Release v1.1.1 token-accounting corrections

Include observed subagent tokens in derived reports, regenerate the reproducibility index, and refresh versioned Croissant metadata. Keep raw records, canonical tables, artifacts, labels, and v1.1.0 unchanged.

CHANGELOG.md

@@ -1,5 +1,12 @@
 # Changelog
 
+## v1.1.1: token-accounting and readability corrections (2026-11-07)
+
+- Include subagent requests in observed vendor-token totals and the derived cost
+  summaries. No routine finder meets the unchanged 150,000 mean-token ceiling;
+  the prior-stress audit finder remains MiMo 2.5 Pro low.
+- Improve figure and table readability with larger text and compact heatmap labels.
+
 ## v1.1.0: camera-ready corrections (2026-10-16)
 
 v1.0.0 is the release as submitted to the NeurIPS 2026 Evaluations & Datasets

README.md

@@ -72,7 +72,7 @@ the paper.
 - Paper: https://openreview.net/forum?id=y48mj8HiWu
 - Code: https://github.com/JacobLinCool/Build-and-Find
 - DOI: [10.57967/hf/11347](https://doi.org/10.57967/hf/11347)
-- Version: 1.1.0 (camera-ready), revision `v1.1.0` of this dataset; changes
+- Version: 1.1.1 (camera-ready), revision `v1.1.1` of this dataset; changes
   from the reviewed v1.0.0 are listed in `CHANGELOG.md`.
 
 ## Dataset structure
@@ -85,7 +85,7 @@ data directories into a checkout of the code, from its root:
 ```bash
 HF_DATASET_URL="https://huggingface.co/datasets/JacobLinCool/Build-and-Find"
 uvx --from huggingface_hub hf download "${HF_DATASET_URL#https://huggingface.co/datasets/}" \
-  --repo-type dataset --revision v1.1.0 \
+  --repo-type dataset --revision v1.1.1 \
   --include "benchmarks/**" --include "runs_release/**" --include "rebuttal/**" \
   --local-dir .
 ```
@@ -93,7 +93,7 @@ uvx --from huggingface_hub hf download "${HF_DATASET_URL#https://huggingface.co/
 Downloading the whole dataset into a checkout would replace the code
 repository's `README.md` with this card. `croissant.json` is the same file as
 `metadata/croissant.json` in the code repository: it gives each file as its URL
-at revision `v1.1.0` and each file set as paths in this repository.
+at revision `v1.1.1` and each file set as paths in this repository.
 
 | path | contents | license |
 |---|---|---|
@@ -107,7 +107,7 @@ at revision `v1.1.0` and each file set as paths in this repository.
 | `rebuttal/conformance_spot_audit/` | post-review spot audit of 24 consensus prelabels | CC BY 4.0 |
 | `croissant.json` | Croissant 1.1 metadata with Responsible AI fields (`metadata/croissant.json` in the code repository) | CC BY 4.0 |
 | `LICENSE-DATA`, `LICENSE-CODE` | license notices: CC BY 4.0 for the data, Apache-2.0 for the generated codebases | |
-| `CHANGELOG.md` | changes from v1.0.0 to v1.1.0 | CC BY 4.0 |
+| `CHANGELOG.md` | changes from v1.0.0 to v1.1.1 | CC BY 4.0 |
 
 The v1.1.0 report layers are `runs_release/reports/eandd_final/control_exclusions/`,
 `runs_release/reports/eandd_final/transcript_audit/`,
@@ -229,6 +229,14 @@ No crowdsourced or human-subject annotation was used. The raw agent sessions of
 the audits are not included; the audit tables keep each session's id and result
 hash.
 
+## Corrections in v1.1.1
+
+v1.1.1 includes subagent requests in observed vendor-token totals and the derived
+cost summaries. No routine finder meets the unchanged 150,000 mean-token ceiling;
+the prior-stress audit finder remains MiMo 2.5 Pro low. The released records,
+canonical tables, artifacts, and labels are unchanged, and v1.1.0 remains
+available at its original revision.
+
 ## Corrections in v1.1.0
 
