dsfb-computer-graphics 0.1.1

Minimal DSFB-for-computer-graphics research artifact for temporal accumulation supervision
Documentation
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
# Final Blocker Removal Plan

This document replaces the earlier upgrade plan with a blocker-complete execution map for the current `dsfb-computer-graphics` crate.

Scope boundary:

- Only files inside `crates/dsfb-computer-graphics` may change.
- No workspace-root edits are permitted.
- If a stronger result would normally require engine integration or external benchmark data, this crate must state that limitation explicitly rather than imply it is solved.

## Current Audit Summary

The existing crate is already substantive:

- deterministic multi-scenario Demo A and Demo B flows exist
- host-realistic DSFB mode exists
- stronger heuristics and ablations exist
- fixed-budget equality exists in Demo B
- integration and cost documents exist

The current binding blockers remain real in the audited code and generated outputs:

1. No timing artifact or hardware-facing measurement path exists.
2. Point-like ROI scenarios are headline-visible but not analytically disclosed as point measurements.
3. `trust_error_rank_correlation` is still emitted even when trust occupies only one to three bins.
4. Trust behaves as a near-binary gate in the current implementation.
5. Motion disagreement has no effect in the current ablation suite.
6. Neighborhood clamp failure is shown numerically but not explained mechanically.
7. Demo B still leans on thin sub-pixel structure and aliasing-sensitive wins.
8. No credible resolution scaling artifact exists.
9. Hazard weights and thresholds remain scattered and hand-set.
10. Non-ROI penalty is surfaced numerically but not bounded or interpreted well enough.

## File-Level Implementation Map

### Existing files to change

- `Cargo.toml`
- `README.md`
- `docs/cost_model.md`
- `docs/gpu_implementation.md`
- `docs/integration_surface.md`
- `src/cli.rs`
- `src/config.rs`
- `src/cost.rs`
- `src/dsfb/mod.rs`
- `src/frame.rs`
- `src/host.rs`
- `src/lib.rs`
- `src/metrics.rs`
- `src/pipeline.rs`
- `src/plots.rs`
- `src/report.rs`
- `src/sampling.rs`
- `src/scene/mod.rs`
- `src/taa/mod.rs`
- `tests/colab_notebook.rs`
- `tests/demo_cli.rs`
- `tests/research_artifact.rs`

### New files to add

- `docs/gpu_path.md`
- `docs/validation_contract.md`
- `src/parameters.rs`
- `src/timing.rs`
- generated reports and figures emitted by code:
  - `generated/timing_report.md`
  - `generated/timing_metrics.json`
  - `generated/trust_diagnostics.md`
  - `generated/trust_diagnostics.json`
  - `generated/resolution_scaling_report.md`
  - `generated/resolution_scaling_metrics.json`
  - `generated/parameter_sensitivity_report.md`
  - `generated/parameter_sensitivity_metrics.json`
  - `generated/demo_b_efficiency_report.md`

## Blocker-by-Blocker Plan

### 1. Timing / hardware path

Files:

- `src/timing.rs`
- `src/cost.rs`
- `src/pipeline.rs`
- `src/report.rs`
- `src/cli.rs`
- `docs/gpu_path.md`

Work:

- add a reproducible timing command that measures crate-local CPU proxy timings for:
  - scalar reference path
  - streamlined minimum supervisory path
  - host-realistic path
  - full debug/reporting path
- emit explicit per-stage timing, operations per pixel, memory traffic estimate, and optimization levers
- add a minimal GPU-oriented compute scaffold description and pass decomposition inside the crate
- label the result explicitly as CPU-measured proxy unless actual GPU timing is available at run time

Acceptance:

- `generated/timing_report.md` and `generated/timing_metrics.json` exist
- timing report explicitly states one of:
  - `GPU-measured`
  - `GPU-estimated`
  - `CPU-only proxy`
- no unsupported production language appears

Fail gates:

- validation fails if timing artifacts are missing
- validation fails if timing classification text is absent
- validation fails if unsupported phrases such as `production-ready` appear

### 2. ROI disclosure / scenario / metric repair

Files:

- `src/scene/mod.rs`
- `src/metrics.rs`
- `src/report.rs`
- `src/plots.rs`
- `src/pipeline.rs`

Work:

- add ROI taxonomy metadata:
  - point-like ROI
  - region ROI
  - negative control
- add ROI size and area-fraction reporting everywhere
- add at least two new benefit-expected region-ROI scenarios with ROI > 15 pixels
- ensure at least one benefit-expected scenario has ROI > 50 pixels
- add imperfect-motion / noisy-reprojection scenarios with subpixel or biased motion behavior
- add resolution scaling study across small, medium, and high-resolution configurations
- separate aggregate reporting for point-like and region ROI scenarios

Acceptance:

- every decision-facing report lists ROI size and scenario category
- point-like ROI scenarios are labeled explicitly
- new region-ROI scenarios are present in both metrics and reports
- `generated/resolution_scaling_report.md` and `generated/resolution_scaling_metrics.json` exist

Fail gates:

- validation fails if ROI size is missing
- validation fails if point-like ROI scenarios are unlabeled
- validation fails if aggregate tables mix point-like and region ROI without disclosure

### 3. Trust calibration cleanup

Files:

- `src/host.rs`
- `src/parameters.rs`
- `src/metrics.rs`
- `src/plots.rs`
- `src/report.rs`
- `src/pipeline.rs`

Work:

- centralize trust mapping and hazard merge behavior
- add trust diagnostics:
  - histogram
  - occupied bins
  - entropy / discreteness score
  - calibration bins
  - effective trust-level count
  - operating-mode classification
- quarantine or remove degenerate rank-correlation as a headline metric
- optionally add a smoother graded mode if it improves interpretability without misrepresenting the method

Acceptance:

- `generated/trust_diagnostics.md` and `generated/trust_diagnostics.json` exist
- decision-facing reports state whether the current mode is gate-like or graded
- trust rank correlation is never used as primary calibration evidence when occupancy is degenerate

Fail gates:

- validation fails if trust diagnostics are missing
- validation fails if near-binary behavior is not disclosed when present
- validation fails if calibration quality is claimed without multi-bin support

### 4. Motion disagreement resolution

Files:

- `src/scene/mod.rs`
- `src/host.rs`
- `src/metrics.rs`
- `src/report.rs`
- `docs/integration_surface.md`
- `docs/cost_model.md`

Work:

- add at least one motion-stress scenario with vector bias, boundary mismatch, or motion inconsistency
- compare minimum path versus motion-augmented path
- either:
  - show where motion materially matters, or
  - demote motion from the minimum path and document it as optional

Acceptance:

- ablation report explains motion contribution with scenario evidence
- the minimum path does not include motion disagreement unless evidence justifies it

Fail gates:

- validation fails if the minimum path still includes motion disagreement while all motion evidence is null
- validation fails if motion is kept but not explained analytically

### 5. Neighborhood clamp explanation

Files:

- `src/metrics.rs`
- `src/report.rs`

Work:

- add mechanistic diagnostics for neighborhood clamp:
  - ROI trigger rate
  - clamp-distance statistics
  - cases where history stays inside the local neighborhood hull
- write explicit explanation for why the clamp misses thin reveal cases

Acceptance:

- main report and blocker report explain the failure mode instead of only listing the bad score

Fail gates:

- validation fails if neighborhood clamp is discussed only numerically and not mechanically

### 6. Demo B anti-confound strengthening

Files:

- `src/scene/mod.rs`
- `src/sampling.rs`
- `src/metrics.rs`
- `src/plots.rs`
- `src/report.rs`
- `src/pipeline.rs`

Work:

- add non-subpixel and mixed-width sampling scenes
- add at least one variance-driven or texture-driven scene where edge guidance is not enough
- add native-trust and hybrid guidance variants under strict fixed-budget equality
- expand budget-efficiency curves to multiple budgets
- separate aliasing, variance, and prioritization interpretations in reporting

Acceptance:

- `generated/demo_b_efficiency_report.md` and updated `generated/demo_b_metrics.json` exist
- Demo B decision reporting explicitly distinguishes aliasing recovery from allocation quality
- fixed-budget equality remains exact for every policy

Fail gates:

- validation fails if Demo B still only demonstrates thin sub-pixel evidence
- validation fails if fixed-budget equality is broken
- validation fails if aliasing-versus-allocation distinction is missing from decision-facing reports

### 7. Hazard parameter centralization and sensitivity

Files:

- `src/parameters.rs`
- `src/host.rs`
- `src/taa/mod.rs`
- `src/metrics.rs`
- `src/pipeline.rs`
- `src/report.rs`
- `src/plots.rs`

