File size: 15,091 Bytes
4adaccc
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
{
  "claims": [
    {
      "actual_model_or_dataset_used": true,
      "assessment": "verified",
      "claim": 1,
      "claim_object_match": "exact",
      "control_artifacts": [
        "outputs/claim1.json"
      ],
      "destructive_control": true,
      "destructive_control_executed": true,
      "destructive_or_boundary_control": "Holding the predictive vector family fixed while changing rho and the spurious vector changes the exact and leading risks; the zero-coupling boundary remains separately visible.",
      "direct_evidence": true,
      "evidence_tier": "literal_claim_experiment",
      "executed_outputs": [
        "outputs/claim1.json",
        "outputs/results.json"
      ],
      "expected_points": 2,
      "independent_evidence": [
        "outputs/claim1.json",
        "replay_a/claim1.json",
        "replay_b/claim1.json"
      ],
      "independent_oracle": "Exact finite-matrix fixed-point evaluation and the displayed trace formula are independent computational routes.",
      "limitation": "Universal theorem quantifiers remain supplied by the pinned proof; finite execution checks the displayed objects and the registered direction/risk consequences.",
      "literal_claim": "In the population setting, Theorem 1 characterizes excess risk as a function of the magnitude and direction of the performative effect together with spurious features (Section 4, Theorem 1).",
      "native_scale_justification": "The registered closed-form population/proportional objects and the released proportional RRM mechanism are executed directly; stochastic cells use p=88>n=80 and fixed paired seeds.",
      "not_proxy_reason": "The displayed population recursion, deterministic-equivalent specialization, and released proportional/perforidge.py update are the registered mechanisms, not a nearby task.",
      "oracle_artifacts": [
        "replay_a/claim1.json",
        "replay_b/claim1.json"
      ],
      "paper_native_mechanism": "Evaluates the paper's population fixed-point recursion and first-order excess-risk expression on three covariance couplings and four performativity scales.",
      "paper_or_released_scale": true,
      "registered_system_executed": true,
      "result": "Twelve block-covariance cells vary performative magnitude, direction and spurious coordinates; the displayed leading risk tracks the exact fixed point with maximum absolute residual 0.000991884.",
      "scope_boundary": "The verdict is bound to the literal registered claim and the executed identity/block-covariance specializations covered by the pinned theorem statements.",
      "source_locator": "Pinned body.tex, Theorem 1 / thm:pop and eq:fppopavg",
      "upstream_source_commit": "370fcd19199313c53da310d861a1a9fbd73b731d",
      "upstream_source_digest": "sha256:d6baf4926bb033040d389d896c3f7d0d8b124404a7e6bd2240a7883a917a0b88"
    },
    {
      "actual_model_or_dataset_used": true,
      "assessment": "falsified",
      "claim": 2,
      "claim_object_match": "exact",
      "control_artifacts": [
        "outputs/claim2.json"
      ],
      "destructive_control": true,
      "destructive_control_executed": true,
      "destructive_or_boundary_control": "A nonuniform b vector at the same mean produces strictly positive optimal risk, so the zero is the literal constant-b exception rather than a broken risk implementation.",
      "direct_evidence": true,
      "evidence_tier": "literal_claim_experiment",
      "executed_outputs": [
        "outputs/claim2.json",
        "outputs/results.json"
      ],
      "expected_points": 2,
      "independent_evidence": [
        "outputs/claim2.json",
        "replay_a/claim2.json",
        "replay_b/claim2.json"
      ],
      "independent_oracle": "Both the exact fixed-point recursion and Corollary-2 leading expression independently return zero at the pinned exception.",
      "limitation": "Universal theorem quantifiers remain supplied by the pinned proof; finite execution checks the displayed objects and the registered direction/risk consequences.",
      "literal_claim": "Corollary 2 shows the optimal regularization parameter in the population regime is proportional to the strength of the performative effect, with optimal risk remaining strictly positive (Section 4, Corollary 2).",
      "native_scale_justification": "The registered closed-form population/proportional objects and the released proportional RRM mechanism are executed directly; stochastic cells use p=88>n=80 and fixed paired seeds.",
      "not_proxy_reason": "The displayed population recursion, deterministic-equivalent specialization, and released proportional/perforidge.py update are the registered mechanisms, not a nearby task.",
      "oracle_artifacts": [
        "replay_a/claim2.json",
        "replay_b/claim2.json"
      ],
      "paper_native_mechanism": "Executes Corollary 2's identity-covariance formula and the exact population fixed-point recursion at the source's explicit constant-b boundary.",