 v1.1.0 adds files and changes analysis outputs; the v1.0.0 records, tables,

croissant.json

@@ -66,15 +66,15 @@
   "@id": "build_and_find_scratch_task_pack_v1",
   "name": "Build-and-Find",
   "alternateName": "Build-and-Find: An Effort-Aware Protocol for Evaluating Agent-Managed Codebases",
-  "description": "Release data of Build-and-Find, a protocol and benchmark for evaluating whether downstream coding agents can recover the intended behavior and design choices of a repository written by another agent, and how much inspection the recovery requires. A builder agent receives a hidden specification and writes a codebase; a finder agent is given only the codebase and a specification-traced multiple-choice question bank. The release contains two Rust build-from-scratch task specifications with their question banks, the 48 generated build artifacts, 1,924 sanitized run records (48 builds, 1,728 artifact-conditioned find runs, 144 question-only and spec-only control find runs, and 4 control build records), canonical CSV and JSON tables with a manifest, derived analysis reports, the v1.1.0 post-review layers (question-only control exclusions, transcript audit, implementation-label corrections, option-free check, builder-written test-suite runs, and per-task R_b), and the post-review implementation-label audit records with the author and model-assisted decisions. Each FileObject contentUrl points to the file in the Hugging Face dataset at revision v1.1.0, and each FileSet is contained in the git repository of the dataset, with includes relative to the dataset root. The dataset root has the layout of the code repository root; this file is stored there as croissant.json and in the code repository as metadata/croissant.json.",
+  "description": "Release data of Build-and-Find, a protocol and benchmark for evaluating whether downstream coding agents can recover the intended behavior and design choices of a repository written by another agent, and how much inspection the recovery requires. A builder agent receives a hidden specification and writes a codebase; a finder agent is given only the codebase and a specification-traced multiple-choice question bank. The release contains two Rust build-from-scratch task specifications with their question banks, the 48 generated build artifacts, 1,924 sanitized run records (48 builds, 1,728 artifact-conditioned find runs, 144 question-only and spec-only control find runs, and 4 control build records), canonical CSV and JSON tables with a manifest, derived analysis reports, the v1.1.0 post-review layers (question-only control exclusions, transcript audit, implementation-label corrections, option-free check, builder-written test-suite runs, and per-task R_b), and the post-review implementation-label audit records with the author and model-assisted decisions. Each FileObject contentUrl points to the file in the Hugging Face dataset at revision v1.1.1, and each FileSet is contained in the git repository of the dataset, with includes relative to the dataset root. The dataset root has the layout of the code repository root; this file is stored there as croissant.json and in the code repository as metadata/croissant.json.",
   "conformsTo": "http://mlcommons.org/croissant/1.1",
   "license": "https://creativecommons.org/licenses/by/4.0/",
   "url": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find",
   "identifier": "https://doi.org/10.57967/hf/11347",
-  "version": "1.1.0",
+  "version": "1.1.1",
   "dateCreated": "2026-05-01",
   "datePublished": "2026-05-04",
-  "dateModified": "2026-10-16",
+  "dateModified": "2026-11-07",
   "creator": [
     {
       "@type": "sc:Person",
@@ -141,7 +141,7 @@
   "rai:dataCollection": "Task specifications and source-traced question banks were authored for the benchmark. Builder and finder runs were collected between 30 April and 3 May 2026 (UTC) with the Build-and-Find harness from twelve agent configurations: GPT-5.5 and GPT-5.4-mini through the Codex CLI, and Claude Opus 4.7, Claude Sonnet 4.6, MiMo-v2.5-pro, and MiMo-v2.5 through the Claude Code CLI, each at high and low reasoning effort. Each builder built each task twice (48 builds). Each finder answered the 15-question bank of each of the 48 artifacts three times (1,728 artifact-conditioned find runs), and three times per task without an artifact (question-only control) and with the specification only (spec-only control), giving 144 control find runs. Option order was shuffled per run with a recorded seed, answers were written to a file in the finder workspace, and usage was recorded as bytes exchanged with the model. Builder and finder workspaces were per-run directories inside the harness repository, and the agent CLIs ran with permission prompts disabled and without an OS-level sandbox. Post-review layers were collected from July to October 2026: a transcript check of the artifact-conditioned and control find runs; an option-free check (GPT-5.5-high and Opus 4.7-high, 24 sessions, 96 free-text answers on eight low-prior questions); one run of each artifact's builder-written test suite (cargo build --release and cargo test); and model-assisted source-trace audits of the implementation labels (rebuttal/conformance_spot_audit/ and rebuttal/full_conformance_audit/).",
   "rai:dataPreprocessingProtocol": "Append-only local run records were sanitized into runs_release/ by scripts/prepare_release.py: records were pruned to the final panel, the controls, and the build-fidelity diagnostics; local trace paths were removed and artifact paths rewritten to release-relative paths; per-run session logs were not released. Repairable malformed or nested answer files were recovered by recorded rules (malformed_answer_recovery_manifest.json and nested_answer_recovery_manifest.json under runs_release/reports/eandd_final/, with the original records kept). Canonical CSV and JSON tables under runs_release/tables are deterministic projections of the records; MANIFEST.json records row counts, sizes, hashes, schema version, and generation time. The v1.1.0 corrections leave the released records, tables, artifacts, and v1.0 labels unchanged: control exclusions and label corrections are added files that the analysis scripts apply. Release CSVs use LF line endings, and recorded checksums are computed on those bytes.",
   "rai:dataAnnotationProtocol": "Gold answers: the question banks and their gold answers were written by the author with LLM assistance and reviewed by the author; every question carries a source trace to a specification intention, which bnf specs validate checks. No crowdsourced annotation or human-subject labeling was used. Implementation labels (runs_release/reports/artifact_conformance_consensus/final/): each artifact-question pair was prelabeled when the stable answers (at least two of three trials) of the Codex GPT-5.5-high and Claude Opus 4.7-high finders agreed, as implemented_gold on the gold answer and implemented_non_gold on another answer (696 pairs); the 24 pairs without such consensus were labeled by manual review. These prelabels are triage labels, not independent conformance annotations. v1.1.0 adds one rule-based correction in label_corrections_v1_1.csv: the 15 pairs of the build without source files become absent_or_ambiguous (rule_empty_artifact). Source-trace audit (rebuttal/full_conformance_audit/): after review, a model-assisted source trace of the 592 scored compile-pass pairs (a primary pass and a cross-check pass; models, prompts, session ids, and result hashes are recorded in the audit tables) flagged 205 pairs. Every flag has a recorded decision: 43 by the author (author_decisions.csv; 41 relabeled, 2 kept) and 162 drafted by model-assisted adjudication (two passes, the second without access to the audit verdicts, with arbitration where they disagreed) and reviewed and approved by the author (model_assisted_decisions.csv; 154 relabeled, 8 kept). The adjudicated labels enter the sensitivity analysis in label_sensitivity_adjudicated/. Option-free check (runs_release/reports/option_free_check/): the author graded all 96 free-text answers against a rubric fixed before the sessions, in two frozen passes (claim correctness, then evidence sufficiency); an LLM labeled the same rows for a consistency check that enters only the agreement counts.",
-  "rai:dataReleaseMaintenancePlan": "The release is versioned. v1.0.0 is the reviewed release, kept as a tagged commit of the code repository; v1.1.0, the version of this dataset, adds the camera-ready corrections as new files and changed analysis outputs, and CHANGELOG.md lists every change. Corrections to specifications, questions, gold answers, labels, or analysis transforms are released as new versions that keep the prior records. Code is maintained in the GitHub repository and data in the Hugging Face dataset; questions and corrections can be sent through repository issues or to jacob.cs14@nycu.edu.tw.",
+  "rai:dataReleaseMaintenancePlan": "The release is versioned. v1.0.0 is the reviewed release, kept as a tagged commit of the code repository; v1.1.0 adds the camera-ready corrections as new files and changed analysis outputs. v1.1.1 includes subagent requests in token totals and updates the derived cost summaries and finder selection; the released records, canonical tables, artifacts, and labels are unchanged. CHANGELOG.md lists every change, and prior versions remain available. Corrections to specifications, questions, gold answers, labels, or analysis transforms are released as new versions that keep the prior records. Code is maintained in the GitHub repository and data in the Hugging Face dataset; questions and corrections can be sent through repository issues or to jacob.cs14@nycu.edu.tw.",
   "prov:wasDerivedFrom": [
     {
       "@id": "benchmarks/specs/scratch_minidb.yaml"
@@ -282,7 +282,7 @@
       "@id": "build_runs_csv",
       "name": "build_runs.csv",
       "description": "Canonical build-run table.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/tables/build_runs.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/tables/build_runs.csv",
       "contentSize": "30278",
       "encodingFormat": "text/csv",
       "sha256": "6ef9dc045cc4b004e591c17c862b121aeae4ccabb20ff63b225ea098e63d031c"
@@ -292,7 +292,7 @@
       "@id": "find_runs_csv",
       "name": "find_runs.csv",
       "description": "Canonical finder-run table with run-level metrics and manifest fields.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/tables/find_runs.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/tables/find_runs.csv",
       "contentSize": "1322367",
       "encodingFormat": "text/csv",
       "sha256": "5f1285bfb746a5c53da54ad57f69e7a451a670bc58b09c283fb9c31082645540"
@@ -302,7 +302,7 @@
       "@id": "question_answers_csv",
       "name": "question_answers.csv",
       "description": "Canonical per-question answer table.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/tables/question_answers.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/tables/question_answers.csv",
       "contentSize": "24554824",
       "encodingFormat": "text/csv",
       "sha256": "30ca51f75d2d87cda16c410f83eb20b89f5a028f03a2c5d8f8e33eca88fa01df"
@@ -312,7 +312,7 @@
       "@id": "panel_membership_csv",
       "name": "panel_membership.csv",
       "description": "Panel membership table linking named analysis panels to table row ids.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/tables/panel_membership.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/tables/panel_membership.csv",
       "contentSize": "18202760",
       "encodingFormat": "text/csv",
       "sha256": "eb0e5d63a33f36cb86fe4346dec14fd1577fc2af5df1e101c0a7c010d8f1d402"
@@ -322,7 +322,7 @@
       "@id": "experiments_json",
       "name": "experiments.json",
       "description": "Experiment and panel metadata for released analysis tables.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/tables/experiments.json",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/tables/experiments.json",
       "contentSize": "15526",
       "encodingFormat": "application/json",
       "sha256": "00c4fa6e46eda9a11bbac210bdda06f5d33faa87905011b5c69f0e9cd3df0b5d"
@@ -332,7 +332,7 @@
       "@id": "table_manifest_json",
       "name": "MANIFEST.json",
       "description": "Canonical table manifest with file sizes, row counts, hashes, schema version, and generation timestamp.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/tables/MANIFEST.json",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/tables/MANIFEST.json",
       "contentSize": "963",
       "encodingFormat": "application/json",
       "sha256": "f6c8edc1f4e7ee96295c7aad33f8cfa554ffa9583963bdeeab0b2c1cb62ef8a4"
@@ -342,10 +342,10 @@
       "@id": "reproducibility_index_md",
       "name": "reproducibility_index.md",
       "description": "Reproducibility index: maps the paper's main results to the scripts, tables, and reports that produce them, with the table hashes.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/reports/reproducibility_index.md",
-      "contentSize": "12173",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/reports/reproducibility_index.md",
+      "contentSize": "12174",
       "encodingFormat": "text/markdown",
-      "sha256": "b2f392c66b33e33058796edca38733c62312f8f708603323cd5aa450ebbfdfc2"
+      "sha256": "84b1e470ab9b5a142e9b6674eb1a1712c33f1bffc41a3a28977bd22783c2904f"
     },
     {
       "@type": "cr:FileSet",
@@ -377,7 +377,7 @@
       "@id": "evidence_citation_audit_md",
       "name": "evidence_citation_audit.md",
       "description": "Evidence-citation audit of a sample of correct finder answers.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/reports/evidence_citation_audit.md",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/reports/evidence_citation_audit.md",
       "contentSize": "1157",
       "encodingFormat": "text/markdown",
       "sha256": "a671d855ce09689f1d36c6c417a30cfc79df2b876c2193aea233af29ef9a5cf8"
@@ -387,7 +387,7 @@
       "@id": "evidence_citation_audit_sample_csv",
       "name": "evidence_citation_audit_sample.csv",
       "description": "Evidence-citation audit sample with the audit labels.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/reports/evidence_citation_audit_sample.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/reports/evidence_citation_audit_sample.csv",
       "contentSize": "63749",
       "encodingFormat": "text/csv",
       "sha256": "d61293e7045244733ba7faae723925b01224f7bd4c85dc564c42886de9a7d584"
@@ -397,7 +397,7 @@
       "@id": "implementation_labels_csv",
       "name": "artifact_conformance_final.csv",
       "description": "v1.0 implementation labels: one row per artifact-question pair with its final label and label source (consensus prelabel or manual review). Unchanged in v1.1.0.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/reports/artifact_conformance_consensus/final/artifact_conformance_final.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/reports/artifact_conformance_consensus/final/artifact_conformance_final.csv",
       "contentSize": "1044174",
       "encodingFormat": "text/csv",
       "sha256": "d89d92dc92b00488984cb52965c5cb1f774894fea845ef90ed9b9b8eee3a2b1e"
@@ -407,7 +407,7 @@
       "@id": "label_corrections_v1_1_csv",
       "name": "label_corrections_v1_1.csv",
       "description": "v1.1.0 implementation-label corrections, applied as an overlay on artifact_conformance_final.csv.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/reports/artifact_conformance_consensus/final/label_corrections_v1_1.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/reports/artifact_conformance_consensus/final/label_corrections_v1_1.csv",
       "contentSize": "5025",
       "encodingFormat": "text/csv",
       "sha256": "6430dd0b42919b7f8b25214ae10d20f0f5958cb8c784093f92bd084e08910b29"
@@ -417,7 +417,7 @@
       "@id": "conditioned_builder_scores_by_task_csv",
       "name": "conditioned_builder_scores_by_task.csv",
       "description": "Per-task builder R_b under the released labels with the v1.1 correction (paper Table 16).",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/reports/artifact_conformance_consensus/impact/conditioned_builder_scores_by_task.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/reports/artifact_conformance_consensus/impact/conditioned_builder_scores_by_task.csv",
       "contentSize": "5427",
       "encodingFormat": "text/csv",
       "sha256": "bfb048a4888774e55d5b4721e07af189424c75ed7d1678adba0d25690d05539c"
@@ -442,7 +442,7 @@
       "@id": "excluded_question_only_runs_csv",
       "name": "excluded_question_only_runs.csv",
       "description": "The 11 question-only control runs excluded in v1.1.0, with the exposure category and transcript evidence.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/reports/eandd_final/control_exclusions/excluded_question_only_runs.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/reports/eandd_final/control_exclusions/excluded_question_only_runs.csv",
       "contentSize": "3226",
       "encodingFormat": "text/csv",
       "sha256": "9b8566d5f3064ae12289d93697fa7ae38c3a9e5e0b789bb06977372b41e87858"
@@ -467,7 +467,7 @@
       "@id": "primary_panel_exposure_csv",
       "name": "primary_panel_exposure.csv",
       "description": "The 62 primary-panel find runs whose transcripts reach outside their workspace, with the exposure category.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/reports/eandd_final/transcript_audit/primary_panel_exposure.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/reports/eandd_final/transcript_audit/primary_panel_exposure.csv",
       "contentSize": "16511",
       "encodingFormat": "text/csv",
       "sha256": "02c7d4fbd53b7cfbc9d3049304c453a026a71c9c3ec9e9af44acfff7f61206d7"
@@ -492,7 +492,7 @@
       "@id": "option_free_grading_key_csv",
       "name": "grading_key.csv",
       "description": "Option-free check: map from grading id to session, condition, finder, trial, seed, and source artifact.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/reports/option_free_check/grading_key.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/reports/option_free_check/grading_key.csv",
       "contentSize": "17074",
       "encodingFormat": "text/csv",
       "sha256": "16701370cf99b9a7f79ff8a1fb2ad9f51cb58f08f29261394de8076ca80d39d7"
@@ -502,7 +502,7 @@
       "@id": "option_free_grading_pass2_csv",
       "name": "grading_sheet.pass2.csv",
       "description": "Option-free check: the frozen author grading (claim correctness from pass 1 and evidence sufficiency from pass 2) of the 96 free-text answers.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/reports/option_free_check/grading_sheet.pass2.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/reports/option_free_check/grading_sheet.pass2.csv",
       "contentSize": "68230",
       "encodingFormat": "text/csv",
       "sha256": "8a479b6c79bbafa0cf876af3be43698ae23062c3c5a3e48d04aacc6c7f0b53a9"
@@ -527,7 +527,7 @@
       "@id": "artifact_tests_csv",
       "name": "artifact_tests.csv",
       "description": "One run of cargo build --release and cargo test per released build artifact.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/runs_release/reports/runtime_tests/artifact_tests.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/runs_release/reports/runtime_tests/artifact_tests.csv",
       "contentSize": "6843",
       "encodingFormat": "text/csv",
       "sha256": "06c693838de9a9bad68dd378ce085ddcb455f10be3609139cf50467c40599597"
@@ -537,7 +537,7 @@
       "@id": "full_audit_readme_md",
       "name": "full_conformance_audit_README.md",
       "description": "Description of the full-coverage source-trace audit files and decision tables.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/rebuttal/full_conformance_audit/README.md",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/rebuttal/full_conformance_audit/README.md",
       "contentSize": "9474",
       "encodingFormat": "text/markdown",
       "sha256": "0a7d93ad470ec2e8d772d94771d31c894dc8ac24c4ada79ddf0e45b649f32ae5"
@@ -547,7 +547,7 @@
       "@id": "full_audit_spec_md",
       "name": "full_conformance_audit_SPEC.md",
       "description": "Protocol of the full-coverage source-trace audit.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/rebuttal/full_conformance_audit/SPEC.md",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/rebuttal/full_conformance_audit/SPEC.md",
       "contentSize": "7943",
       "encodingFormat": "text/markdown",
       "sha256": "c9f59d63534813c1dc2952bd81b31a463968d84ce0fbf0f1f417a00ee872f18e"
@@ -557,7 +557,7 @@
       "@id": "full_audit_report_md",
       "name": "full_conformance_audit_report.md",
       "description": "Summary of the full-coverage source-trace audit run.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/rebuttal/full_conformance_audit/report.md",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/rebuttal/full_conformance_audit/report.md",
       "contentSize": "8752",
       "encodingFormat": "text/markdown",
       "sha256": "3cfd238539421effffb2ffdfe3c1c3881b8ca3944f29d0d105641fea1267924e"
@@ -567,7 +567,7 @@
       "@id": "full_audit_prompt_md",
       "name": "live_path_v1.md",
       "description": "Prompt template of the source-trace audit passes.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/rebuttal/full_conformance_audit/prompts/live_path_v1.md",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/rebuttal/full_conformance_audit/prompts/live_path_v1.md",
       "contentSize": "2073",
       "encodingFormat": "text/markdown",
       "sha256": "d02a81bb42f2f3cb685c4c8e3d9712f091f3508859a8f82381a79d3c24caef87"
@@ -577,7 +577,7 @@
       "@id": "full_audit_population_manifest_csv",
       "name": "population_manifest.csv",
       "description": "The 592 scored compile-pass artifact-question pairs audited.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/rebuttal/full_conformance_audit/population_manifest.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/rebuttal/full_conformance_audit/population_manifest.csv",
       "contentSize": "176048",
       "encodingFormat": "text/csv",
       "sha256": "3ef3d5cac41f3b26b1a6ca5fe133bb8a8f07724ba29603210e13e1d06f489efd"
@@ -587,7 +587,7 @@
       "@id": "full_audit_minidb_adjudication_csv",
       "name": "full_audit_minidb_adjudication.csv",
       "description": "Primary and cross-check audit results for the scratch_minidb pairs.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/rebuttal/full_conformance_audit/minidb_adjudication.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/rebuttal/full_conformance_audit/minidb_adjudication.csv",
       "contentSize": "557671",
       "encodingFormat": "text/csv",
       "sha256": "a1c01056c196698fd3cdebe4e4a94b47d94d4e4c57e3f8982e2a6533d56023aa"
@@ -597,7 +597,7 @@
       "@id": "full_audit_nanoweb_adjudication_csv",
       "name": "full_audit_nanoweb_adjudication.csv",
       "description": "Primary and cross-check audit results for the scratch_nanoweb pairs.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/rebuttal/full_conformance_audit/nanoweb_adjudication.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/rebuttal/full_conformance_audit/nanoweb_adjudication.csv",
       "contentSize": "588251",
       "encodingFormat": "text/csv",
       "sha256": "df3abc72153d61164bc93ae4c5653cc1ad59a796d6625472069afd39783e7b23"
@@ -607,7 +607,7 @@
       "@id": "full_audit_disagreements_csv",
       "name": "full_audit_disagreements.csv",
       "description": "The 205 pairs flagged by either audit pass.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/rebuttal/full_conformance_audit/disagreements.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/rebuttal/full_conformance_audit/disagreements.csv",
       "contentSize": "495897",
       "encodingFormat": "text/csv",
       "sha256": "07832ae110093b9a55b044dd97a94bf35e8836412e33be9390141bd31ce28ac1"
@@ -617,7 +617,7 @@
       "@id": "full_audit_evidence_excerpts_jsonl",
       "name": "evidence_excerpts.jsonl",
       "description": "Source excerpts cited by the audit passes, one line per cited span, with the SHA-256 of the cited artifact file.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/rebuttal/full_conformance_audit/evidence_excerpts.jsonl",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/rebuttal/full_conformance_audit/evidence_excerpts.jsonl",
       "contentSize": "5801598",
       "encodingFormat": "application/jsonlines",
       "sha256": "532d30809cb34412dd349a6f72f05d4ec6615cf8e98fea0f99940f9d4dd943da"
@@ -627,7 +627,7 @@
       "@id": "author_decisions_csv",
       "name": "author_decisions.csv",
       "description": "The 43 author decisions on audit flags.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/rebuttal/full_conformance_audit/author_decisions.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/rebuttal/full_conformance_audit/author_decisions.csv",
       "contentSize": "7443",
       "encodingFormat": "text/csv",
       "sha256": "0d7826a49385678bf79f22daae945fa67d71ac5d85f7d21a159f3ff0a577d0a3"
@@ -637,7 +637,7 @@
       "@id": "model_assisted_decisions_csv",
       "name": "model_assisted_decisions.csv",
       "description": "The 162 decisions drafted by model-assisted adjudication and approved by the author.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/rebuttal/full_conformance_audit/model_assisted_decisions.csv",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/rebuttal/full_conformance_audit/model_assisted_decisions.csv",
       "contentSize": "199702",
       "encodingFormat": "text/csv",
       "sha256": "82092af15ef8dcd097bf45c34a333ba0ea53a192ffbaa25ab1b752cf4d4d91cc"
@@ -647,7 +647,7 @@
       "@id": "full_audit_core_generation_manifest_json",
       "name": "core_generation_manifest.json",
       "description": "Manifest of the core audit generation; make label-sensitivity checks the audit tables against its hashes.",
-      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.0/rebuttal/full_conformance_audit/generations/345e9f073af7b83f6ed35a069afe4c1dc78f9aac5c74e40b09c34b33cc428bfd/generation_manifest.json",
+      "contentUrl": "https://huggingface.co/datasets/JacobLinCool/Build-and-Find/resolve/v1.1.1/rebuttal/full_conformance_audit/generations/345e9f073af7b83f6ed35a069afe4c1dc78f9aac5c74e40b09c34b33cc428bfd/generation_manifest.json",
       "contentSize": "2639",
       "encodingFormat": "application/json",
       "sha256": "35b1834f10c68a4e9f743f0e63117bdb2041a678386831e610abf7954a59c9e4"

runs_release/reports/eandd_final/builder_summary.csv

@@ -1,13 +1,13 @@
 builder_id,family,base_model,effort,downstream_accuracy,downstream_run_ci95_low,downstream_run_ci95_high,median_build_tokens,mean_build_tokens,total_build_tokens,total_build_cost_usd,build_runs,question_rows
-claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.994444,0.991161,0.997728,5891623,6418914,25675659,31.4761,4,2160
+claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.994444,0.991161,0.997728,5963041,6565830,26263322,31.4761,4,2160
 claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.992593,0.985651,0.999534,7278835,7278835,7278835,3.5121,1,540
-claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.992593,0.989159,0.996027,4507726,4744350,18977400,21.9518,4,2160
+claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.992593,0.989159,0.996027,4588497,4820855,19283421,21.9518,4,2160
 codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.991667,0.987620,0.995713,7304000,7470920,29883683,0.0000,4,2160
 codex_gpt5_5_high,codex,codex_gpt5_5,high,0.988889,0.984618,0.993160,3185167,3216598,12866393,0.0000,4,2160
 codex_gpt5_5_low,codex,codex_gpt5_5,low,0.986574,0.981829,0.991320,1884608,1921738,7686954,0.0000,4,2160
-claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.967130,0.961517,0.972742,6900882,7709126,30836507,19.4583,4,2160
-claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.960185,0.951747,0.968624,22845676,22845676,45691352,19.8670,2,1080
-claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.949074,0.939018,0.959131,3034817,3280869,13123479,12.5481,4,2160
-claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.944444,0.930964,0.957925,11909618,13859479,41578437,18.0834,3,1620
+claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.967130,0.961517,0.972742,7065720,7867195,31468782,19.4583,4,2160
+claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.960185,0.951747,0.968624,23108807,23108807,46217615,19.8670,2,1080
+claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.949074,0.939018,0.959131,5529057,5092624,20370498,12.5481,4,2160
+claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.944444,0.930964,0.957925,12084066,13976005,41928015,18.0834,3,1620
 codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.921605,0.909111,0.934099,3043780,2536506,7609518,0.0000,3,1620
-claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.915278,0.903046,0.927509,14120147,15334640,61338562,26.1003,4,2160
+claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.915278,0.903046,0.927509,14214290,15424765,61699061,26.1003,4,2160

