARDWARE: A100 large (80GB) on HF Jobs at $2.50/hr ($0.0417/min)

   FAST_MODE (200-800 steps):
     dataset gen:    ~2 min
     model load:     ~3 min
     training:       ~10-25 min  (depends on steps)
     eval:           ~3 min
     plot + upload:  ~2 min
     ────────────────────────────
     total:          20-35 min   ($0.80-1.50 per iter)

   FULL RUN (2000 steps):
     dataset gen:    ~3 min
     model load:     ~3 min
     training:       ~60-90 min
     eval:           ~3 min
     plot + upload:  ~2 min
     ────────────────────────────
     total:          70-100 min  ($3-4)

  Iter 1 (200 steps):    $0.50  ❌ mode collapse (single action)
  Iter 2 (400 steps):    $1.50  ❌ mode collapse (2-cycle)
  Iter 3 (800 steps):    $5     ⏳ in flight (control)
  Iter 4 (800 steps):    $5     ⏳ in flight (with full fixes)
  Final (2000 steps):    $4     ⏳ pending iter 3+4 results
                       ──────
                       ~$16    of $30 budget
```