      "paper_or_released_scale": true,
      "registered_system_executed": true,
      "result": "The registered strict-positivity conjunction is false: at Sigma=I, constant nonzero b=0.2 and lambda*=b, the exact fixed-point risk is 0.0 and displayed leading risk is 6.939e-18; a nonuniform-b control is positive at 0.0132.",
      "scope_boundary": "The verdict is bound to the literal registered claim and the executed identity/block-covariance specializations covered by the pinned theorem statements.",
      "source_locator": "Pinned body.tex, contribution paragraph and Corollary 2 / cor:pop",
      "upstream_source_commit": "370fcd19199313c53da310d861a1a9fbd73b731d",
      "upstream_source_digest": "sha256:d6baf4926bb033040d389d896c3f7d0d8b124404a7e6bd2240a7883a917a0b88"
    },
    {
      "actual_model_or_dataset_used": true,
      "assessment": "verified",
      "claim": 3,
      "claim_object_match": "exact",
      "control_artifacts": [
        "outputs/claim3.json"
      ],
      "destructive_control": true,
      "destructive_control_executed": true,
      "destructive_or_boundary_control": "The theorem expression is computed without samples while the finite estimator uses forty independent Gaussian matrices; agreement therefore cannot be a shared-data identity.",
      "direct_evidence": true,
      "evidence_tier": "literal_claim_experiment",
      "executed_outputs": [
        "outputs/claim3.json",
        "outputs/results.json"
      ],
      "expected_points": 2,
      "independent_evidence": [
        "outputs/claim3.json",
        "replay_a/claim3.json",
        "replay_b/claim3.json"
      ],
      "independent_oracle": "Closed-form deterministic equivalent and independent Monte Carlo execution of the released ridge recursion agree within sampling error.",
      "limitation": "Universal theorem quantifiers remain supplied by the pinned proof; finite execution checks the displayed objects and the registered direction/risk consequences.",
      "literal_claim": "Theorem 3 establishes a deterministic equivalent of the performative fixed point for over-parameterized ridge regression when the number of features exceeds the number of samples (Section 5, Theorem 3).",
      "native_scale_justification": "The registered closed-form population/proportional objects and the released proportional RRM mechanism are executed directly; stochastic cells use p=88>n=80 and fixed paired seeds.",
      "not_proxy_reason": "The displayed population recursion, deterministic-equivalent specialization, and released proportional/perforidge.py update are the registered mechanisms, not a nearby task.",
      "oracle_artifacts": [
        "replay_a/claim3.json",
        "replay_b/claim3.json"
      ],
      "paper_native_mechanism": "Specializes the Theorem-3 deterministic equivalent to Sigma=I and compares it with the released two-deployment proportional ridge recursion.",
      "paper_or_released_scale": true,
      "registered_system_executed": true,
      "result": "For p=88>n=80, 40 two-deployment Gaussian runs give mean excess risk 0.336192 versus deterministic equivalent 0.328403; gap 0.007789 is below one empirical SE 0.010002.",
      "scope_boundary": "The verdict is bound to the literal registered claim and the executed identity/block-covariance specializations covered by the pinned theorem statements.",
      "source_locator": "Pinned body.tex, Theorem 3 / thm:over and pinned official proportional/perforidge.py",
      "upstream_source_commit": "370fcd19199313c53da310d861a1a9fbd73b731d",
      "upstream_source_digest": "sha256:d6baf4926bb033040d389d896c3f7d0d8b124404a7e6bd2240a7883a917a0b88"
    },
    {
      "actual_model_or_dataset_used": true,
      "assessment": "verified",
      "claim": 4,
      "claim_object_match": "exact",
      "control_artifacts": [
        "outputs/claim4.json"
      ],
      "destructive_control": true,
      "destructive_control_executed": true,
      "destructive_or_boundary_control": "The zero-performativity optimum is recomputed at every noise level and subtraction reverses sign only across the registered noise regimes.",
      "direct_evidence": true,
      "evidence_tier": "literal_claim_experiment",
      "executed_outputs": [
        "outputs/claim4.json",
        "outputs/results.json"
      ],
      "expected_points": 2,
      "independent_evidence": [
        "outputs/claim4.json",
        "replay_a/claim4.json",
        "replay_b/claim4.json"
      ],
      "independent_oracle": "Bounded scalar optimization is independently repeated for b=0 and b>0 at each noise value.",
      "limitation": "Universal theorem quantifiers remain supplied by the pinned proof; finite execution checks the displayed objects and the registered direction/risk consequences.",
      "literal_claim": "Theorem 4 shows the optimal regularization moves in the same direction as the performative effect on predictive features under low noise, but in the opposite direction under high noise, in the over-parameterized regime (Section 5, Theorem 4).",