runs_release/reports/eandd_final/control_exclusions/finder_calibration.csv

@@ -1,13 +1,13 @@
 finder_id,formal_all_rows_correct,formal_all_rows_total,formal_all_rows_pct,question_only_runs,formal_matched_all_correct,formal_matched_all_total,formal_matched_all_pct,question_only_all_correct,question_only_all_total,question_only_all_pct,delta_all_pp,formal_matched_low_prior_correct,formal_matched_low_prior_total,formal_matched_low_prior_pct,question_only_low_prior_correct,question_only_low_prior_total,question_only_low_prior_pct,delta_low_prior_pp,formal_matched_implementation_signal_correct,formal_matched_implementation_signal_total,formal_matched_implementation_signal_pct,question_only_implementation_signal_correct,question_only_implementation_signal_total,question_only_implementation_signal_pct,delta_implementation_signal_pp,mean_tokens_k
-claude_code_sonnet_4_6_low,1821,1845,98.7,6,1821,1845,98.7,88,90,97.8,+0.9,595,615,96.7,28,30,93.3,+3.4,1219,1233,98.9,58,60,96.7,+2.2,116
-claude_code_sonnet_4_6_high,1809,1845,98.0,6,1809,1845,98.0,86,90,95.6,+2.5,596,615,96.9,26,30,86.7,+10.2,1220,1233,98.9,56,60,93.3,+5.6,133
-claude_code_opus_4_7_low,1806,1845,97.9,6,1806,1845,97.9,90,90,100.0,-2.1,599,615,97.4,30,30,100.0,-2.6,1224,1233,99.3,60,60,100.0,-0.7,470
-claude_code_opus_4_7_high,1792,1845,97.1,6,1792,1845,97.1,90,90,100.0,-2.9,596,615,96.9,30,30,100.0,-3.1,1223,1233,99.2,60,60,100.0,-0.8,889
+claude_code_sonnet_4_6_low,1821,1845,98.7,6,1821,1845,98.7,88,90,97.8,+0.9,595,615,96.7,28,30,93.3,+3.4,1219,1233,98.9,58,60,96.7,+2.2,671
+claude_code_sonnet_4_6_high,1809,1845,98.0,6,1809,1845,98.0,86,90,95.6,+2.5,596,615,96.9,26,30,86.7,+10.2,1220,1233,98.9,56,60,93.3,+5.6,729
+claude_code_opus_4_7_low,1806,1845,97.9,6,1806,1845,97.9,90,90,100.0,-2.1,599,615,97.4,30,30,100.0,-2.6,1224,1233,99.3,60,60,100.0,-0.7,521
+claude_code_opus_4_7_high,1792,1845,97.1,6,1792,1845,97.1,90,90,100.0,-2.9,596,615,96.9,30,30,100.0,-3.1,1223,1233,99.2,60,60,100.0,-0.8,913
 claude_code_mimo_v2_5_high,1786,1845,96.8,4,1190,1230,96.7,57,60,95.0,+1.7,387,410,94.4,17,20,85.0,+9.4,805,822,97.9,37,40,92.5,+5.4,404
-claude_code_mimo_v2_5_low,1786,1845,96.8,2,591,615,96.1,28,30,93.3,+2.8,190,205,92.7,8,10,80.0,+12.7,399,411,97.1,18,20,90.0,+7.1,397
-claude_code_mimo_v2_5_pro_low,1785,1845,96.7,3,891,915,97.4,38,45,84.4,+12.9,295,305,96.7,9,15,60.0,+36.7,584,591,98.8,22,29,75.9,+23.0,409
+claude_code_mimo_v2_5_low,1786,1845,96.8,2,591,615,96.1,28,30,93.3,+2.8,190,205,92.7,8,10,80.0,+12.7,399,411,97.1,18,20,90.0,+7.1,398
+claude_code_mimo_v2_5_pro_low,1785,1845,96.7,3,891,915,97.4,38,45,84.4,+12.9,295,305,96.7,9,15,60.0,+36.7,584,591,98.8,22,29,75.9,+23.0,412
 claude_code_mimo_v2_5_pro_high,1784,1845,96.7,5,1484,1530,97.0,69,75,92.0,+5.0,489,510,95.9,19,25,76.0,+19.9,985,1002,98.3,43,49,87.8,+10.5,413
-codex_gpt5_5_low,1781,1845,96.5,6,1781,1845,96.5,90,90,100.0,-3.5,594,615,96.6,30,30,100.0,-3.4,1221,1233,99.0,60,60,100.0,-1.0,390
+codex_gpt5_5_low,1781,1845,96.5,6,1781,1845,96.5,90,90,100.0,-3.5,594,615,96.6,30,30,100.0,-3.4,1221,1233,99.0,60,60,100.0,-1.0,389
 codex_gpt5_5_high,1770,1845,95.9,6,1770,1845,95.9,90,90,100.0,-4.1,589,615,95.8,30,30,100.0,-4.2,1215,1233,98.5,60,60,100.0,-1.5,547
-codex_gpt5_4_mini_low,1756,1845,95.2,5,1472,1545,95.3,73,75,97.3,-2.1,476,515,92.4,23,25,92.0,+0.4,1017,1053,96.6,49,51,96.1,+0.5,331
-codex_gpt5_4_mini_high,1737,1845,94.1,6,1737,1845,94.1,90,90,100.0,-5.9,570,615,92.7,30,30,100.0,-7.3,1189,1233,96.4,60,60,100.0,-3.6,639
+codex_gpt5_4_mini_low,1756,1845,95.2,5,1472,1545,95.3,73,75,97.3,-2.1,476,515,92.4,23,25,92.0,+0.4,1017,1053,96.6,49,51,96.1,+0.5,329
+codex_gpt5_4_mini_high,1737,1845,94.1,6,1737,1845,94.1,90,90,100.0,-5.9,570,615,92.7,30,30,100.0,-7.3,1189,1233,96.4,60,60,100.0,-3.6,638

runs_release/reports/eandd_final/control_exclusions/finder_selection.csv

@@ -1,13 +1,13 @@
 selection,finder_id,family,base_model,effort,formal_accuracy,formal_run_ci95_low,formal_run_ci95_high,formal_ci_width,question_only_accuracy,formal_minus_question_only,low_prior_formal_accuracy,low_prior_question_only_accuracy,low_prior_lift,implementation_signal_formal_accuracy,implementation_signal_question_only_accuracy,implementation_signal_lift,spec_only_accuracy,mean_find_tokens,median_find_tokens,rank_stability_mean_tau,rank_stability_median_tau,same_family_minus_different_family,same_base_minus_different_base,selection_reason
-core_default,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.986992,0.981417,0.992567,0.011150,0.977778,0.009214,0.967480,0.933333,0.034146,0.988646,0.966667,0.021979,1.000000,116083.000000,108429.000000,0.613333,0.666667,0.005389,-0.004545,passes all default core gates
-core_default,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.980488,0.973787,0.987189,0.013402,0.955556,0.024932,0.969106,0.866667,0.102439,0.989457,0.933333,0.056123,1.000000,133456.000000,115841.000000,0.493333,0.600000,0.009222,-0.006818,passes all default core gates
-prior_stress_audit,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.967480,0.959588,0.975371,0.015783,0.844444,0.129326,0.967213,0.600000,0.367213,0.988156,0.758621,0.229535,1.000000,409182.000000,369992.000000,0.600000,0.666667,-0.018710,0.011111,strong finder-specific prior lift; calibration audits only
-not_selected,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.978862,0.971735,0.985988,0.014253,1.000000,-0.021138,0.973984,1.000000,-0.026016,0.992701,1.000000,-0.007299,1.000000,469590.000000,393682.000000,0.106667,0.100000,0.002778,0.026263,mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only; rank-stability mean tau < 0.40
-not_selected,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.971274,0.962597,0.979950,0.017353,1.000000,-0.028726,0.969106,1.000000,-0.030894,0.991890,1.000000,-0.008110,1.000000,889263.000000,929709.000000,0.293333,0.200000,0.012944,0.035690,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only; rank-stability mean tau < 0.40
-not_selected,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.968022,0.959053,0.976991,0.017938,0.933333,0.027642,0.926829,0.800000,0.126829,0.970803,0.900000,0.070803,1.000000,397099.000000,357823.000000,0.600000,0.666667,-0.028244,-0.034174,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000
+prior_stress_audit,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.967480,0.959588,0.975371,0.015783,0.844444,0.129326,0.967213,0.600000,0.367213,0.988156,0.758621,0.229535,1.000000,412001.000000,371420.000000,0.600000,0.666667,-0.018710,0.011111,strong finder-specific prior lift; calibration audits only
+not_selected,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.986992,0.981417,0.992567,0.011150,0.977778,0.009214,0.967480,0.933333,0.034146,0.988646,0.966667,0.021979,1.000000,670960.000000,629543.000000,0.613333,0.666667,0.005389,-0.004545,mean tokens > 150000
+not_selected,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.980488,0.973787,0.987189,0.013402,0.955556,0.024932,0.969106,0.866667,0.102439,0.989457,0.933333,0.056123,1.000000,729460.000000,684900.000000,0.493333,0.600000,0.009222,-0.006818,mean tokens > 150000
+not_selected,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.978862,0.971735,0.985988,0.014253,1.000000,-0.021138,0.973984,1.000000,-0.026016,0.992701,1.000000,-0.007299,1.000000,521407.000000,442934.000000,0.106667,0.100000,0.002778,0.026263,mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only; rank-stability mean tau < 0.40
+not_selected,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.971274,0.962597,0.979950,0.017353,1.000000,-0.028726,0.969106,1.000000,-0.030894,0.991890,1.000000,-0.008110,1.000000,912645.000000,939938.000000,0.293333,0.200000,0.012944,0.035690,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only; rank-stability mean tau < 0.40
+not_selected,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.968022,0.959053,0.976991,0.017938,0.933333,0.027642,0.926829,0.800000,0.126829,0.970803,0.900000,0.070803,1.000000,397627.000000,357823.000000,0.600000,0.666667,-0.028244,-0.034174,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000
 not_selected,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.968022,0.959849,0.976194,0.016345,0.950000,0.017480,0.943902,0.850000,0.093902,0.979319,0.925000,0.054319,0.988889,404049.000000,341233.000000,0.546667,0.600000,-0.040000,-0.049486,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000
 not_selected,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.966938,0.958223,0.975652,0.017429,0.920000,0.049935,0.958824,0.760000,0.198824,0.983034,0.877551,0.105483,1.000000,413189.000000,367753.000000,0.666667,0.666667,-0.023871,0.019688,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000
-not_selected,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.965312,0.955611,0.975012,0.019401,1.000000,-0.034688,0.965854,1.000000,-0.034146,0.990268,1.000000,-0.009732,1.000000,389857.000000,376231.000000,0.573333,0.666667,0.017322,0.036195,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only
+not_selected,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.965312,0.955611,0.975012,0.019401,1.000000,-0.034688,0.965854,1.000000,-0.034146,0.990268,1.000000,-0.009732,1.000000,389499.000000,376231.000000,0.573333,0.666667,0.017322,0.036195,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only
 not_selected,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.959350,0.948199,0.970500,0.022301,1.000000,-0.040650,0.957724,1.000000,-0.042276,0.985401,1.000000,-0.014599,1.000000,546827.000000,520718.000000,0.693333,0.666667,0.026724,0.043603,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only
-not_selected,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.951762,0.940243,0.963280,0.023037,0.973333,-0.020583,0.924272,0.920000,0.004272,0.965812,0.960784,0.005028,1.000000,330733.000000,312134.000000,0.573333,0.733333,0.022336,-0.014566,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only
-not_selected,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.941463,0.927755,0.955172,0.027417,1.000000,-0.058537,0.926829,1.000000,-0.073171,0.964315,1.000000,-0.035685,1.000000,638768.000000,608496.000000,0.693333,0.733333,0.026895,-0.005976,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only
+not_selected,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.951762,0.940243,0.963280,0.023037,0.973333,-0.020583,0.924272,0.920000,0.004272,0.965812,0.960784,0.005028,1.000000,329206.000000,310677.000000,0.573333,0.733333,0.022336,-0.014566,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only
+not_selected,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.941463,0.927755,0.955172,0.027417,1.000000,-0.058537,0.926829,1.000000,-0.073171,0.964315,1.000000,-0.035685,1.000000,638067.000000,608496.000000,0.693333,0.733333,0.026895,-0.005976,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only

runs_release/reports/eandd_final/control_exclusions/printed_values.csv

@@ -146,7 +146,7 @@ table8.codex_gpt5_4_mini_high.delta_all,Table 9 (tab:finder_calibration),-5.9,-5
 table8.codex_gpt5_4_mini_high.delta_low_prior,Table 9 (tab:finder_calibration),-7.3,-7.3171 pp,0
 table8.codex_gpt5_4_mini_high.delta_implementation_signal,Table 9 (tab:finder_calibration),-3.6,-3.5685 pp,0
 app_e.largest_question_only_lift.claude_code_mimo_v2_5_pro_low,"App. E (app:finder_calibration), selector paragraph",+12.9,+12.9326 pp,0
-app_e.selection_unchanged,"App. E (app:finder_calibration), selector paragraph",unchanged,"core claude_code_sonnet_4_6_low, claude_code_sonnet_4_6_high; audit claude_code_mimo_v2_5_pro_low",0
+app_e.selection_unchanged,"App. E (app:finder_calibration), selector paragraph",unchanged,core none; audit claude_code_mimo_v2_5_pro_low,0
 app_f.clusters_overall,"App. F (app:metric_details), Uncertainty",30,,0
 app_f.clusters_low_prior,"App. F (app:metric_details), Uncertainty",10,,0
 app_f.clusters_per_task,"App. F (app:metric_details), Uncertainty",15,,0

runs_release/reports/eandd_final/control_exclusions/report.md

@@ -12,7 +12,7 @@ Differences from the question-only control use matched finder x task x trial str
 
 - Released `bootstrap_lift.csv` reproduced exactly: yes
 - Frozen low-prior subset equals the released-control rule (agreement < 90%): yes
-- Core-finder selection unchanged: yes (core `claude_code_sonnet_4_6_low`, `claude_code_sonnet_4_6_high`; audit `claude_code_mimo_v2_5_pro_low`)
+- Core-finder selection unchanged: yes (core none; audit `claude_code_mimo_v2_5_pro_low`)
 - Label sources after v1.1 corrections: consensus_prelabel 681, manual_review 24, rule_empty_artifact 15
 
 ## Printed values
@@ -231,7 +231,7 @@ printed precision.
 | value | printed | exact |
 | --- | ---: | --- |
 | `app_e.largest_question_only_lift.claude_code_mimo_v2_5_pro_low` | +12.9 | +12.9326 pp |
-| `app_e.selection_unchanged` | unchanged | core claude_code_sonnet_4_6_low, claude_code_sonnet_4_6_high; audit claude_code_mimo_v2_5_pro_low |
+| `app_e.selection_unchanged` | unchanged | core none; audit claude_code_mimo_v2_5_pro_low |
 
 ### App. F (app:metric_details), Uncertainty
 
@@ -257,9 +257,9 @@ of Table 9, so a lift can differ from formal minus question-only accuracy.
 
