[qmc] frustration-free: authenticated long-range percolation pilot - #278

Open

jiangweiqi001 wants to merge 102 commits into

QuantumBFS:mainfrom

jiangweiqi001:challenge/194-standalone

jiangweiqi001 commented Jul 30, 2026

Summary

Implements a pinned q=1 long-range FK/percolation model with independent exact and geometric oracles, counter-based Philox RNG, an incremental Poisson/Newman–Ziff production engine, immutable artifacts, restart support, and cluster wrappers.
Provides an authenticated correctness package with 22,755 checks and records two completed exploratory campaigns: P0 plus extension v1, totaling 192 verified trajectories at L = 2^10, 2^14, 2^18.
Publishes a Chinese Challenge 194 report and deterministic SVG figures derived from hash-pinned evidence. The current selector identifies a transition-refinement window for sigma=0.8 and a crossover-refinement window for sigma=1.1; sigma=0.9 and 1.0 remain candidates for further grid refinement under the preregistered dual-estimator rule.

Scientific scope

This PR establishes and validates the model, simulation engine, HPC workflow, provenance chain, exploratory data, and deterministic window-selection analysis. It does not claim final eta, nu, finite-size scaling, confirmatory P1, or extension-v2 results.

Verification

Authenticated 120-cell correctness package: 22,755 checks
Full solution test suite on the source worktree: 835 passed
Report generator focused tests: 6 passed
Standalone branch matches the committed Challenge 194 source tree, excluding unrelated scratch-only records
git diff --check passes
Report SVG generation is byte-deterministic and source-hash checked

Team and challenge

Team: frustration-free — 蒋玮琪 (jiangweiqi001), 陈硕 (ChS-YHWH), 马追景 (desitterf)
Track: QMC
Addresses [challenge]: Critical scaling in one-dimensional long-range random-cluster models #194
This is a standalone submission and intentionally does not modify or merge into registration PR [ed+mps] frustration-free: interacting pumps and finite-temperature impurity solvers #152.

Made with Cursor

Codex and others added 30 commits

July 30, 2026 19:44


          Define Challenge 194 scientific protocol

78d0a30

Pin the long-range percolation model, validation gates, scalable sampling strategy, and preregistered inference rules so later numerical claims remain reproducible and uncertainty-controlled.

Co-authored-by: Cursor <cursoragent@cursor.com>


          Plan Challenge 194 Day-0 validation

03beb6c

Break the approved protocol into test-driven model, kernel, graph-oracle, exact-enumeration, and accelerated-sampler milestones with an independent acceptance gate.

Co-authored-by: Cursor <cursoragent@cursor.com>


          Add pinned Challenge 194 model

b526085

Co-authored-by: Cursor <cursoragent@cursor.com>


          Implement periodic long-range kernel

629f990

Co-authored-by: Cursor <cursoragent@cursor.com>


          Add deterministic percolation components

5552f82

Co-authored-by: Cursor <cursoragent@cursor.com>


          Reject non-integer GraphSample arrays and expand tests

3033aae

Validate edges and labels are safe integer dtypes before int64 copy,
and add focused coverage for mutation safety, validation edges, and exports.

Co-authored-by: Cursor <cursoragent@cursor.com>


          Add independent quadratic graph oracle

895c3ec

Co-authored-by: Cursor <cursoragent@cursor.com>


          Add exact small-graph product oracle

b9fabac

Co-authored-by: Cursor <cursoragent@cursor.com>


          Fix enumeration log weights for saturated edges

c773df4

Use kappa*J rate-based closed (-rate) and open (log(-expm1(-rate)))
log weights instead of log1p(-p) on rounded probabilities.

Co-authored-by: Cursor <cursoragent@cursor.com>


          Fix tiny-kappa open-edge log weight underflow

284caad

Compute log_rate = log(kappa)+log(J) and use a stable helper for
log(1-exp(-exp(log_rate))) so saturated and underflowed rates both work.

Co-authored-by: Cursor <cursoragent@cursor.com>


          Fix huge-kappa overflow in enumeration log weights

fd226d2

Branch on log_rate before exponentiation with documented saturation,
underflow, and overflow thresholds; map -inf log mass to probability 0.

Co-authored-by: Cursor <cursoragent@cursor.com>


          Add geometric distance-class sampler

e9be31e

Sample each distance class with geometric skipping so extreme kappa regimes avoid pairwise enumeration while preserving canonical edge coverage.

Co-authored-by: Cursor <cursoragent@cursor.com>


          Strengthen geometric skipping tests

6b5ecc1

Add helper-level deterministic and statistical checks so the geometric sampler evidence catches boundary and marginal-distribution regressions.

Co-authored-by: Cursor <cursoragent@cursor.com>


          Validate Challenge 194 graph generation

384357a

Co-authored-by: Cursor <cursoragent@cursor.com>


          Tighten Challenge 194 acceptance checks

fbbe65b

Co-authored-by: Cursor <cursoragent@cursor.com>


          fix: reject subnormal sigma boundary in Challenge 194

937792c

Protect Day-0 kernel validation from sigma values that round the exponent back to one, so the model fails safely instead of reaching Hurwitz-zeta poles and NaNs. Add the final targeted regressions for geometry validation, kernel conventions, oracle exports, and the requested fix report.