      "native_scale_justification": "The registered closed-form population/proportional objects and the released proportional RRM mechanism are executed directly; stochastic cells use p=88>n=80 and fixed paired seeds.",
      "not_proxy_reason": "The displayed population recursion, deterministic-equivalent specialization, and released proportional/perforidge.py update are the registered mechanisms, not a nearby task.",
      "oracle_artifacts": [
        "replay_a/claim4.json",
        "replay_b/claim4.json"
      ],
      "paper_native_mechanism": "Numerically minimizes the displayed deterministic-equivalent specialization with identical kappa and b on both sides of the noise transition.",
      "paper_or_released_scale": true,
      "registered_system_executed": true,
      "result": "Exact deterministic-equivalent optimization moves lambda by +0.000692 at noise 0.2, but by -0.002341 and -0.006853 at noise 0.7 and 1.0, reproducing the registered low/high-noise sign reversal.",
      "scope_boundary": "The verdict is bound to the literal registered claim and the executed identity/block-covariance specializations covered by the pinned theorem statements.",
      "source_locator": "Pinned body.tex, Theorem 4 consequences / relations1a, relations1b",
      "upstream_source_commit": "370fcd19199313c53da310d861a1a9fbd73b731d",
      "upstream_source_digest": "sha256:d6baf4926bb033040d389d896c3f7d0d8b124404a7e6bd2240a7883a917a0b88"
    },
    {
      "actual_model_or_dataset_used": true,
      "assessment": "verified",
      "claim": 5,
      "claim_object_match": "exact",
      "control_artifacts": [
        "outputs/claim5.json"
      ],
      "destructive_control": true,
      "destructive_control_executed": true,
      "destructive_or_boundary_control": "The paired b=0 curve is a destructive contrast: it removes performativity while preserving every Gaussian draw and the full lambda grid.",
      "direct_evidence": true,
      "evidence_tier": "literal_claim_experiment",
      "executed_outputs": [
        "outputs/claim5.json",
        "outputs/results.json"
      ],
      "expected_points": 2,
      "independent_evidence": [
        "outputs/claim5.json",
        "replay_a/claim5.json",
        "replay_b/claim5.json"
      ],
      "independent_oracle": "Sixteen paired seeds and a twelve-point lambda sweep independently identify both minima and their risk difference.",
      "limitation": "Universal theorem quantifiers remain supplied by the pinned proof; finite execution checks the displayed objects and the registered direction/risk consequences.",
      "literal_claim": "Numerical experiments in Section 6 confirm that in the over-parameterized setting, performative effects can improve optimally-regularized risk when performativity reinforces existing trends, contrasting with the population-regime degradation (Section 6).",
      "native_scale_justification": "The registered closed-form population/proportional objects and the released proportional RRM mechanism are executed directly; stochastic cells use p=88>n=80 and fixed paired seeds.",
      "not_proxy_reason": "The displayed population recursion, deterministic-equivalent specialization, and released proportional/perforidge.py update are the registered mechanisms, not a nearby task.",
      "oracle_artifacts": [
        "replay_a/claim5.json",
        "replay_b/claim5.json"
      ],
      "paper_native_mechanism": "Runs five deployments of the released proportional/perforidge.py mechanism for b=0 and reinforcing b=0.2 with identical seeds.",
      "paper_or_released_scale": true,
      "registered_system_executed": true,
      "result": "Sixteen paired released-mechanism runs over twelve lambdas move the optimum from 0.05 to 0.07 and reduce optimal risk from 0.274612 to 0.268981, an improvement of 0.005630.",
      "scope_boundary": "The verdict is bound to the literal registered claim and the executed identity/block-covariance specializations covered by the pinned theorem statements.",
      "source_locator": "Pinned body.tex, Section 6 / Figure propa, and pinned official proportional/perforidge.py",
      "upstream_source_commit": "370fcd19199313c53da310d861a1a9fbd73b731d",
      "upstream_source_digest": "sha256:d6baf4926bb033040d389d896c3f7d0d8b124404a7e6bd2240a7883a917a0b88"
    }
  ],
  "paper_id": "G4ve69pimc",
  "release_quality_gate": {
    "algebraic_bound_substitution_counted": false,
    "exact_derivation_cells": 26,
    "direct_rate_claims": 0,
    "expected_verified_points": 10,
    "formula_only_support_counted": false,
    "independent_seeded_trials": 56,
    "judge_target": "verified_or_literal_falsification",
    "literal_falsifications": 1,
    "proxy_support_counted": false,
    "registered_claims": 5,
    "semantic_quality_gate_version": 4,
    "status": "pass",
    "supported_by_independent_evidence": 5
  },
  "upstream_pin": {
    "commit": "370fcd19199313c53da310d861a1a9fbd73b731d",
    "digest": "sha256:d6baf4926bb033040d389d896c3f7d0d8b124404a7e6bd2240a7883a917a0b88"
  }
}