 | finder | selection | q-only lift | low-prior lift | impl.-signal lift |
 | --- | --- | ---: | ---: | ---: |
-| `claude_code_sonnet_4_6_low` | core_default | +0.9 | +3.4 | +2.2 |
-| `claude_code_sonnet_4_6_high` | core_default | +2.5 | +10.2 | +5.6 |
 | `claude_code_mimo_v2_5_pro_low` | prior_stress_audit | +12.9 | +36.7 | +23.0 |
+| `claude_code_sonnet_4_6_low` | not_selected | +0.9 | +3.4 | +2.2 |
+| `claude_code_sonnet_4_6_high` | not_selected | +2.5 | +10.2 | +5.6 |
 | `claude_code_opus_4_7_low` | not_selected | -2.1 | -2.6 | -0.7 |
 | `claude_code_opus_4_7_high` | not_selected | -2.9 | -3.1 | -0.8 |
 | `claude_code_mimo_v2_5_low` | not_selected | +2.8 | +12.7 | +7.1 |

runs_release/reports/eandd_final/control_exclusions/summary.json

@@ -10,10 +10,7 @@
     "seed": 20260602
   },
   "core_finder_selection": {
-    "default_core_finders": [
-      "claude_code_sonnet_4_6_low",
-      "claude_code_sonnet_4_6_high"
-    ],
+    "default_core_finders": [],
     "inputs_replaced": [
       "question_only_accuracy",
       "formal_minus_question_only",
@@ -27,10 +24,7 @@
     "prior_stress_audit_finders": [
       "claude_code_mimo_v2_5_pro_low"
     ],
-    "released_default_core_finders": [
-      "claude_code_sonnet_4_6_low",
-      "claude_code_sonnet_4_6_high"
-    ],
+    "released_default_core_finders": [],
     "released_prior_stress_audit_finders": [
       "claude_code_mimo_v2_5_pro_low"
     ],
@@ -48,8 +42,8 @@
     "runs_release/reports/artifact_conformance_consensus/final/artifact_conformance_final.csv": "d89d92dc92b00488984cb52965c5cb1f774894fea845ef90ed9b9b8eee3a2b1e",
     "runs_release/reports/artifact_conformance_consensus/final/label_corrections_v1_1.csv": "6430dd0b42919b7f8b25214ae10d20f0f5958cb8c784093f92bd084e08910b29",
     "runs_release/reports/eandd_final/control_exclusions/excluded_question_only_runs.csv": "9b8566d5f3064ae12289d93697fa7ae38c3a9e5e0b789bb06977372b41e87858",
-    "runs_release/reports/eandd_final/core_finder_selection/core_finders.json": "d6e370e13a168f32059e83b09715ea87f5f84521704dc00f3732bcdbf75c8e30",
-    "runs_release/reports/eandd_final/finder_summary.csv": "d3e32cbf72eb436caf03335a2221387994c23ed916b4868a0b5f02534567e790",
+    "runs_release/reports/eandd_final/core_finder_selection/core_finders.json": "651cdab5eaa7d5256413a3f58ae5bc444ada80db02b6950f64d14d0bc14552ff",
+    "runs_release/reports/eandd_final/finder_summary.csv": "d5d5d6df99eecbbdc457a6591a6513f0b61031d042573d654d98654d5672aba2",
     "runs_release/reports/eandd_final/low_prior/manifest.json": "f47e43ba9ec702bdb79241dee001aab094d63be95102d80c1fa9aff740e8f050",
     "runs_release/reports/eandd_final/pivots/affinity_by_agent.csv": "9fcbaff2b04874b1a6ac96bc91536ffada4503085f0c230bc864aadea30e955b",
     "runs_release/reports/eandd_final/pivots/manifest.json": "f327294d9f0e37343fdff0393ea1ecade43d7e6576739ea18b7031ff10416817",

runs_release/reports/eandd_final/core_finder_selection/core_finders.json

@@ -1,12 +1,9 @@
 {
   "analysis_script": "scripts/select_eandd_core_finders.py",
-  "default_core_finders": [
-    "claude_code_sonnet_4_6_low",
-    "claude_code_sonnet_4_6_high"
-  ],
+  "default_core_finders": [],
   "generated_at": "2026-05-03T20:04:26.141382+00:00",
   "input_hashes": {
-    "runs_release/reports/eandd_final/finder_summary.csv": "d3e32cbf72eb436caf03335a2221387994c23ed916b4868a0b5f02534567e790",
+    "runs_release/reports/eandd_final/finder_summary.csv": "d5d5d6df99eecbbdc457a6591a6513f0b61031d042573d654d98654d5672aba2",
     "runs_release/reports/eandd_final/pivots/affinity_by_agent.csv": "9fcbaff2b04874b1a6ac96bc91536ffada4503085f0c230bc864aadea30e955b",
     "runs_release/reports/eandd_final/pivots/manifest.json": "f327294d9f0e37343fdff0393ea1ecade43d7e6576739ea18b7031ff10416817",
     "runs_release/reports/eandd_final/within_finder_efficiency/within_finder_cells.csv": "00ef3820d18abca3219eca0f79266f5001c655a881f1d8b8cd7df7696ef4e94b",

runs_release/reports/eandd_final/core_finder_selection/core_finders.md

@@ -2,12 +2,12 @@
 
 - Selection id: `eandd_core_finders_v1`
 - Script: `scripts/select_eandd_core_finders.py`
-- Default core finders: `claude_code_sonnet_4_6_low`, `claude_code_sonnet_4_6_high`
+- Default core finders: none
 - Prior-stress audit finders: `claude_code_mimo_v2_5_pro_low`
 
 ## Interpretation
 
-The default core panel is intentionally smaller than the four highest formal-accuracy finders. The Opus and Codex finders have high artifact-conditioned accuracy, but their finder-specific question-only controls are saturated, so they are not clean routine instruments for future artifact-conditioned scoring in this task pack.
+The default core gates select 0 of 12 candidates.
 
 Spec-only controls are also saturated. The selector records that fact as a validity limitation and uses finder-specific question-only lift, low-prior lift, and implementation-signal lift as the prior controls that can still discriminate finder behavior.
 
@@ -15,20 +15,20 @@ Spec-only controls are also saturated. The selector records that fact as a valid
 
 | role | finder | formal | q-only lift | low-prior lift | impl.-signal lift | mean tokens | tau | affinity |
 | --- | --- | ---: | ---: | ---: | ---: | ---: | ---: | ---: |
-| core_default | `claude_code_sonnet_4_6_low` | 98.7% | 0.9% | 3.4% | 2.2% | 116083 | 0.613 | 0.5% |
-| core_default | `claude_code_sonnet_4_6_high` | 98.0% | 2.5% | 10.2% | 5.6% | 133456 | 0.493 | 0.9% |
-| prior_stress_audit | `claude_code_mimo_v2_5_pro_low` | 96.7% | 32.3% | 37.2% | 37.6% | 409182 | 0.600 | -1.9% |
+| prior_stress_audit | `claude_code_mimo_v2_5_pro_low` | 96.7% | 32.3% | 37.2% | 37.6% | 412001 | 0.600 | -1.9% |
 
 ## Non-Selected Summary
 
 | finder | formal | q-only lift | mean tokens | tau | main reason |
 | --- | ---: | ---: | ---: | ---: | --- |
-| `claude_code_opus_4_7_low` | 97.9% | -2.1% | 469590 | 0.107 | mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only; rank-stability mean tau < 0.40 |
-| `claude_code_opus_4_7_high` | 97.1% | -2.9% | 889263 | 0.293 | formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only; rank-stability mean tau < 0.40 |
-| `claude_code_mimo_v2_5_low` | 96.8% | 1.2% | 397099 | 0.600 | formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000 |
+| `claude_code_sonnet_4_6_low` | 98.7% | 0.9% | 670960 | 0.613 | mean tokens > 150000 |
+| `claude_code_sonnet_4_6_high` | 98.0% | 2.5% | 729460 | 0.493 | mean tokens > 150000 |
+| `claude_code_opus_4_7_low` | 97.9% | -2.1% | 521407 | 0.107 | mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only; rank-stability mean tau < 0.40 |
+| `claude_code_opus_4_7_high` | 97.1% | -2.9% | 912645 | 0.293 | formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only; rank-stability mean tau < 0.40 |
+| `claude_code_mimo_v2_5_low` | 96.8% | 1.2% | 397627 | 0.600 | formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000 |
 | `claude_code_mimo_v2_5_high` | 96.8% | 3.5% | 404049 | 0.547 | formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000 |
 | `claude_code_mimo_v2_5_pro_high` | 96.7% | 6.7% | 413189 | 0.667 | formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000 |
-| `codex_gpt5_5_low` | 96.5% | -3.5% | 389857 | 0.573 | formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only |
+| `codex_gpt5_5_low` | 96.5% | -3.5% | 389499 | 0.573 | formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only |
 | `codex_gpt5_5_high` | 95.9% | -4.1% | 546827 | 0.693 | formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only |
-| `codex_gpt5_4_mini_low` | 95.2% | -2.6% | 330733 | 0.573 | formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only |
-| `codex_gpt5_4_mini_high` | 94.1% | -5.9% | 638768 | 0.693 | formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only |
+| `codex_gpt5_4_mini_low` | 95.2% | -2.6% | 329206 | 0.573 | formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only |
+| `codex_gpt5_4_mini_high` | 94.1% | -5.9% | 638067 | 0.693 | formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only |

runs_release/reports/eandd_final/core_finder_selection/finder_selection.csv

@@ -1,13 +1,13 @@
 selection,finder_id,family,base_model,effort,formal_accuracy,formal_run_ci95_low,formal_run_ci95_high,formal_ci_width,question_only_accuracy,formal_minus_question_only,low_prior_formal_accuracy,low_prior_question_only_accuracy,low_prior_lift,implementation_signal_formal_accuracy,implementation_signal_question_only_accuracy,implementation_signal_lift,spec_only_accuracy,mean_find_tokens,median_find_tokens,rank_stability_mean_tau,rank_stability_median_tau,same_family_minus_different_family,same_base_minus_different_base,selection_reason
-core_default,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.986992,0.981417,0.992567,0.011150,0.977778,0.009214,0.967480,0.933333,0.034146,0.988646,0.966667,0.021979,1.000000,116083.000000,108429.000000,0.613333,0.666667,0.005389,-0.004545,passes all default core gates
-core_default,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.980488,0.973787,0.987189,0.013402,0.955556,0.024932,0.969106,0.866667,0.102439,0.989457,0.933333,0.056123,1.000000,133456.000000,115841.000000,0.493333,0.600000,0.009222,-0.006818,passes all default core gates
-prior_stress_audit,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.967480,0.959588,0.975371,0.015783,0.644444,0.323035,0.938211,0.566667,0.371545,0.975669,0.600000,0.375669,1.000000,409182.000000,369992.000000,0.600000,0.666667,-0.018710,0.011111,strong finder-specific prior lift; calibration audits only
-not_selected,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.978862,0.971735,0.985988,0.014253,1.000000,-0.021138,0.973984,1.000000,-0.026016,0.992701,1.000000,-0.007299,1.000000,469590.000000,393682.000000,0.106667,0.100000,0.002778,0.026263,mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only; rank-stability mean tau < 0.40
-not_selected,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.971274,0.962597,0.979950,0.017353,1.000000,-0.028726,0.969106,1.000000,-0.030894,0.991890,1.000000,-0.008110,1.000000,889263.000000,929709.000000,0.293333,0.200000,0.012944,0.035690,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only; rank-stability mean tau < 0.40
-not_selected,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.968022,0.959053,0.976991,0.017938,0.955556,0.012466,0.944715,0.866667,0.078049,0.979724,0.933333,0.046391,1.000000,397099.000000,357823.000000,0.600000,0.666667,-0.028244,-0.034174,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000
+prior_stress_audit,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.967480,0.959588,0.975371,0.015783,0.644444,0.323035,0.938211,0.566667,0.371545,0.975669,0.600000,0.375669,1.000000,412001.000000,371420.000000,0.600000,0.666667,-0.018710,0.011111,strong finder-specific prior lift; calibration audits only
+not_selected,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.986992,0.981417,0.992567,0.011150,0.977778,0.009214,0.967480,0.933333,0.034146,0.988646,0.966667,0.021979,1.000000,670960.000000,629543.000000,0.613333,0.666667,0.005389,-0.004545,mean tokens > 150000
+not_selected,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.980488,0.973787,0.987189,0.013402,0.955556,0.024932,0.969106,0.866667,0.102439,0.989457,0.933333,0.056123,1.000000,729460.000000,684900.000000,0.493333,0.600000,0.009222,-0.006818,mean tokens > 150000
+not_selected,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.978862,0.971735,0.985988,0.014253,1.000000,-0.021138,0.973984,1.000000,-0.026016,0.992701,1.000000,-0.007299,1.000000,521407.000000,442934.000000,0.106667,0.100000,0.002778,0.026263,mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only; rank-stability mean tau < 0.40
+not_selected,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.971274,0.962597,0.979950,0.017353,1.000000,-0.028726,0.969106,1.000000,-0.030894,0.991890,1.000000,-0.008110,1.000000,912645.000000,939938.000000,0.293333,0.200000,0.012944,0.035690,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only; rank-stability mean tau < 0.40
+not_selected,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.968022,0.959053,0.976991,0.017938,0.955556,0.012466,0.944715,0.866667,0.078049,0.979724,0.933333,0.046391,1.000000,397627.000000,357823.000000,0.600000,0.666667,-0.028244,-0.034174,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000
 not_selected,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.968022,0.959849,0.976194,0.016345,0.933333,0.034688,0.947967,0.800000,0.147967,0.978913,0.900000,0.078913,0.988889,404049.000000,341233.000000,0.546667,0.600000,-0.040000,-0.049486,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000
 not_selected,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.966938,0.958223,0.975652,0.017429,0.900000,0.066938,0.946341,0.700000,0.246341,0.977291,0.866667,0.110624,1.000000,413189.000000,367753.000000,0.666667,0.666667,-0.023871,0.019688,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000
-not_selected,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.965312,0.955611,0.975012,0.019401,1.000000,-0.034688,0.965854,1.000000,-0.034146,0.990268,1.000000,-0.009732,1.000000,389857.000000,376231.000000,0.573333,0.666667,0.017322,0.036195,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only
+not_selected,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.965312,0.955611,0.975012,0.019401,1.000000,-0.034688,0.965854,1.000000,-0.034146,0.990268,1.000000,-0.009732,1.000000,389499.000000,376231.000000,0.573333,0.666667,0.017322,0.036195,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only
 not_selected,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.959350,0.948199,0.970500,0.022301,1.000000,-0.040650,0.957724,1.000000,-0.042276,0.985401,1.000000,-0.014599,1.000000,546827.000000,520718.000000,0.693333,0.666667,0.026724,0.043603,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only
-not_selected,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.951762,0.940243,0.963280,0.023037,0.977778,-0.026016,0.925203,0.933333,-0.008130,0.965126,0.966667,-0.001541,1.000000,330733.000000,312134.000000,0.573333,0.733333,0.022336,-0.014566,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only
-not_selected,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.941463,0.927755,0.955172,0.027417,1.000000,-0.058537,0.926829,1.000000,-0.073171,0.964315,1.000000,-0.035685,1.000000,638768.000000,608496.000000,0.693333,0.733333,0.026895,-0.005976,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only
+not_selected,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.951762,0.940243,0.963280,0.023037,0.977778,-0.026016,0.925203,0.933333,-0.008130,0.965126,0.966667,-0.001541,1.000000,329206.000000,310677.000000,0.573333,0.733333,0.022336,-0.014566,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only
+not_selected,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.941463,0.927755,0.955172,0.027417,1.000000,-0.058537,0.926829,1.000000,-0.073171,0.964315,1.000000,-0.035685,1.000000,638067.000000,608496.000000,0.693333,0.733333,0.026895,-0.005976,formal CI low < 0.970; formal CI width > 0.015; mean tokens > 150000; formal <= question-only; low-prior formal <= question-only; implementation-signal formal <= question-only

runs_release/reports/eandd_final/finder_summary.csv

@@ -1,13 +1,13 @@
 finder_id,family,base_model,effort,formal_accuracy,formal_run_ci95_low,formal_run_ci95_high,spec_only_accuracy,question_only_accuracy,formal_minus_spec_only,formal_minus_question_only,spec_only_stable_questions,spec_only_total_questions,question_only_stable_questions,question_only_total_questions,median_find_tokens,mean_find_tokens,total_find_tokens,total_find_cost_usd,find_runs,question_rows
-claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.986992,0.981417,0.992567,1.000000,0.977778,-0.013008,0.009214,30,30,29,30,108429,116083,14278298,36.5433,123,1845
-claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.980488,0.973787,0.987189,1.000000,0.955556,-0.019512,0.024932,30,30,27,30,115841,133456,16415091,45.5810,123,1845
-claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.978862,0.971735,0.985988,1.000000,1.000000,-0.021138,-0.021138,30,30,30,30,393682,469590,57759611,59.7287,123,1845
-claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.971274,0.962597,0.979950,1.000000,1.000000,-0.028726,-0.028726,30,30,30,30,929709,889263,109379454,112.9163,123,1845
+claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.986992,0.981417,0.992567,1.000000,0.977778,-0.013008,0.009214,30,30,29,30,629543,670960,82528195,36.5433,123,1845
+claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.980488,0.973787,0.987189,1.000000,0.955556,-0.019512,0.024932,30,30,27,30,684900,729460,89723641,45.5810,123,1845
+claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.978862,0.971735,0.985988,1.000000,1.000000,-0.021138,-0.021138,30,30,30,30,442934,521407,64133147,59.7287,123,1845
+claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.971274,0.962597,0.979950,1.000000,1.000000,-0.028726,-0.028726,30,30,30,30,939938,912645,112255420,112.9163,123,1845
 claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.968022,0.959849,0.976194,0.988889,0.933333,-0.020867,0.034688,29,30,25,30,341233,404049,49698111,51.5040,123,1845
-claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.968022,0.959053,0.976991,1.000000,0.955556,-0.031978,0.012466,30,30,27,30,357823,397099,48843244,51.0743,123,1845
-claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.967480,0.959588,0.975371,1.000000,0.644444,-0.032520,0.323035,30,30,3,30,369992,409182,50329438,52.6152,123,1845
+claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.968022,0.959053,0.976991,1.000000,0.955556,-0.031978,0.012466,30,30,27,30,357823,397627,48908233,51.0743,123,1845
+claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.967480,0.959588,0.975371,1.000000,0.644444,-0.032520,0.323035,30,30,3,30,371420,412001,50676169,52.6152,123,1845
 claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.966938,0.958223,0.975652,1.000000,0.900000,-0.033062,0.066938,30,30,23,30,367753,413189,50822350,53.0428,123,1845
-codex_gpt5_5_low,codex,codex_gpt5_5,low,0.965312,0.955611,0.975012,1.000000,1.000000,-0.034688,-0.034688,30,30,30,30,376231,389857,47952491,0.0000,123,1845
+codex_gpt5_5_low,codex,codex_gpt5_5,low,0.965312,0.955611,0.975012,1.000000,1.000000,-0.034688,-0.034688,30,30,30,30,376231,389499,47908484,0.0000,123,1845
 codex_gpt5_5_high,codex,codex_gpt5_5,high,0.959350,0.948199,0.970500,1.000000,1.000000,-0.040650,-0.040650,30,30,30,30,520718,546827,67259729,0.0000,123,1845
-codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.951762,0.940243,0.963280,1.000000,0.977778,-0.048238,-0.026016,30,30,28,30,312134,330733,40680176,0.0000,123,1845
-codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.941463,0.927755,0.955172,1.000000,1.000000,-0.058537,-0.058537,30,30,30,30,608496,638768,78568500,0.0000,123,1845
+codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.951762,0.940243,0.963280,1.000000,0.977778,-0.048238,-0.026016,30,30,28,30,310677,329206,40492384,0.0000,123,1845
+codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.941463,0.927755,0.955172,1.000000,1.000000,-0.058537,-0.058537,30,30,30,30,608496,638067,78482341,0.0000,123,1845

runs_release/reports/eandd_final/insights.md

@@ -98,8 +98,8 @@ Positive values mean high effort outperformed low effort for the same base model
 
 Token and cost totals are reported only as run-cost diagnostics. They are not used as direct recovery-efficiency metrics because formal, spec-only, and question-only runs have different inputs.
 
-- Formal tokens: 934,533,272
-- Control tokens: 19,189,616
+- Formal tokens: 1,095,444,301
+- Control tokens: 19,311,382
 - Formal cost USD: 616.00
 - Control cost USD: 19.42
 

runs_release/reports/eandd_final/metrics.json

@@ -72,8 +72,8 @@
       "downstream_run_ci95_low": 0.9911613209145976,
       "effort": "high",
       "family": "claude",
-      "median_build_tokens": 5891623.0,
-      "total_build_tokens": 25675659
+      "median_build_tokens": 5963041.0,
+      "total_build_tokens": 26263322
     },
     {
       "agent_id": "claude_code_opus_4_7_low",
@@ -83,8 +83,8 @@
       "downstream_run_ci95_low": 0.9891586006040337,
       "effort": "low",
       "family": "claude",
-      "median_build_tokens": 4507726.0,
-      "total_build_tokens": 18977400
+      "median_build_tokens": 4588497.5,
+      "total_build_tokens": 19283421
     },
     {
       "agent_id": "claude_code_mimo_v2_5_pro_low",
@@ -138,8 +138,8 @@
       "downstream_run_ci95_low": 0.9615169835322208,
       "effort": "high",
       "family": "claude",
-      "median_build_tokens": 6900882.0,
-      "total_build_tokens": 30836507
+      "median_build_tokens": 7065720.5,
+      "total_build_tokens": 31468782
     },
     {
       "agent_id": "claude_code_mimo_v2_5_pro_high",
@@ -149,8 +149,8 @@
       "downstream_run_ci95_low": 0.9517468391366029,
       "effort": "high",
       "family": "mimo",
-      "median_build_tokens": 22845676.0,
-      "total_build_tokens": 45691352
+      "median_build_tokens": 23108807.5,
+      "total_build_tokens": 46217615
     },
     {
       "agent_id": "claude_code_sonnet_4_6_low",
@@ -160,8 +160,8 @@
       "downstream_run_ci95_low": 0.9390175935771679,
       "effort": "low",
       "family": "claude",
-      "median_build_tokens": 3034817.0,
-      "total_build_tokens": 13123479
+      "median_build_tokens": 5529057.5,
+      "total_build_tokens": 20370498
     },
     {
       "agent_id": "claude_code_mimo_v2_5_low",
@@ -171,8 +171,8 @@
       "downstream_run_ci95_low": 0.9309637085170341,
       "effort": "low",
       "family": "mimo",
-      "median_build_tokens": 11909618,
-      "total_build_tokens": 41578437
+      "median_build_tokens": 12084066,
+      "total_build_tokens": 41928015
     },
     {
       "agent_id": "codex_gpt5_4_mini_low",
@@ -193,8 +193,8 @@
       "downstream_run_ci95_low": 0.9030462995068412,
       "effort": "high",
       "family": "mimo",
-      "median_build_tokens": 14120147.5,
-      "total_build_tokens": 61338562
+      "median_build_tokens": 14214290.0,
+      "total_build_tokens": 61699061
     }
   ],
   "control_diagnostics": {
@@ -312,14 +312,14 @@
       "formal_minus_spec_only": -0.013008130081300862,
       "formal_run_ci95_high": 0.9925669283590308,
       "formal_run_ci95_low": 0.9814168114783677,
-      "median_find_tokens": 108429,
+      "median_find_tokens": 629543,
       "question_only_accuracy": 0.9777777777777777,
       "question_only_stable_questions": 29,
       "question_only_total_questions": 30,
       "spec_only_accuracy": 1.0,
       "spec_only_stable_questions": 30,
       "spec_only_total_questions": 30,
-      "total_find_tokens": 14278298
+      "total_find_tokens": 82528195
     },
     {
       "agent_id": "claude_code_sonnet_4_6_high",
@@ -331,14 +331,14 @@
       "formal_minus_spec_only": -0.019512195121951237,
       "formal_run_ci95_high": 0.9871886514103919,
       "formal_run_ci95_low": 0.9737869583457056,
-      "median_find_tokens": 115841,
+      "median_find_tokens": 684900,
       "question_only_accuracy": 0.9555555555555556,
       "question_only_stable_questions": 27,
       "question_only_total_questions": 30,
       "spec_only_accuracy": 1.0,
       "spec_only_stable_questions": 30,
       "spec_only_total_questions": 30,
-      "total_find_tokens": 16415091
+      "total_find_tokens": 89723641
     },
     {
       "agent_id": "claude_code_opus_4_7_low",
@@ -350,14 +350,14 @@
       "formal_minus_spec_only": -0.021138211382113803,
       "formal_run_ci95_high": 0.9859881190827636,
       "formal_run_ci95_low": 0.9717354581530088,
-      "median_find_tokens": 393682,
+      "median_find_tokens": 442934,
       "question_only_accuracy": 1.0,
       "question_only_stable_questions": 30,
       "question_only_total_questions": 30,
       "spec_only_accuracy": 1.0,
       "spec_only_stable_questions": 30,
       "spec_only_total_questions": 30,
-      "total_find_tokens": 57759611
+      "total_find_tokens": 64133147
     },
     {
       "agent_id": "claude_code_opus_4_7_high",
@@ -369,14 +369,14 @@
       "formal_minus_spec_only": -0.02872628726287263,
       "formal_run_ci95_high": 0.9799500940097176,
       "formal_run_ci95_low": 0.9625973314645372,
-      "median_find_tokens": 929709,
+      "median_find_tokens": 939938,
       "question_only_accuracy": 1.0,
       "question_only_stable_questions": 30,
       "question_only_total_questions": 30,
       "spec_only_accuracy": 1.0,
       "spec_only_stable_questions": 30,
       "spec_only_total_questions": 30,
-      "total_find_tokens": 109379454
+      "total_find_tokens": 112255420
     },
     {
       "agent_id": "claude_code_mimo_v2_5_high",
@@ -414,7 +414,7 @@
       "spec_only_accuracy": 1.0,
       "spec_only_stable_questions": 30,
       "spec_only_total_questions": 30,
-      "total_find_tokens": 48843244
+      "total_find_tokens": 48908233
     },
     {
       "agent_id": "claude_code_mimo_v2_5_pro_low",
@@ -426,14 +426,14 @@
       "formal_minus_spec_only": -0.03252032520325199,
       "formal_run_ci95_high": 0.9753710344885602,
       "formal_run_ci95_low": 0.9595883151049358,
-      "median_find_tokens": 369992,
+      "median_find_tokens": 371420,
       "question_only_accuracy": 0.6444444444444445,
       "question_only_stable_questions": 3,
       "question_only_total_questions": 30,
       "spec_only_accuracy": 1.0,
       "spec_only_stable_questions": 30,
       "spec_only_total_questions": 30,
-      "total_find_tokens": 50329438
+      "total_find_tokens": 50676169
     },
     {
       "agent_id": "claude_code_mimo_v2_5_pro_high",
@@ -471,7 +471,7 @@
       "spec_only_accuracy": 1.0,
       "spec_only_stable_questions": 30,
       "spec_only_total_questions": 30,
-      "total_find_tokens": 47952491
+      "total_find_tokens": 47908484
     },
     {
       "agent_id": "codex_gpt5_5_high",
@@ -502,14 +502,14 @@
       "formal_minus_spec_only": -0.048238482384823866,
       "formal_run_ci95_high": 0.9632797183931744,
       "formal_run_ci95_low": 0.9402433168371779,
-      "median_find_tokens": 312134,
+      "median_find_tokens": 310677,
       "question_only_accuracy": 0.9777777777777777,
       "question_only_stable_questions": 28,
       "question_only_total_questions": 30,
       "spec_only_accuracy": 1.0,
       "spec_only_stable_questions": 30,
       "spec_only_total_questions": 30,
-      "total_find_tokens": 40680176
+      "total_find_tokens": 40492384
     },
     {
       "agent_id": "codex_gpt5_4_mini_high",
@@ -528,7 +528,7 @@
       "spec_only_accuracy": 1.0,
       "spec_only_stable_questions": 30,
       "spec_only_total_questions": 30,
-      "total_find_tokens": 78568500
+      "total_find_tokens": 78482341
     }
   ],
   "generated_at": "2026-05-03T20:04:26.141382+00:00",
@@ -536,10 +536,10 @@
     "control_question_only_accuracy": 0.9453703703703704,
     "control_spec_only_accuracy": 0.9990740740740741,
     "control_total_cost_usd": 19.42142055,
-    "control_total_tokens": 19189616,
+    "control_total_tokens": 19311382,
     "formal_accuracy": 0.967163504968383,
     "formal_total_cost_usd": 616.00270425,
-    "formal_total_tokens": 934533272
+    "formal_total_tokens": 1095444301
   },
   "outputs": {
     "build_exclusions_csv": "runs_release/reports/eandd_final/build_exclusions.csv",