Co-authored-by: Cursor <cursoragent@cursor.com>


          Design Challenge 194 production workflow

745336d

Freeze the Numba-first event sampler, restartable artifacts, validation and performance gates, pilot protocol, cluster policy, and confirmatory analysis boundaries.

Co-authored-by: Cursor <cursoragent@cursor.com>


          Plan Challenge 194 production engine

63ded55

Co-authored-by: Cursor <cursoragent@cursor.com>


          Pin production Numba runtime

c18923a


          Harden runtime provenance validation

7bd7aeb


          Add deterministic Philox counter streams

fdb5106


          Add deterministic distance alias table

d728f9e

Co-authored-by: Cursor <cursoragent@cursor.com>


          Fix high-sigma alias validation

Co-authored-by: Cursor <cursoragent@cursor.com>


          Add deterministic uint64 edge set

16e6002


          Harden uint64 edge set boundaries

f0426c8


          Add incremental production observables

48ece81


          Add independent monotone Poisson reference

a40f530

Co-authored-by: Cursor <cursoragent@cursor.com>


          Harden Poisson reference numerics and laws

515eeeb

Co-authored-by: Cursor <cursoragent@cursor.com>


          Fix terminal Poisson event inclusion

8ed558b

Co-authored-by: Cursor <cursoragent@cursor.com>


          Add Numba Poisson sweep engine

Co-authored-by: Cursor <cursoragent@cursor.com>

codex and others added 30 commits

July 30, 2026 19:44


          Publish deterministic P1 refinement protocol

0b4726c

Co-authored-by: Cursor <cursoragent@cursor.com>


          Document P0 analysis and P1 protocol workflow

46f5f21

Co-authored-by: Cursor <cursoragent@cursor.com>


          Design targeted P0 extension

b1078b7

Co-authored-by: Cursor <cursoragent@cursor.com>


          Plan versioned P0 extension implementation

f552b11

Co-authored-by: Cursor <cursoragent@cursor.com>


          Add versioned P0 extension protocol

19ae0fe

Co-authored-by: Cursor <cursoragent@cursor.com>


          Bind extension protocol to frozen P0 evidence

5731e5c

Co-authored-by: Cursor <cursoragent@cursor.com>


          Publish immutable P0 extension protocol

77aac9c

Co-authored-by: Cursor <cursoragent@cursor.com>


          Add authenticated P0 extension runtime

f5e5cd5

Co-authored-by: Cursor <cursoragent@cursor.com>


          Harden P0 extension runtime evidence

bdab704

Co-authored-by: Cursor <cursoragent@cursor.com>


          Prepare P0 extension cluster campaign

35a3b52


          Harden P0 extension Slurm dispatch

ed187a4


          Correct P0 extension deployment contract

64e7786


          Harden P0 extension bundle publication

bc888b8

Co-authored-by: Cursor <cursoragent@cursor.com>


          Preserve P0 extension venv launchers

e798587

Co-authored-by: Cursor <cursoragent@cursor.com>


          Require explicit P0 extension evidence root

341be9a

Co-authored-by: Cursor <cursoragent@cursor.com>


          Add bounded P0 extension aggregation

8747d30

Co-authored-by: Cursor <cursoragent@cursor.com>


          Combine verified P0 extension evidence

013676e

Co-authored-by: Cursor <cursoragent@cursor.com>


          Require canonical combined evidence integers

d4148da

Co-authored-by: Cursor <cursoragent@cursor.com>


          Rerun frozen selector on combined evidence

2e20958

Co-authored-by: Cursor <cursoragent@cursor.com>


          Require source validation for combined selector

ec07fe8

Co-authored-by: Cursor <cursoragent@cursor.com>


          Document P0 extension evidence workflow

5d655f1


          Format P0 extension plan files

ba78122


          Record P0 extension boundary evidence

ee480d1

Co-authored-by: Cursor <cursoragent@cursor.com>


          Authenticate combined P0 extension evidence

740b31d

Require exact deeply verified P0 and extension roots plus the immutable protocol before combined evidence can authorize selection or P1.

Co-authored-by: Cursor <cursoragent@cursor.com>


          Close extension evidence ABA window

cf61476

Bind retained-snapshot aggregation back to the pinned extension run and progress hashes before combined evidence can reach selection or P1.

Co-authored-by: Cursor <cursoragent@cursor.com>


          Preregister standalone coarse-grid P0 extension v2

0e9b112

Co-authored-by: Cursor <cursoragent@cursor.com>


          Plan standalone P0 extension v2 execution

3dbba52

Co-authored-by: Cursor <cursoragent@cursor.com>


          Document Challenge 194 exploratory evidence

591439f

Present the authenticated pilot results, reproducible figures, and scientific boundaries without overstating incomplete scaling work.

Co-authored-by: Cursor <cursoragent@cursor.com>


          Clean standalone Challenge 194 submission

b0c4eeb

Remove trailing whitespace so the isolated submission passes repository diff checks.

Co-authored-by: Cursor <cursoragent@cursor.com>


          Reframe Challenge 194 around verified成果

06d639a

Lead with the authenticated engine, cluster trajectories, and reusable evidence chain while keeping unfinished scaling claims explicitly bounded.

Co-authored-by: Cursor <cursoragent@cursor.com>

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet