D-RPF-CONF-0: measure the confidence revision surface before any policy change by AdaWorldAPI · Pull Request #1424 · AdaWorldAPI/lance-graph

AdaWorldAPI · 2026-10-09T03:59:56Z

Tests and a board entry only. This PR changes no production semantics, chooses no c=255 policy, unifies no quantizer and adds no contract declaration.

crates/causal-edge/tests/d_rpf_conf_0.rs covers T1, the singular cell, T3, R1 and NarsTables invariance.
crates/lance-graph-planner/tests/d_rpf_conf_0_r3.rs builds R3, a disagreement matrix over the distinct live revision laws.
.claude/board/entries/2026-10-09-d-rpf-conf-0-confidence-surface.md has the full report.

Findings

T1. Rust's confidence surface matches the exact integer form on 65,532 of 65,535 ordinary cells. It is one code off at (45,45), (146,148) and (148,146), all from f32 rounding at or near a .5 boundary. The earlier numpy prediction was right on only one of those three cells. Monotonicity and symmetry are clean.
Singular cell. (255,255) produces (0,0) because 0/0 evaluates to NaN, which encodes as 0. It is kept separate from the ordinary domain and recorded, not endorsed.
T3. Revising against a zero-weight partner returns the input exactly whenever c > 0. The only drift is at c = 0 with f ≠ 128, where the result is (128,0).
R1. Five c=255 candidates are measured side by side; none is chosen.
- The status quo is exact on the one-sided boundary.
- Every cap or saturate variant changes between 101,900 and 194,671 cells, by up to 128 codes.
- "Singular only" removes all NaN and changes exactly one cell.
G4 census. There are 24 revise/revision definitions. Nine are homonyms and three are delegations, which leaves the live laws A, B(L), C, D, E, F, G, J and LIN.
- On the replay/counterfactual path, replay_step overwrites forward's truth with NarsTables::build(1).
- That table is confidence-inert: c is always 170 and f is the mean.
R3. The matrix covers seven laws × 1,638,400 cells. Pairs that never differ by more than one code in the broad domain:
- A and D are byte-identical.
- C differs from them by at most one code.
- E differs only through its 0.99 cap.
- B1, B16 and LIN are genuinely different. B1, the law the replay path emits, differs from A on 99.9% of cells, by up to 168 confidence codes.

Checks

Six disable runs, all red as required. One test was vacuous because it read its own pinned constant; it was fixed and re-verified.
fmt and clippy are clean.
The entries index is current and the supersession index is current.

Open

These are left for the operator:

G1: the (255,255) cell.
G2: whether replay should emit B1.
T5 stays blocked.
J needs a declared rule for turning c into a weight before it can be compared.

Do not merge automatically.

🤖 Generated with Claude Code

https://claude.ai/code/session_01R2EuNHrariL6B3G3agyASK

Generated by Claude Code

…e (T1, T3, R1, NarsTables) Measurement only; no production change, no policy chosen. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01R2EuNHrariL6B3G3agyASK

…ngularOnly + one-sided f-side columns Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01R2EuNHrariL6B3G3agyASK

…laws Seven laws from the G4 census (A ISA, B1/B16 NarsTables, C ndarray evidence split, D planner TruthValue, E contract NarsTruth, LIN revise_truth), each called through its real entry point and compared on f in {0,51,128,204,255}^2 by the full 256x256 confidence grid. Pins distinct output confidence per law, the 21-pair matrix (cells differing, broad cells, broad max dF/dC), the set of cosmetic pairs (broad max <= 1 code: A-C, A-D, C-D), and an anti-vacuity test. Tests only; no production change. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01R2EuNHrariL6B3G3agyASK

The first version filtered the MATRIX constant, so it passed under a disable that broke law D. It now filters the measured matrix. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01R2EuNHrariL6B3G3agyASK

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01R2EuNHrariL6B3G3agyASK

coderabbitai · 2026-10-09T04:00:05Z

Note

Currently processing new changes in this PR. This may take a few minutes, please wait...

⚙️ Run configuration

Configuration used: Organization UI
Review profile: CHILL
Plan: Essentials
Run ID: ffaa4b91-db71-40cd-8e5d-ead6aac5f189

📥 Commits

Reviewing files that changed from the base of the PR and between 4eb2734 and f185b31.

📒 Files selected for processing (4)

.claude/board/entries/2026-10-09-d-rpf-conf-0-confidence-surface.md
.claude/board/entries/README.md
crates/causal-edge/tests/d_rpf_conf_0.rs
crates/lance-graph-planner/tests/d_rpf_conf_0_r3.rs

 ________________________________________________________________
< Fully armed and operationally intelligent code reviewer bunny. >
 ----------------------------------------------------------------
  \
   \   \
        \ /\
        ( )
      .( o ).

✨ Finishing Touches

📝 Generate docstrings

Commit to this branch
Create a new PR

Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Warning

Billing warning: we have not been able to collect payment for this subscription for more than 72 hours. Please update the payment method or pay any pending invoices in Billing to avoid service interruption.