runs_release/reports/eandd_final/pivots/agent_token_diagnostics.csv

@@ -1,205 +1,205 @@
 role,view_id,description,rank,agent_id,family,base_model,effort,accuracy,question_rows,mean_tokens,accuracy_per_million_tokens,pareto_frontier,within_0_005_least_cost,within_0_010_least_cost,within_0_020_least_cost
-builder,all_finders,All finders judging each builder,1,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9944444444444445,2160,6418914.75,0.15492407722729834,1,0,0,0
-builder,all_finders,All finders judging each builder,2,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9925925925925926,2160,4744350.0,0.20921571818955023,1,1,0,0
+builder,all_finders,All finders judging each builder,1,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9944444444444445,2160,6565830.5,0.15145752611866,1,0,0,0
+builder,all_finders,All finders judging each builder,2,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9925925925925926,2160,4820855.25,0.2058955395088024,1,1,0,0
 builder,all_finders,All finders judging each builder,3,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.9925925925925926,540,7278835.0,0.13636695880489014,0,0,0,0
 builder,all_finders,All finders judging each builder,4,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9916666666666667,2160,7470920.75,0.13273687405487022,0,0,0,0
 builder,all_finders,All finders judging each builder,5,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.9888888888888889,2160,3216598.25,0.3074331365096306,1,0,0,0
 builder,all_finders,All finders judging each builder,6,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.986574074074074,2160,1921738.5,0.513375817820205,1,0,1,1
-builder,all_finders,All finders judging each builder,7,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9671296296296297,2160,7709126.75,0.12545255266812544,0,0,0,0
-builder,all_finders,All finders judging each builder,8,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9601851851851851,1080,22845676.0,0.042029186844161896,0,0,0,0
-builder,all_finders,All finders judging each builder,9,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9490740740740741,2160,3280869.75,0.2892751454318094,0,0,0,0
-builder,all_finders,All finders judging each builder,10,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9444444444444444,1620,13859479.0,0.06814429636528505,0,0,0,0
+builder,all_finders,All finders judging each builder,7,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9671296296296297,2160,7867195.5,0.12293194310852319,0,0,0,0
+builder,all_finders,All finders judging each builder,8,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9601851851851851,1080,23108807.5,0.04155061593659409,0,0,0,0
+builder,all_finders,All finders judging each builder,9,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9490740740740741,2160,5092624.5,0.18636246871805962,0,0,0,0
+builder,all_finders,All finders judging each builder,10,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9444444444444444,1620,13976005.0,0.06757613813421248,0,0,0,0
 builder,all_finders,All finders judging each builder,11,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.921604938271605,1620,2536506.0,0.36333639197841633,0,0,0,0
-builder,all_finders,All finders judging each builder,12,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9152777777777777,2160,15334640.5,0.059686940673814796,0,0,0,0
-builder,cross_base_finders,Only judgments where builder and finder are not the same base model,1,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9933333333333333,1800,6418914.75,0.15475097769967008,1,0,0,0
+builder,all_finders,All finders judging each builder,12,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9152777777777777,2160,15424765.25,0.05933819821198107,0,0,0,0
+builder,cross_base_finders,Only judgments where builder and finder are not the same base model,1,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9933333333333333,1800,6565830.5,0.1512882998324939,1,0,0,0
 builder,cross_base_finders,Only judgments where builder and finder are not the same base model,2,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9933333333333333,1800,7470920.75,0.1329599612381557,0,0,0,0
 builder,cross_base_finders,Only judgments where builder and finder are not the same base model,3,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.9933333333333333,450,7278835.0,0.13646872519205797,0,0,0,0
-builder,cross_base_finders,Only judgments where builder and finder are not the same base model,4,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9911111111111112,1800,4744350.0,0.2089034559235957,1,1,0,0
+builder,cross_base_finders,Only judgments where builder and finder are not the same base model,4,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9911111111111112,1800,4820855.25,0.20558823273341617,1,1,0,0
 builder,cross_base_finders,Only judgments where builder and finder are not the same base model,5,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.9872222222222222,1800,3216598.25,0.3069149907739402,1,0,0,0
 builder,cross_base_finders,Only judgments where builder and finder are not the same base model,6,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9855555555555555,1800,1921738.5,0.5128458193222207,1,0,1,1
-builder,cross_base_finders,Only judgments where builder and finder are not the same base model,7,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9633333333333334,1800,7709126.75,0.12496011086253489,0,0,0,0
-builder,cross_base_finders,Only judgments where builder and finder are not the same base model,8,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9566666666666667,900,22845676.0,0.04187517439478117,0,0,0,0
-builder,cross_base_finders,Only judgments where builder and finder are not the same base model,9,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9444444444444444,1800,3280869.75,0.2878640471614103,0,0,0,0
-builder,cross_base_finders,Only judgments where builder and finder are not the same base model,10,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9444444444444444,1350,13859479.0,0.06814429636528505,0,0,0,0
+builder,cross_base_finders,Only judgments where builder and finder are not the same base model,7,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9633333333333334,1800,7867195.5,0.12244939550991625,0,0,0,0
+builder,cross_base_finders,Only judgments where builder and finder are not the same base model,8,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9566666666666667,900,23108807.5,0.04139835717038478,0,0,0,0
+builder,cross_base_finders,Only judgments where builder and finder are not the same base model,9,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9444444444444444,1800,5092624.5,0.18545338350480078,0,0,0,0
+builder,cross_base_finders,Only judgments where builder and finder are not the same base model,10,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9444444444444444,1350,13976005.0,0.06757613813421248,0,0,0,0
 builder,cross_base_finders,Only judgments where builder and finder are not the same base model,11,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9303703703703704,1350,2536506.0,0.3667921031412385,0,0,0,0
-builder,cross_base_finders,Only judgments where builder and finder are not the same base model,12,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9133333333333333,1800,15334640.5,0.05956013988937878,0,0,0,0
-builder,cross_family_finders,Only cross-family finder judgments,1,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9923611111111111,1440,6418914.75,0.15459951561299534,1,0,0,0
+builder,cross_base_finders,Only judgments where builder and finder are not the same base model,12,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9133333333333333,1800,15424765.25,0.05921213830682664,0,0,0,0
+builder,cross_family_finders,Only cross-family finder judgments,1,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9923611111111111,1440,6565830.5,0.15114022683209857,1,0,0,0
 builder,cross_family_finders,Only cross-family finder judgments,2,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9916666666666667,1440,7470920.75,0.13273687405487022,0,0,0,0
 builder,cross_family_finders,Only cross-family finder judgments,3,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.9916666666666667,360,7278835.0,0.13623975082093037,0,0,0,0
-builder,cross_family_finders,Only cross-family finder judgments,4,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9895833333333334,1440,4744350.0,0.20858143546183006,0,0,0,0
+builder,cross_family_finders,Only cross-family finder judgments,4,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9895833333333334,1440,4820855.25,0.20527132262129907,0,0,0,0
 builder,cross_family_finders,Only cross-family finder judgments,5,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.9895833333333334,1440,3216598.25,0.30764903056616827,1,1,0,0
 builder,cross_family_finders,Only cross-family finder judgments,6,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9847222222222223,1440,1921738.5,0.5124121841875064,1,0,1,1
-builder,cross_family_finders,Only cross-family finder judgments,7,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9631944444444445,1440,7709126.75,0.12494209469891573,0,0,0,0
-builder,cross_family_finders,Only cross-family finder judgments,8,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9597222222222223,720,22845676.0,0.04200892204819075,0,0,0,0
-builder,cross_family_finders,Only cross-family finder judgments,9,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9416666666666667,1440,3280869.75,0.2870173881991709,0,0,0,0
-builder,cross_family_finders,Only cross-family finder judgments,10,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9416666666666667,1080,13859479.0,0.06794387196421067,0,0,0,0
+builder,cross_family_finders,Only cross-family finder judgments,7,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9631944444444445,1440,7867195.5,0.12243174132947941,0,0,0,0
+builder,cross_family_finders,Only cross-family finder judgments,8,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9597222222222223,720,23108807.5,0.04153058188840866,0,0,0,0
+builder,cross_family_finders,Only cross-family finder judgments,9,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9416666666666667,1440,5092624.5,0.1849079323768455,0,0,0,0
+builder,cross_family_finders,Only cross-family finder judgments,10,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9416666666666667,1080,13976005.0,0.06737738478675892,0,0,0,0
 builder,cross_family_finders,Only cross-family finder judgments,11,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9388888888888889,1080,2536506.0,0.37015047032764314,0,0,0,0
-builder,cross_family_finders,Only cross-family finder judgments,12,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9104166666666667,1440,15334640.5,0.05936993871272474,0,0,0,0
-builder,hard_formal_questions,Questions with formal accuracy <= 0.90,1,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,1.0,144,6418914.75,0.15578957486543968,0,0,0,0
-builder,hard_formal_questions,Questions with formal accuracy <= 0.90,2,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,1.0,144,4744350.0,0.21077702951932298,0,0,0,0
+builder,cross_family_finders,Only cross-family finder judgments,12,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9104166666666667,1440,15424765.25,0.05902304844909498,0,0,0,0
+builder,hard_formal_questions,Questions with formal accuracy <= 0.90,1,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,1.0,144,6565830.5,0.15230365754949052,0,0,0,0
+builder,hard_formal_questions,Questions with formal accuracy <= 0.90,2,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,1.0,144,4820855.25,0.20743207338573377,0,0,0,0
 builder,hard_formal_questions,Questions with formal accuracy <= 0.90,3,codex_gpt5_5_high,codex,codex_gpt5_5,high,1.0,144,3216598.25,0.3108874414142332,0,0,0,0
 builder,hard_formal_questions,Questions with formal accuracy <= 0.90,4,codex_gpt5_5_low,codex,codex_gpt5_5,low,1.0,144,1921738.5,0.5203621616572702,1,1,1,1
 builder,hard_formal_questions,Questions with formal accuracy <= 0.90,5,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,1.0,144,7470920.75,0.1338523099712977,0,0,0,0
 builder,hard_formal_questions,Questions with formal accuracy <= 0.90,6,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,1.0,36,7278835.0,0.13738462267656842,0,0,0,0
-builder,hard_formal_questions,Questions with formal accuracy <= 0.90,7,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.8703703703703703,108,13859479.0,0.06279964566996857,0,0,0,0
+builder,hard_formal_questions,Questions with formal accuracy <= 0.90,7,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.8703703703703703,108,13976005.0,0.06227604886878406,0,0,0,0
 builder,hard_formal_questions,Questions with formal accuracy <= 0.90,8,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.6481481481481481,108,2536506.0,0.2555279380960061,0,0,0,0
-builder,hard_formal_questions,Questions with formal accuracy <= 0.90,9,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.6180555555555556,144,7709126.75,0.08017192810528839,0,0,0,0
-builder,hard_formal_questions,Questions with formal accuracy <= 0.90,10,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.6111111111111112,72,22845676.0,0.02674953068191596,0,0,0,0
-builder,hard_formal_questions,Questions with formal accuracy <= 0.90,11,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.5694444444444444,144,15334640.5,0.037134515441978856,0,0,0,0
-builder,hard_formal_questions,Questions with formal accuracy <= 0.90,12,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.5555555555555556,144,3280869.75,0.16933179244788843,0,0,0,0
+builder,hard_formal_questions,Questions with formal accuracy <= 0.90,9,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.6180555555555556,144,7867195.5,0.07856110294393416,0,0,0,0
+builder,hard_formal_questions,Questions with formal accuracy <= 0.90,10,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.6111111111111112,72,23108807.5,0.026444943604775416,0,0,0,0
+builder,hard_formal_questions,Questions with formal accuracy <= 0.90,11,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.5694444444444444,144,15424765.25,0.03691754365237062,0,0,0,0
+builder,hard_formal_questions,Questions with formal accuracy <= 0.90,12,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.5555555555555556,144,5092624.5,0.1090902255910593,0,0,0,0
 builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,1,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,1.0,324,7278835.0,0.13738462267656842,1,0,0,0
-builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,2,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9951388888888889,1440,3280869.75,0.30331557322228014,1,1,0,0
-builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,3,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9944444444444445,1440,6418914.75,0.15492407722729834,0,0,0,0
+builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,2,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9951388888888889,1440,5092624.5,0.19540786658998496,1,1,0,0
+builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,3,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9944444444444445,1440,6565830.5,0.15145752611866,0,0,0,0
 builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,4,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9923371647509579,1044,2536506.0,0.3912220845331956,1,0,1,0
-builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,5,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9895833333333334,1440,7709126.75,0.12836516578655727,0,0,0,0
-builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,6,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9888888888888889,1440,4744350.0,0.20843506252466384,0,0,0,0
+builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,5,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9895833333333334,1440,7867195.5,0.12578603561247884,0,0,0,0
+builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,6,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9888888888888889,1440,4820855.25,0.20512727257033675,0,0,0,0
 builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,7,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9881944444444445,1440,7470920.75,0.1322721090896921,0,0,0,0
 builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,8,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.9861111111111112,1440,3216598.25,0.30656956028348,0,0,0,0
 builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,9,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9826388888888888,1440,1921738.5,0.5113280963507204,1,0,0,1
-builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,10,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9810606060606061,792,22845676.0,0.04294294491704277,0,0,0,0
-builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,11,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9458333333333333,1440,15334640.5,0.061679524429238054,0,0,0,0
-builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,12,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9435483870967742,1116,13859479.0,0.06807964333268042,0,0,0,0
+builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,10,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9810606060606061,792,23108807.5,0.04245396938204648,0,0,0,0
+builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,11,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9458333333333333,1440,15424765.25,0.06131913957869364,0,0,0,0
+builder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,12,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9435483870967742,1116,13976005.0,0.06751202415116295,0,0,0,0
 builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,1,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,1.0,180,7278835.0,0.13738462267656842,1,1,0,0
 builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,2,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9925925925925926,540,2536506.0,0.3913227851984551,1,0,1,1
-builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,3,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9888888888888889,720,6418914.75,0.15405857958915703,0,0,0,0
-builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,4,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9805555555555555,720,7709126.75,0.12719411515131146,0,0,0,0
-builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,5,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9791666666666666,720,4744350.0,0.20638584140433708,0,0,0,0
+builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,3,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9888888888888889,720,6565830.5,0.1506113946878295,0,0,0,0
+builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,4,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9805555555555555,720,7867195.5,0.1246385138840843,0,0,0,0
+builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,5,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9791666666666666,720,4820855.25,0.20311057185686432,0,0,0,0
 builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,6,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9777777777777777,720,7470920.75,0.13087781419415775,0,0,0,0
 builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,7,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.9736111111111111,720,3216598.25,0.30268346726580203,0,0,0,0
 builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,8,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9652777777777778,720,1921738.5,0.5022940310441706,1,0,0,0
-builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,9,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9430555555555555,720,3280869.75,0.2874407176802906,0,0,0,0
-builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,10,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.8962962962962963,540,13859479.0,0.06467027341332933,0,0,0,0
-builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,11,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.8861111111111111,360,22845676.0,0.03878681948877814,0,0,0,0
-builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,12,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.8472222222222222,720,15334640.5,0.055248913218553915,0,0,0,0
+builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,9,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9430555555555555,720,5092624.5,0.18518065794082314,0,0,0,0
+builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,10,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.8962962962962963,540,13976005.0,0.064131080111684,0,0,0,0
+builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,11,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.8861111111111111,360,23108807.5,0.038345168226924346,0,0,0,0
+builder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,12,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.8472222222222222,720,15424765.25,0.0549261015315758,0,0,0,0
 builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,1,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.9962962962962963,270,7278835.0,0.1368757907407293,1,0,0,0
-builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,2,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9953703703703703,1080,6418914.75,0.15506832683365523,1,0,0,0
-builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,3,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9925925925925926,1080,4744350.0,0.20921571818955023,1,1,1,0
+builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,2,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9953703703703703,1080,6565830.5,0.1515985480237984,1,0,0,0
+builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,3,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9925925925925926,1080,4820855.25,0.2058955395088024,1,1,1,0
 builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,4,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9907407407407407,1080,7470920.75,0.13261293673082272,0,0,0,0
 builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,5,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.9861111111111112,1080,3216598.25,0.30656956028348,1,0,0,0
 builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,6,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9824074074074074,1080,1921738.5,0.5112076421466331,1,0,0,1
-builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,7,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9731481481481481,1080,7709126.75,0.12623325309162262,0,0,0,0
-builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,8,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9722222222222222,540,22845676.0,0.04255607153941175,0,0,0,0
-builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,9,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9580246913580247,810,13859479.0,0.06912414899275973,0,0,0,0
-builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,10,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9546296296296296,1080,3280869.75,0.2909684633562883,0,0,0,0
+builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,7,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9731481481481481,1080,7867195.5,0.12369695759411954,0,0,0,0
+builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,8,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9722222222222222,540,23108807.5,0.042071501189415425,0,0,0,0
+builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,9,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9580246913580247,810,13976005.0,0.0685478211662077,0,0,0,0
+builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,10,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9546296296296296,1080,5092624.5,0.18745337097397022,0,0,0,0
 builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,11,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9469135802469136,810,2536506.0,0.3733141495612128,0,0,0,0
-builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,12,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9342592592592592,1080,15334640.5,0.0609247578552141,0,0,0,0
+builder,positive_lift_finders,Only finders whose formal accuracy exceeds question-only control,12,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9342592592592592,1080,15424765.25,0.06056878300039342,0,0,0,0
 builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,1,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,1.0,180,7278835.0,0.13738462267656842,1,0,0,0
-builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,2,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9986111111111111,720,6418914.75,0.15557320045590434,0,0,0,0
-builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,3,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9986111111111111,720,4744350.0,0.2104842836449906,1,0,0,0
+builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,2,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9986111111111111,720,6565830.5,0.15209212469178288,0,0,0,0
+builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,3,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9986111111111111,720,4820855.25,0.20714397328380915,1,0,0,0
 builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,4,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.9972222222222222,720,3216598.25,0.31002386518808256,1,1,0,0
 builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,5,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9944444444444445,720,7470920.75,0.13310868602701273,0,0,0,0
 builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,6,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9902777777777778,720,1921738.5,0.5153030850856023,1,0,1,1
-builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,7,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9805555555555555,360,22845676.0,0.04292083786689242,0,0,0,0
-builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,8,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.975,720,7709126.75,0.12647346860654482,0,0,0,0
-builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,9,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9722222222222222,540,13859479.0,0.07014854037602872,0,0,0,0
-builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,10,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9638888888888889,720,3280869.75,0.2937906598970864,0,0,0,0
-builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,11,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9472222222222222,720,15334640.5,0.06177009641812093,0,0,0,0
+builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,7,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9805555555555555,360,23108807.5,0.042432114056753276,0,0,0,0
+builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,8,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.975,720,7867195.5,0.12393234666661074,0,0,0,0
+builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,9,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9722222222222222,540,13976005.0,0.06956367160874814,0,0,0,0
+builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,10,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9638888888888889,720,5092624.5,0.18927154140048788,0,0,0,0
+builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,11,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9472222222222222,720,15424765.25,0.06140918236808966,0,0,0,0
 builder,top4_formal_finders,Only the top four formal-accuracy finders judge builders,12,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9388888888888889,540,2536506.0,0.37015047032764314,0,0,0,0
 builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,1,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,1.0,270,7278835.0,0.13738462267656842,1,0,0,0
-builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,2,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9972222222222222,1080,6418914.75,0.15535682604636902,0,0,0,0
-builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,3,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9972222222222222,1080,4744350.0,0.21019153777065822,1,1,0,0
+builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,2,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9972222222222222,1080,6565830.5,0.15188059183407526,0,0,0,0
+builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,3,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9972222222222222,1080,4820855.25,0.20685587318188453,1,1,0,0
 builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,4,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.9944444444444445,1080,3216598.25,0.3091602889619319,1,0,1,0
 builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,5,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9944444444444445,1080,7470920.75,0.13310868602701273,0,0,0,0
 builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,6,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.987037037037037,1080,1921738.5,0.5136167262283797,1,0,0,1
-builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,7,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9731481481481481,1080,7709126.75,0.12623325309162262,0,0,0,0
-builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,8,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9685185185185186,540,22845676.0,0.042393953171642566,0,0,0,0
-builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,9,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9629629629629629,810,13859479.0,0.06948045903911416,0,0,0,0
-builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,10,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9574074074074074,1080,3280869.75,0.2918151223185277,0,0,0,0
-builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,11,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9398148148148148,1080,15334640.5,0.061287045810745594,0,0,0,0
+builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,7,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9731481481481481,1080,7867195.5,0.12369695759411954,0,0,0,0
+builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,8,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9685185185185186,540,23108807.5,0.041911228803931944,0,0,0,0
+builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,9,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9629629629629629,810,13976005.0,0.06890116045056958,0,0,0,0
+builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,10,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9574074074074074,1080,5092624.5,0.1879988221019255,0,0,0,0
+builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,11,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9398148148148148,1080,15424765.25,0.06092895415797753,0,0,0,0
 builder,top6_formal_finders,Only the top six formal-accuracy finders judge builders,12,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.937037037037037,810,2536506.0,0.3694203905045117,0,0,0,0
-finder,all_builders,All builders judged by each finder,1,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9869918699186991,1845,116083.72357723577,8.502413943174458,1,1,1,1
-finder,all_builders,All builders judged by each finder,2,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9804878048780488,1845,133456.0243902439,7.346898046437879,0,0,0,0