Work:

- move all DSFB hazard weights, trust mapping constants, smoothstep edges, grammar strengths, and alpha bounds into centralized parameter structs
- centralize baseline thresholds as well
- run systematic one-at-a-time sensitivity sweeps over key parameters
- classify parameters as robust, fragile, or neutral

Acceptance:

- `generated/parameter_sensitivity_report.md` and `generated/parameter_sensitivity_metrics.json` exist
- reports state whether parameters are hand-set, sensitivity-vetted, or calibrated
- magic constants are no longer scattered in code paths

Fail gates:

- validation fails if sensitivity artifacts are missing
- validation fails if reports still describe weights opaquely
- validation fails if key thresholds remain hard-coded in multiple logic sites

### 8. Reporting overhaul

Files:

- `src/report.rs`
- `src/pipeline.rs`
- `src/plots.rs`

Work:

- regenerate all decision-facing reports from the richer metrics
- add explicit sections for:
  - what is proven
  - what is not proven
  - remaining blockers
  - ROI disclosure
  - trust operating mode
  - motion relevance
  - neighborhood clamp mechanics
  - non-ROI penalties
  - timing path honesty

Acceptance:

- every required report exists
- every decision-facing report contains `What Is Not Proven`
- every decision-facing report contains `Remaining Blockers`

Fail gates:

- validation fails if any required report is missing
- validation fails if unsupported claims appear

### 9. Hard validation gates and CLI usability

Files:

- `src/cli.rs`
- `src/pipeline.rs`
- `tests/demo_cli.rs`
- `tests/research_artifact.rs`
- `tests/colab_notebook.rs`
- `README.md`
- `docs/validation_contract.md`

Work:

- add commands for:
  - `run-all`
  - `run-scenario`
  - `run-ablations`
  - `run-timing`
  - `run-resolution-scaling`
  - `run-sensitivity`
  - `run-demo-b`
  - `run-demo-b-efficiency`
  - `validate`
  - `export-minimal-report`
- make validation enforce artifact completeness, honesty sections, timing classification, ROI disclosure, and budget equality

Acceptance:

- one command generates the complete artifact bundle
- one command validates it
- single-purpose study commands also work

Fail gates:

- validation fails if required artifacts, figures, JSON files, or disclosures are missing
- tests fail if CLI commands regress

## Required Experiments and Outputs

### New or expanded scenario studies

- existing point-like scenarios retained and explicitly labeled:
  - `thin_reveal`
  - `diagonal_reveal`
- existing benefit region scenario retained:
  - `fast_pan`
- new region-ROI scenarios to add:
  - multi-segment reveal band
  - mixed-width reveal band
  - imperfect-motion / reprojection-bias reveal
- existing negative controls retained:
  - `contrast_pulse`
  - `stability_holdout`

### Generated outputs that must exist after `run-all`

- `generated/report.md`
- `generated/reviewer_summary.md`
- `generated/five_mentor_audit.md`
- `generated/check_signing_blockers.md`
- `generated/ablation_report.md`
- `generated/trust_diagnostics.md`
- `generated/trust_diagnostics.json`
- `generated/timing_report.md`
- `generated/timing_metrics.json`
- `generated/resolution_scaling_report.md`
- `generated/resolution_scaling_metrics.json`
- `generated/parameter_sensitivity_report.md`
- `generated/parameter_sensitivity_metrics.json`
- `generated/demo_b_decision_report.md`
- `generated/demo_b_efficiency_report.md`
- updated `generated/demo_b/metrics.json`

### Figures that must be emitted from code

- trust histogram
- trust calibration plot
- ROI taxonomy / ROI size figure
- timing breakdown figure
- resolution scaling figure
- parameter sensitivity figure
- motion relevance figure
- Demo B efficiency figure

## Execution Order

1. Write this plan and the validation contract.
2. Centralize parameters and expand scenario metadata.
3. Add region-ROI and imperfect-motion scenarios.
4. Implement timing pipeline and GPU-path documentation.
5. Add trust diagnostics and motion-resolution logic.
6. Strengthen Demo B against aliasing-only interpretation.
7. Add sensitivity and resolution studies.
8. Rewrite reports and figures from generated metrics.
9. Add hard validation gates and final CLI/README usability updates.