_{Comment @coderabbitai help to get the list of available commands.}

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: f185b31785

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-10-09T05:31:40Z

+    let got = edge(200, 200).forward(CausalEdge64(0), &tab, &tab, &tab);
+    assert_eq!(got, Err(IsaFault::Unsupported { mantissa: 0 }));


Pin all eight operand-zero forward cases

The parent D-RPF-CONF task defines T3 as both the revision identity grid and an eight-cell forward operand-zero table, but this test exercises only one all-zero word, which immediately fails because its mantissa is unsupported. None of the zero-weight behavior under legal forward opcodes is pinned, so regressions in those seven remaining cells can pass while the report presents T3 as covered.

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-10-09T05:31:40Z

+    // `ws` never reads f, so c(c1, c2) must be the same for every frequency
+    // pair. A probe that silently depended on f = 128 would hide a defect.
+    for &(f1, f2) in &[(0u8, 0u8), (255, 255), (0, 255), (77, 200)] {


Exercise the frequency pairs claimed as exhaustive

This loop checks only four selected (f1, f2) pairs, despite the test comment saying every pair and the board report claiming the frequency-independence result was exhaustively tested. A future confidence calculation that accidentally depends on any other frequency pair would therefore pass this guard; iterate the full frequency domain or narrow the reported conclusion to the sampled pairs.

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-10-09T05:31:40Z

+So M = 7 laws compared, which collapses to 4 numerical behaviours on the u8
+grid away from the boundary: {A, C, D} (E is this one plus a cap), B16, B1, LIN.


Count E as a separate broad behavior

The report's own matrix shows E differing from A on 13,327 broad cells with maxima of 15 frequency codes and 2 confidence codes, exceeding the stated cosmetic threshold. Collapsing {A,C,D} leaves E, B16, B1, and LIN as four additional behaviors, so the measured result is five broad numerical behaviors rather than four; the current summary understates the alternatives presented for the G2 decision.

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-10-09T05:31:45Z

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-09T05:31:43.692188Z	`f185b31`	Draft marked ready

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

… conditions Also corrects #1424's claim that the replay path emits B1 in production: no production code constructs a CutContext or reads NarsEngine::tables. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01R2EuNHrariL6B3G3agyASK

claude added 7 commits October 9, 2026 03:50

test(causal-edge): D-RPF-CONF-0 fmt, drop unsafe table closure, R1 Si…

8d716f9

…ngularOnly + one-sided f-side columns Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01R2EuNHrariL6B3G3agyASK

D-RPF-CONF-0: board entry with the T1/singular/T3/R1/G4/R3 report

6fa05ad

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01R2EuNHrariL6B3G3agyASK

Regenerate board entries index

b22b8fb

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01R2EuNHrariL6B3G3agyASK

Merge remote-tracking branch 'origin/main' into ccr-bd5a6e03-niaink

f185b31

AdaWorldAPI marked this pull request as ready for review October 9, 2026 05:29

AdaWorldAPI merged commit d6f9fef into main Oct 9, 2026
9 of 10 checks passed

chatgpt-codex-connector Bot reviewed Oct 9, 2026

View reviewed changes

This was referenced Oct 9, 2026

D-RPF-TABLE-1: does the confidence-inert NarsTables fast path earn its semantic cost? #1426

Merged

D-SELF-CALIBRATING-LAB-0 phases I–IV: G2 law-A boundary, P5 Beta correspondence, P1 certified argmax, lab v0 (Gomoku) #1440

Merged

Navigation Menu

Repository navigation

D-RPF-CONF-0: measure the confidence revision surface before any policy change - #1424

D-RPF-CONF-0: measure the confidence revision surface before any policy change#1424
AdaWorldAPI merged 7 commits into
mainfrom
ccr-bd5a6e03-niaink

AdaWorldAPI commented Oct 9, 2026

Uh oh!

coderabbitai Bot commented Oct 9, 2026 •

edited

Loading

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Oct 9, 2026

Uh oh!

chatgpt-codex-connector Bot Oct 9, 2026

Uh oh!

chatgpt-codex-connector Bot Oct 9, 2026

Uh oh!

chatgpt-codex-connector Bot commented Oct 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		let got = edge(200, 200).forward(CausalEdge64(0), &tab, &tab, &tab);
		assert_eq!(got, Err(IsaFault::Unsupported { mantissa: 0 }));

		So M = 7 laws compared, which collapses to 4 numerical behaviours on the u8
		grid away from the boundary: {A, C, D} (E is this one plus a cap), B16, B1, LIN.

Conversation

AdaWorldAPI commented Oct 9, 2026

Findings

Checks

Open

Uh oh!

coderabbitai Bot commented Oct 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Oct 9, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Oct 9, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Oct 9, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot commented Oct 9, 2026

Codex Review Summary

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

coderabbitai Bot commented Oct 9, 2026 •

edited

Loading