-finder,all_builders,All builders judged by each finder,3,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9788617886178862,1845,469590.3333333333,2.084501573253324,0,0,0,0
-finder,all_builders,All builders judged by each finder,4,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9712737127371274,1845,889263.8536585366,1.0922221888826276,0,0,0,0
+finder,all_builders,All builders judged by each finder,1,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9869918699186991,1845,670960.9349593496,1.4710124218759417,1,1,0,0
+finder,all_builders,All builders judged by each finder,2,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9804878048780488,1845,729460.4959349594,1.344127352121165,0,0,0,0
+finder,all_builders,All builders judged by each finder,3,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9788617886178862,1845,521407.69918699184,1.877344331785247,1,0,1,0
+finder,all_builders,All builders judged by each finder,4,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9712737127371274,1845,912645.6910569106,1.0642396301814796,0,0,0,0
 finder,all_builders,All builders judged by each finder,5,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9680216802168021,1845,404049.68292682926,2.3957986384365126,0,0,0,0
-finder,all_builders,All builders judged by each finder,6,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9680216802168021,1845,397099.5447154472,2.4377305214753275,0,0,0,0
-finder,all_builders,All builders judged by each finder,7,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.967479674796748,1845,409182.42276422767,2.3644213948902033,0,0,0,0
+finder,all_builders,All builders judged by each finder,6,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9680216802168021,1845,397627.9105691057,2.434491278118076,1,0,0,1
+finder,all_builders,All builders judged by each finder,7,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.967479674796748,1845,412001.3739837398,2.3482438066697586,0,0,0,0
 finder,all_builders,All builders judged by each finder,8,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9669376693766938,1845,413189.837398374,2.340177762998628,0,0,0,0
-finder,all_builders,All builders judged by each finder,9,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9653116531165311,1845,389857.65040650405,2.476061844906729,0,0,0,0
+finder,all_builders,All builders judged by each finder,9,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9653116531165311,1845,389499.8699186992,2.4783362657297467,1,0,0,0
 finder,all_builders,All builders judged by each finder,10,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.959349593495935,1845,546827.0650406504,1.7543930336085656,0,0,0,0
-finder,all_builders,All builders judged by each finder,11,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9517615176151761,1845,330733.1382113821,2.8777325512718206,0,0,0,0
-finder,all_builders,All builders judged by each finder,12,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9414634146341463,1845,638768.2926829269,1.4738731170889094,0,0,0,0
-finder,cross_base_builders,Only inputs where builder and finder are not the same base model,1,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9878787878787879,1485,116083.72357723577,8.51005427321176,1,1,1,1
-finder,cross_base_builders,Only inputs where builder and finder are not the same base model,2,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9818181818181818,1485,133456.0243902439,7.356866700503602,0,0,0,0
-finder,cross_base_builders,Only inputs where builder and finder are not the same base model,3,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9764705882352941,1530,404049.68292682926,2.416709205566006,0,0,0,0
-finder,cross_base_builders,Only inputs where builder and finder are not the same base model,4,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9738562091503268,1530,397099.5447154472,2.4524233837844633,0,0,0,0
-finder,cross_base_builders,Only inputs where builder and finder are not the same base model,5,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9737373737373738,1485,469590.3333333333,2.0735890511744786,0,0,0,0
-finder,cross_base_builders,Only inputs where builder and finder are not the same base model,6,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.9666666666666667,1710,409182.42276422767,2.362434486155001,0,0,0,0
+finder,all_builders,All builders judged by each finder,11,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9517615176151761,1845,329206.3739837398,2.8910786449784402,1,0,0,0
+finder,all_builders,All builders judged by each finder,12,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9414634146341463,1845,638067.8130081301,1.4754911553925232,0,0,0,0
+finder,cross_base_builders,Only inputs where builder and finder are not the same base model,1,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9878787878787879,1485,670960.9349593496,1.4723342841690759,1,1,1,0
+finder,cross_base_builders,Only inputs where builder and finder are not the same base model,2,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9818181818181818,1485,729460.4959349594,1.3459511341457526,0,0,0,0
+finder,cross_base_builders,Only inputs where builder and finder are not the same base model,3,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9764705882352941,1530,404049.68292682926,2.416709205566006,1,0,0,0
+finder,cross_base_builders,Only inputs where builder and finder are not the same base model,4,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9738562091503268,1530,397627.9105691057,2.4491646166298873,1,0,0,1
+finder,cross_base_builders,Only inputs where builder and finder are not the same base model,5,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9737373737373738,1485,521407.69918699184,1.8675162934028013,0,0,0,0
+finder,cross_base_builders,Only inputs where builder and finder are not the same base model,6,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.9666666666666667,1710,412001.3739837398,2.3462704925465068,0,0,0,0
 finder,cross_base_builders,Only inputs where builder and finder are not the same base model,7,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9654970760233919,1710,413189.837398374,2.3366912460930513,0,0,0,0
-finder,cross_base_builders,Only inputs where builder and finder are not the same base model,8,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9643097643097643,1485,889263.8536585366,1.0843910503530307,0,0,0,0
-finder,cross_base_builders,Only inputs where builder and finder are not the same base model,9,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9582491582491582,1485,389857.65040650405,2.4579462715429417,0,0,0,0
-finder,cross_base_builders,Only inputs where builder and finder are not the same base model,10,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.954248366013072,1530,330733.1382113821,2.8852517506219213,0,0,0,0
+finder,cross_base_builders,Only inputs where builder and finder are not the same base model,8,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9643097643097643,1485,912645.6910569106,1.0566091241750377,0,0,0,0
+finder,cross_base_builders,Only inputs where builder and finder are not the same base model,9,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9582491582491582,1485,389499.8699186992,2.4602040520557162,1,0,0,0
+finder,cross_base_builders,Only inputs where builder and finder are not the same base model,10,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.954248366013072,1530,329206.3739837398,2.8986327162067775,1,0,0,0
 finder,cross_base_builders,Only inputs where builder and finder are not the same base model,11,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.9508417508417508,1485,546827.0650406504,1.738834471865555,0,0,0,0
-finder,cross_base_builders,Only inputs where builder and finder are not the same base model,12,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9424836601307189,1530,638768.2926829269,1.4754703245712775,0,0,0,0
-finder,cross_family_builders,Only cross-family builder inputs,1,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9848888888888889,1125,116083.72357723577,8.484297871730464,1,1,1,1
-finder,cross_family_builders,Only cross-family builder inputs,2,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9777777777777777,1395,404049.68292682926,2.4199444253860407,0,0,0,0
-finder,cross_family_builders,Only cross-family builder inputs,3,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9777777777777777,1125,469590.3333333333,2.0821931551212605,0,0,0,0
-finder,cross_family_builders,Only cross-family builder inputs,4,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9768888888888889,1125,133456.0243902439,7.319930991143049,0,0,0,0
-finder,cross_family_builders,Only cross-family builder inputs,5,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.974910394265233,1395,397099.5447154472,2.4550780962587915,0,0,0,0
+finder,cross_base_builders,Only inputs where builder and finder are not the same base model,12,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9424836601307189,1530,638067.8130081301,1.4770901163113677,0,0,0,0
+finder,cross_family_builders,Only cross-family builder inputs,1,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9848888888888889,1125,670960.9349593496,1.467878139505333,1,1,0,0
+finder,cross_family_builders,Only cross-family builder inputs,2,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9777777777777777,1395,404049.68292682926,2.4199444253860407,1,0,0,0
+finder,cross_family_builders,Only cross-family builder inputs,3,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9777777777777777,1125,521407.69918699184,1.8752653236658834,0,0,0,0
+finder,cross_family_builders,Only cross-family builder inputs,4,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9768888888888889,1125,729460.4959349594,1.3391936840072431,0,0,0,0
+finder,cross_family_builders,Only cross-family builder inputs,5,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.974910394265233,1395,397627.9105691057,2.4518158015363927,1,0,1,1
 finder,cross_family_builders,Only cross-family builder inputs,6,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9727598566308244,1395,413189.837398374,2.354268591782777,0,0,0,0
-finder,cross_family_builders,Only cross-family builder inputs,7,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.9720430107526882,1395,409182.42276422767,2.3755737213394004,0,0,0,0
-finder,cross_family_builders,Only cross-family builder inputs,8,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9662222222222222,1125,889263.8536585366,1.0865416583020548,0,0,0,0
-finder,cross_family_builders,Only cross-family builder inputs,9,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.958974358974359,1170,389857.65040650405,2.4598064395412984,0,0,0,0
+finder,cross_family_builders,Only cross-family builder inputs,7,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.9720430107526882,1395,412001.3739837398,2.3593198278776883,0,0,0,0
+finder,cross_family_builders,Only cross-family builder inputs,8,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9662222222222222,1125,912645.6910569106,1.0587046338905803,0,0,0,0
+finder,cross_family_builders,Only cross-family builder inputs,9,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.958974358974359,1170,389499.8699186992,2.4620659287370925,1,0,0,0
 finder,cross_family_builders,Only cross-family builder inputs,10,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.9495726495726495,1170,546827.0650406504,1.7365136261169876,0,0,0,0
-finder,cross_family_builders,Only cross-family builder inputs,11,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9435897435897436,1170,330733.1382113821,2.853024491868926,0,0,0,0
-finder,cross_family_builders,Only cross-family builder inputs,12,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9316239316239316,1170,638768.2926829269,1.4584692795426104,0,0,0,0
-finder,hard_formal_questions,Questions with formal accuracy <= 0.90,1,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9349593495934959,123,116083.72357723577,8.054181247652906,1,1,1,1
-finder,hard_formal_questions,Questions with formal accuracy <= 0.90,2,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.8780487804878049,123,133456.0243902439,6.579311683377204,0,0,0,0
+finder,cross_family_builders,Only cross-family builder inputs,11,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9435897435897436,1170,329206.3739837398,2.866255996721222,1,0,0,0
+finder,cross_family_builders,Only cross-family builder inputs,12,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9316239316239316,1170,638067.8130081301,1.460070407300205,0,0,0,0
+finder,hard_formal_questions,Questions with formal accuracy <= 0.90,1,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9349593495934959,123,670960.9349593496,1.3934631673454145,1,1,1,1
+finder,hard_formal_questions,Questions with formal accuracy <= 0.90,2,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.8780487804878049,123,729460.4959349594,1.2036961362279088,0,0,0,0
 finder,hard_formal_questions,Questions with formal accuracy <= 0.90,3,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.8373983739837398,123,413189.837398374,2.026667401251615,0,0,0,0
-finder,hard_formal_questions,Questions with formal accuracy <= 0.90,4,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.8373983739837398,123,409182.42276422767,2.046515997257907,0,0,0,0
-finder,hard_formal_questions,Questions with formal accuracy <= 0.90,5,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.8292682926829268,123,469590.3333333333,1.7659398710285634,0,0,0,0
-finder,hard_formal_questions,Questions with formal accuracy <= 0.90,6,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.8292682926829268,123,397099.5447154472,2.088313380659155,0,0,0,0
+finder,hard_formal_questions,Questions with formal accuracy <= 0.90,4,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.8373983739837398,123,412001.3739837398,2.032513546949455,1,0,0,0
+finder,hard_formal_questions,Questions with formal accuracy <= 0.90,5,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.8292682926829268,123,521407.69918699184,1.5904412113130828,0,0,0,0
+finder,hard_formal_questions,Questions with formal accuracy <= 0.90,6,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.8292682926829268,123,397627.9105691057,2.085538440941017,1,0,0,0
 finder,hard_formal_questions,Questions with formal accuracy <= 0.90,7,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.8211382113821138,123,404049.68292682926,2.0322704015852837,0,0,0,0
-finder,hard_formal_questions,Questions with formal accuracy <= 0.90,8,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.7967479674796748,123,330733.1382113821,2.4090357917822187,0,0,0,0
+finder,hard_formal_questions,Questions with formal accuracy <= 0.90,8,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.7967479674796748,123,329206.3739837398,2.4202082050787626,1,0,0,0
 finder,hard_formal_questions,Questions with formal accuracy <= 0.90,9,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.7886178861788617,123,546827.0650406504,1.4421705445765325,0,0,0,0
-finder,hard_formal_questions,Questions with formal accuracy <= 0.90,10,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.7804878048780488,123,389857.65040650405,2.0019815029004437,0,0,0,0
-finder,hard_formal_questions,Questions with formal accuracy <= 0.90,11,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.7804878048780488,123,638768.2926829269,1.2218637240115313,0,0,0,0
-finder,hard_formal_questions,Questions with formal accuracy <= 0.90,12,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.7642276422764228,123,889263.8536585366,0.859393574957871,0,0,0,0
-finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,1,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9927007299270073,1233,469590.3333333333,2.113971816413755,1,0,0,0
-finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,2,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.991889699918897,1233,889263.8536585366,1.1154053949658986,0,0,0,0
-finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,3,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9902676399026764,1233,389857.65040650405,2.540074919320233,1,0,0,0
-finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,4,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9894566098945661,1233,133456.0243902439,7.414102243906636,1,0,0,0
-finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,5,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9886455798864558,1233,116083.72357723577,8.516659781581396,1,1,1,1
+finder,hard_formal_questions,Questions with formal accuracy <= 0.90,10,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.7804878048780488,123,389499.8699186992,2.003820450674248,0,0,0,0
+finder,hard_formal_questions,Questions with formal accuracy <= 0.90,11,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.7804878048780488,123,638067.8130081301,1.2232051029160815,0,0,0,0
+finder,hard_formal_questions,Questions with formal accuracy <= 0.90,12,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.7642276422764228,123,912645.6910569106,0.8373760483012759,0,0,0,0
+finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,1,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9927007299270073,1233,521407.69918699184,1.9038858296010628,1,0,0,0
+finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,2,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.991889699918897,1233,912645.6910569106,1.0868288862134616,0,0,0,0
+finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,3,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9902676399026764,1233,389499.8699186992,2.5424081402373155,1,1,1,1
+finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,4,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9894566098945661,1233,729460.4959349594,1.3564224730584842,0,0,0,0
+finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,5,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9886455798864558,1233,670960.9349593496,1.4734771107744944,0,0,0,0
 finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,6,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.9854014598540146,1233,546827.0650406504,1.8020349080211697,0,0,0,0
-finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,7,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9797242497972425,1233,397099.5447154472,2.467200637309447,0,0,0,0
+finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,7,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9797242497972425,1233,397627.9105691057,2.463922234218947,0,0,0,0
 finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,8,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9789132197891321,1233,404049.68292682926,2.422754579828261,0,0,0,0
 finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,9,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9772911597729116,1233,413189.837398374,2.3652352292262777,0,0,0,0
-finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,10,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.975669099756691,1233,409182.42276422767,2.384435512076908,0,0,0,0
-finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,11,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9651257096512571,1233,330733.1382113821,2.918140331720901,0,0,0,0
-finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,12,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9643146796431468,1233,638768.2926829269,1.50964706715932,0,0,0,0
-finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,1,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9739837398373984,615,469590.3333333333,2.0741136916590386,1,0,0,0
-finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,2,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9691056910569106,615,889263.8536585366,1.0897841929253003,0,0,0,0
-finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,3,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9691056910569106,615,133456.0243902439,7.26161067276447,1,1,0,0
-finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,4,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.967479674796748,615,116083.72357723577,8.334326682353877,1,0,1,1
-finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,5,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9658536585365853,615,389857.65040650405,2.477452109839299,0,0,0,0
+finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,10,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.975669099756691,1233,412001.3739837398,2.368120985429522,0,0,0,0
+finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,11,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9651257096512571,1233,329206.3739837398,2.931673825060649,1,0,0,0
+finder,implementation_signal_questions,Questions where formal three-trial agreement exceeds question-only by >= 0.05,12,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9643146796431468,1233,638067.8130081301,1.5113043786003664,0,0,0,0
+finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,1,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9739837398373984,615,521407.69918699184,1.8679887952481111,1,1,0,0
+finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,2,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9691056910569106,615,912645.6910569106,1.0618640952926817,0,0,0,0
+finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,3,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9691056910569106,615,729460.4959349594,1.3285238836885809,0,0,0,0
+finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,4,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.967479674796748,615,670960.9349593496,1.441931451426994,0,0,0,0
+finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,5,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9658536585365853,615,389499.8699186992,2.4797278077093816,1,0,1,1
 finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,6,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.9577235772357724,615,546827.0650406504,1.7514194860939747,0,0,0,0
 finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,7,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9479674796747968,615,404049.68292682926,2.3461656319291495,0,0,0,0
 finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,8,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9463414634146341,615,413189.837398374,2.2903309272396886,0,0,0,0
-finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,9,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9447154471544715,615,397099.5447154472,2.379039361103861,0,0,0,0
-finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,10,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.9382113821138212,615,409182.42276422767,2.2928926804229364,0,0,0,0
-finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,11,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.926829268292683,615,638768.2926829269,1.4509631722636933,0,0,0,0
-finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,12,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9252032520325203,615,330733.1382113821,2.797431358212413,0,0,0,0
-finder,top4_builders,Finder quality on top four builders,1,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9982905982905983,585,889263.8536585366,1.122603369273937,0,0,0,0
-finder,top4_builders,Finder quality on top four builders,2,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9982905982905983,585,469590.3333333333,2.125875529004923,1,0,0,0
+finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,9,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9447154471544715,615,397627.9105691057,2.375878106248492,0,0,0,0
+finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,10,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.9382113821138212,615,412001.3739837398,2.27720449823269,0,0,0,0
+finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,11,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.926829268292683,615,638067.8130081301,1.4525560597128468,0,0,0,0
+finder,low_prior_questions,Questions with question-only three-trial agreement < 0.90,12,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9252032520325203,615,329206.3739837398,2.8104050381424814,1,0,0,0
+finder,top4_builders,Finder quality on top four builders,1,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9982905982905983,585,912645.6910569106,1.0938424495649617,0,0,0,0
+finder,top4_builders,Finder quality on top four builders,2,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9982905982905983,585,521407.69918699184,1.9146065542323005,1,0,0,0
 finder,top4_builders,Finder quality on top four builders,3,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.9982905982905983,585,546827.0650406504,1.8256056843426116,0,0,0,0
-finder,top4_builders,Finder quality on top four builders,4,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9965811965811966,585,133456.0243902439,7.46748751983691,0,0,0,0
-finder,top4_builders,Finder quality on top four builders,5,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9965811965811966,585,116083.72357723577,8.585020930329874,1,1,1,1
-finder,top4_builders,Finder quality on top four builders,6,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9965811965811966,585,389857.65040650405,2.556269437170369,0,0,0,0
+finder,top4_builders,Finder quality on top four builders,4,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9965811965811966,585,729460.4959349594,1.366189399062474,0,0,0,0
+finder,top4_builders,Finder quality on top four builders,5,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9965811965811966,585,670960.9349593496,1.4853043517974334,0,0,0,0
+finder,top4_builders,Finder quality on top four builders,6,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9965811965811966,585,389499.8699186992,2.5586175337856063,1,1,1,0
 finder,top4_builders,Finder quality on top four builders,7,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9948717948717949,585,404049.68292682926,2.462251146109573,0,0,0,0
-finder,top4_builders,Finder quality on top four builders,8,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9948717948717949,585,397099.5447154472,2.505346098003457,0,0,0,0
+finder,top4_builders,Finder quality on top four builders,8,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9948717948717949,585,397627.9105691057,2.5020170074275794,0,0,0,0
 finder,top4_builders,Finder quality on top four builders,9,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9880341880341881,585,413189.837398374,2.3912354530675013,0,0,0,0
-finder,top4_builders,Finder quality on top four builders,10,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.9880341880341881,585,409182.42276422767,2.414654523426332,0,0,0,0
-finder,top4_builders,Finder quality on top four builders,11,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9829059829059829,585,638768.2926829269,1.5387519921779833,0,0,0,0
-finder,top4_builders,Finder quality on top four builders,12,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9811965811965812,585,330733.1382113821,2.966731989733267,0,0,0,0
-finder,top6_builders,Finder quality on top six builders,1,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9989417989417989,945,469590.3333333333,2.127262271032976,1,0,0,0
+finder,top4_builders,Finder quality on top four builders,10,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.9880341880341881,585,412001.3739837398,2.398133235529409,0,0,0,0
+finder,top4_builders,Finder quality on top four builders,11,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9829059829059829,585,638067.8130081301,1.5404412554084732,0,0,0,0
+finder,top4_builders,Finder quality on top four builders,12,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9811965811965812,585,329206.3739837398,2.9804908371702563,1,0,0,1
+finder,top6_builders,Finder quality on top six builders,1,claude_code_opus_4_7_low,claude,claude_code_opus_4_7,low,0.9989417989417989,945,521407.69918699184,1.91585548218679,1,0,0,0
 finder,top6_builders,Finder quality on top six builders,2,codex_gpt5_5_high,codex,codex_gpt5_5,high,0.9968253968253968,945,546827.0650406504,1.8229262239448485,0,0,0,0
-finder,top6_builders,Finder quality on top six builders,3,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9957671957671957,945,889263.8536585366,1.1197657384481465,0,0,0,0
-finder,top6_builders,Finder quality on top six builders,4,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9957671957671957,945,389857.65040650405,2.554181493498744,1,0,0,0
-finder,top6_builders,Finder quality on top six builders,5,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9947089947089947,945,133456.0243902439,7.453458914678349,0,0,0,0
-finder,top6_builders,Finder quality on top six builders,6,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9947089947089947,945,116083.72357723577,8.568892899504293,1,1,1,1
-finder,top6_builders,Finder quality on top six builders,7,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9915343915343915,945,397099.5447154472,2.4969416478301514,0,0,0,0
+finder,top6_builders,Finder quality on top six builders,3,claude_code_opus_4_7_high,claude,claude_code_opus_4_7,high,0.9957671957671957,945,912645.6910569106,1.091077518389447,0,0,0,0
+finder,top6_builders,Finder quality on top six builders,4,codex_gpt5_5_low,codex,codex_gpt5_5,low,0.9957671957671957,945,389499.8699186992,2.5565276722044694,1,1,1,0
+finder,top6_builders,Finder quality on top six builders,5,claude_code_sonnet_4_6_high,claude,claude_code_sonnet_4_6,high,0.9947089947089947,945,729460.4959349594,1.3636228421582481,0,0,0,0
+finder,top6_builders,Finder quality on top six builders,6,claude_code_sonnet_4_6_low,claude,claude_code_sonnet_4_6,low,0.9947089947089947,945,670960.9349593496,1.4825140226222848,0,0,0,0
+finder,top6_builders,Finder quality on top six builders,7,claude_code_mimo_v2_5_low,mimo,claude_code_mimo_v2_5,low,0.9915343915343915,945,397627.9105691057,2.4936237250429834,0,0,0,0
 finder,top6_builders,Finder quality on top six builders,8,claude_code_mimo_v2_5_high,mimo,claude_code_mimo_v2_5,high,0.9904761904761905,945,404049.68292682926,2.451372275066379,0,0,0,0
 finder,top6_builders,Finder quality on top six builders,9,claude_code_mimo_v2_5_pro_high,mimo,claude_code_mimo_v2_5_pro,high,0.9862433862433863,945,413189.837398374,2.3869013634343257,0,0,0,0
-finder,top6_builders,Finder quality on top six builders,10,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9841269841269841,945,330733.1382113821,2.975592314242177,0,0,0,0
-finder,top6_builders,Finder quality on top six builders,11,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9809523809523809,945,638768.2926829269,1.535693603125207,0,0,0,0
-finder,top6_builders,Finder quality on top six builders,12,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.9809523809523809,945,409182.42276422767,2.3973473110735477,0,0,0,0
+finder,top6_builders,Finder quality on top six builders,10,codex_gpt5_4_mini_low,codex,codex_gpt5_4_mini,low,0.9841269841269841,945,329206.3739837398,2.989392253309142,1,0,0,1
+finder,top6_builders,Finder quality on top six builders,11,codex_gpt5_4_mini_high,codex,codex_gpt5_4_mini,high,0.9809523809523809,945,638067.8130081301,1.5373795088138726,0,0,0,0
+finder,top6_builders,Finder quality on top six builders,12,claude_code_mimo_v2_5_pro_low,mimo,claude_code_mimo_v2_5_pro,low,0.9809523809523809,945,412001.3739837398,2.3809444407122182,0,0,0,0

