Edition README
RRCT
README.md
1 # RRCT v0.60 Persistent Fuzzy-to-J Workspace Bridge
2
3 RRCT `0.60.0.dev1` turns the proposed fuzzy-workspace/J-space connection into a bounded, falsifiable experiment.
4
5 The new family is:
6
7 ```toml
8 experiment.family = "persistent_fuzzy_j_workspace"
9 ```
10
11 It tests a serial causal hypothesis:
12
13 ```text
14 learned fuzzy membership
15 → moving Jacobian-discovered workspace
16 → sparse nonnegative accessible content
17 → causal broadcast
18 → provenance-conditioned planning
19 → later consequence and recurrence
20 ```
21
22 The package preserves the v0.59 reliability control plane, adds stricter resume and sweep-integrity checks, and retains the canonical RRCT/Komorebi release archive.
23
24 ## What is new
25
26 - evaluator-blind synthetic controller inputs; hidden event labels and terminal targets are withheld until post-commit scoring/feedback
27 - task-independent event schedules to prevent event-type leakage
28 - learned fuzzy gates optimized from terminal consumer loss rather than membership labels
29 - a public NumPy Jacobian-frame API with exact JSON resume
30 - moving J-lens estimation, fixed-rank projector discovery, and sparse nonnegative codes
31 - same-rank PCA and seeded random-subspace controls
32 - measured membership shuffle, J-space cut, broadcast cut, provenance cut, planner cut, and feedback cut
33 - longest-supported-prefix reporting, so a late link cannot hide an earlier failed link
34 - an optional local-only Hugging Face adapter for residual capture, target-token VJPs, norm-matched patching, and projector ablation
35 - stricter completed-run and sweep-resume digest validation
36
37 ## Package scope
38
39 - **308 configuration keys**
40 - **39 major feature switches**
41 - **33 intervention controls**
42 - **9 scientific/engineering families plus one all-family smoke matrix**
43 - **15 presets and 8 sweep templates**
44 - **14 canonical prior release archives**
45 - Windows PowerShell and Bash launchers
46 - no required network access
47
48 This is an engineering and bounded functional-research platform. Its metrics are not probabilities of consciousness, subjective experience, sentience, or moral status.
49
50 ## Fast start
51
52 Run the new bounded synthetic bridge:
53
54 ```powershell
55 .\RUN_FUZZY_JSPACE.ps1
56 ```
57
58 Run a shorter diagnostic:
59
60 ```powershell
61 .\RUN_FUZZY_JSPACE.ps1 -Set "jspace.steps=48","jspace.calibration_steps=12","run.name=fuzzy_j_local_trial"
62 ```
63
64 Verify the package, run tests, and execute the all-family smoke matrix:
65
66 ```powershell
67 .\RUN_V060_LOCAL_CHECKS.ps1
68 ```
69
70 ## Causal protocol
71
72 The synthetic bridge separates controller-visible task cues from evaluator-only event truth and terminal targets. A terminal outcome is revealed only after the controller commits its prediction and plan. Evaluation reports every link separately:
73
74 1. formation: learned fuzzy membership predicts useful access
75 2. geometry: the moving Jacobian frame captures task-relevant sensitivity
76 3. access: sparse content remains decodable and beats matched controls
77 4. broadcast: intervention changes downstream report/inference outputs while preserving upstream code
78 5. planning: provenance and workspace state affect committed actions
79 6. recurrence: later access changes as a consequence of earlier outcomes
80
81 The result also reports the longest consecutive prefix that passes its declared diagnostic gates. Those gates are smoke diagnostics unless a freeze and replicated holdout cohort are run.
82
83 The bundled single-seed preset presently passes the first three diagnostic links—formation, Jacobian-space discrimination, and selective broadcast—and fails planner necessity and recurrence specificity. This negative boundary is recorded in `VALIDATION_REPORT.md`; it is not smoothed into a closure score.
84
85 See `docs/FUZZY_JSPACE_CAUSAL_PROTOCOL.md` for endpoints, controls, and falsifying outcomes.
86
87 ## Resumable sweeps
88
89 Rank/sparsity frontier:
90
91 ```powershell
92 .\RUN_SWEEP.ps1 -Config .\sweeps\07_jspace_rank_sparsity_frontier.toml
93 ```
94
95 Causal intervention ladder:
96
97 ```powershell
98 .\RUN_SWEEP.ps1 -Config .\sweeps\08_jspace_intervention_ladder.toml
99 ```
100
101 Sweep cell identities now bind the resolved base configuration. A ledger from a different base configuration is rejected instead of silently reused.
102
103 ## Optional local transformer probe
104
105 The adapter accepts only an existing local model path. It forces `local_files_only=True` and `trust_remote_code=False`; it does not train, download, or call a network API.
106
107 ```powershell
108 .\RUN_LOCAL_JSPACE_PROBE.ps1 `
109 -Model C:\models\small-local-causal-lm `
110 -Prompts .\my_prompts.jsonl `
111 -Output .\runs\local_jspace_probe.json `
112 -TargetTokenId 123
113 ```
114
115 PyTorch, Transformers, and compatible local weights are optional and are not bundled. The adapter is a measurement path, not evidence that the synthetic bridge has already been found inside a pretrained transformer.
116
117 ## Reliability commands
118
119 ```powershell
120 .\CHECK_CONTRACT.ps1 -Config .\configs\14_persistent_fuzzy_j_workspace.toml
121 .\RESUME_RUN.ps1 -RunDirectory .\runs\persistent_fuzzy_j_workspace_seed600101
122 .\AUDIT_RUN.ps1 -RunDirectory .\runs\persistent_fuzzy_j_workspace_seed600101
123 .\FREEZE_CONFIG.ps1 -Config .\configs\14_persistent_fuzzy_j_workspace.toml -Output .\validation\fuzzy_j_freeze.json
124 .\VERIFY_FREEZE.ps1 -Freeze .\validation\fuzzy_j_freeze.json
125 .\RUN_FROZEN.ps1 -Freeze .\validation\fuzzy_j_freeze.json
126 ```
127
128 Checkpoints bind the stable configuration digest and complete serialized experiment state. A completed run is skipped only after its resolved configuration, summary, run state, and checkpoint digests are all verified.
129
130 ## Scientific correction to v0.59
131
132 An audit found that the legacy `continuous_open_set_runtime` repair path directly copied evaluator-side latent state into its estimates. It remains available as an oracle-assisted engineering scaffold and now emits that limitation explicitly. Its attribution and repair numbers must not be cited as evaluator-blind evidence.
133
134 The new `persistent_fuzzy_j_workspace` family was built specifically to remove that shortcut. Evaluator truth is retained only for post-commit scoring.
135
136 ## Safety and interpretation boundary
137
138 This offline build hard-blocks network access, external-tool autonomy, autonomous persistence, self-preservation reward, valence and negative welfare-like states, and unrestricted file writing.
139
140 The new experiment can support or falsify a synthetic causal workspace chain. It does not establish a natural transformer J-space, open-world awareness, consciousness, sentience, or moral status.