docs(v3): Fisher-z analytic codec ruled THE canonical cosine/centroid replacement; helix : Fisher-2z :: cosine-replacement : Fisher-z#704
Conversation
…cal cosine/centroid replacement; helix : Fisher-2z :: cosine-replacement : Fisher-z EPIPHANIES E-FISHERZ-CANONICAL-COSINE-REPLACEMENT-1 prepended + le-contract §3 canonical-note subsection (append-only): - Ruling 1: the canonical cosine/centroid replacement = the analytic Fisher-z codec (FamilyGamma 8-byte codebook + normalized i8; certified rho>=0.999, E-PALETTE-NNUE-COSINE-GREEN-1). A materialized k x k table is a CACHE of the formula, never the canon: ranking = raw-i8 monotone compares; distance = |delta-i8| x (z_range/254); hydration = affine + tanh. i8 SIMD applies at 4x f32 density with F16/F32-equivalent RANK fidelity (measured; reconstruction-grade beyond certified corpora stays [H]). Boundary: replaces the distance/rank READ, not the semiring COMPOSE. - Ruling 2: helix is to Fisher-2z what the cosine-replacement is to Fisher-z — grounded in the tested identity hyperbolic_depth = 2 x fisher_z (helix fisher_z.rs:110-121) and the residue pipeline's batch_fisher_z; Signed360 = the hemisphere-doubled register. 1z = similarity scale, 2z = hyperbolic location/depth scale; same analytic-codec pattern, zero materialization at both rungs. - Queued follow-up: extract the FamilyGamma contract into lance-graph-contract (BBB-safe); i8 kernels stay ndarray-side. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01K3RyLEbuNSHxxB3NTTrGki
Bugbot couldn't run - usage limit reachedBugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit. A user or team admin can review and increase usage limits in the Cursor dashboard. (requestId: serverGenReqId_d5737c94-1775-49cd-ba0b-9c154fea284c) |
|
Caution Review failedThe pull request is closed. ℹ️ Recent review info⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (2)
📝 WalkthroughWalkthroughDocuments establish an analytic Fisher-z codec as the canonical cosine/centroid replacement, with raw i8 ranking and distance reads, optional materialized caches, defined composition boundaries, and a helix/Fisher-2z relationship. ChangesFisher-z replacement contract
Estimated code review effort: 1 (Trivial) | ~3 minutes Possibly related PRs
Suggested reviewers: Poem
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
Summary
Operator ruling landed on the V3 substrate (EPIPHANIES
E-FISHERZ-CANONICAL-COSINE-REPLACEMENT-1+ an append-only le-contract §3 subsection):bgz-tensor::fisher_z::{FamilyGamma, FisherZTable}— cosine →atanh→ 8-byte per-family affine → normalized i8; hydrate viatanh; certified ρ≥0.999, E-PALETTE-NNUE-COSINE-GREEN-1). A materialized k×k table is a cache of the formula, never the canon: ranking compares raw i8 (monotone), distance is|Δi8|·(z_range/254), an L4 ClassView may declare an analytic codebook with zero table materialization. i8 SIMD synergy at 4× f32 density; F16/F32-equivalent rank fidelity is the measured claim (reconstruction-grade beyond certified corpora stays [H]). Honest boundary preserved: this replaces the distance/rank READ, not the semiring COMPOSE.hyperbolic_depth = 2·fisher_zat 1e-15 (crates/helix/src/fisher_z.rs:110-121) and names the Fisher-Z arctanh the residue pipeline's one transcendental (simd.rs,batch_fisher_z);Signed360is the hemisphere-doubled register. 1z carries similarity, 2z carries hyperbolic location/depth — the same analytic-codec pattern at both rungs.FamilyGammacontract intolance-graph-contract(BBB-safe, 2×f32 + two formulas) so L4 ClassViews, ndarray, thinking-engine lenses, and stockfish-rs speak the canon without the bgz-tensor dep; i8 SIMD kernels stay ndarray-side.Doc-only, two files, pure prepend/append; no layout or code changed.
🤖 Generated with Claude Code
https://claude.ai/code/session_01K3RyLEbuNSHxxB3NTTrGki
Generated by Claude Code
Summary by CodeRabbit