runs_release/reports/eandd_final/pivots/pivot_insights.md

@@ -210,40 +210,44 @@ Diagnostic only; not the paper's within-finder builder-efficiency metric.
 
 | role | view | frontier agent | accuracy | mean tokens | acc / M tok |
 | --- | --- | --- | ---: | ---: | ---: |
-| builder | all_finders | claude_code_opus_4_7_high | 0.994 | 6418915 | 0.155 |
-| builder | all_finders | claude_code_opus_4_7_low | 0.993 | 4744350 | 0.209 |
+| builder | all_finders | claude_code_opus_4_7_high | 0.994 | 6565830 | 0.151 |
+| builder | all_finders | claude_code_opus_4_7_low | 0.993 | 4820855 | 0.206 |
 | builder | all_finders | codex_gpt5_5_high | 0.989 | 3216598 | 0.307 |
 | builder | all_finders | codex_gpt5_5_low | 0.987 | 1921738 | 0.513 |
 | builder | implementation_signal_questions | claude_code_mimo_v2_5_pro_low | 1.000 | 7278835 | 0.137 |
-| builder | implementation_signal_questions | claude_code_sonnet_4_6_low | 0.995 | 3280870 | 0.303 |
+| builder | implementation_signal_questions | claude_code_sonnet_4_6_low | 0.995 | 5092624 | 0.195 |
 | builder | implementation_signal_questions | codex_gpt5_4_mini_low | 0.992 | 2536506 | 0.391 |
 | builder | implementation_signal_questions | codex_gpt5_5_low | 0.983 | 1921738 | 0.511 |
 | builder | low_prior_questions | claude_code_mimo_v2_5_pro_low | 1.000 | 7278835 | 0.137 |
 | builder | low_prior_questions | codex_gpt5_4_mini_low | 0.993 | 2536506 | 0.391 |
 | builder | low_prior_questions | codex_gpt5_5_low | 0.965 | 1921738 | 0.502 |
 | builder | top4_formal_finders | claude_code_mimo_v2_5_pro_low | 1.000 | 7278835 | 0.137 |
-| builder | top4_formal_finders | claude_code_opus_4_7_low | 0.999 | 4744350 | 0.210 |
+| builder | top4_formal_finders | claude_code_opus_4_7_low | 0.999 | 4820855 | 0.207 |
 | builder | top4_formal_finders | codex_gpt5_5_high | 0.997 | 3216598 | 0.310 |
 | builder | top4_formal_finders | codex_gpt5_5_low | 0.990 | 1921738 | 0.515 |
-| finder | all_builders | claude_code_sonnet_4_6_low | 0.987 | 116084 | 8.502 |
-| finder | implementation_signal_questions | claude_code_opus_4_7_low | 0.993 | 469590 | 2.114 |
-| finder | implementation_signal_questions | codex_gpt5_5_low | 0.990 | 389858 | 2.540 |
-| finder | implementation_signal_questions | claude_code_sonnet_4_6_high | 0.989 | 133456 | 7.414 |
-| finder | implementation_signal_questions | claude_code_sonnet_4_6_low | 0.989 | 116084 | 8.517 |
-| finder | low_prior_questions | claude_code_opus_4_7_low | 0.974 | 469590 | 2.074 |
-| finder | low_prior_questions | claude_code_sonnet_4_6_high | 0.969 | 133456 | 7.262 |
-| finder | low_prior_questions | claude_code_sonnet_4_6_low | 0.967 | 116084 | 8.334 |
-| finder | top4_builders | claude_code_opus_4_7_low | 0.998 | 469590 | 2.126 |
-| finder | top4_builders | claude_code_sonnet_4_6_low | 0.997 | 116084 | 8.585 |
+| finder | all_builders | claude_code_sonnet_4_6_low | 0.987 | 670961 | 1.471 |
+| finder | all_builders | claude_code_opus_4_7_low | 0.979 | 521408 | 1.877 |
+| finder | all_builders | claude_code_mimo_v2_5_low | 0.968 | 397628 | 2.434 |
+| finder | all_builders | codex_gpt5_5_low | 0.965 | 389500 | 2.478 |
+| finder | all_builders | codex_gpt5_4_mini_low | 0.952 | 329206 | 2.891 |
+| finder | implementation_signal_questions | claude_code_opus_4_7_low | 0.993 | 521408 | 1.904 |
+| finder | implementation_signal_questions | codex_gpt5_5_low | 0.990 | 389500 | 2.542 |
+| finder | implementation_signal_questions | codex_gpt5_4_mini_low | 0.965 | 329206 | 2.932 |
+| finder | low_prior_questions | claude_code_opus_4_7_low | 0.974 | 521408 | 1.868 |
+| finder | low_prior_questions | codex_gpt5_5_low | 0.966 | 389500 | 2.480 |
+| finder | low_prior_questions | codex_gpt5_4_mini_low | 0.925 | 329206 | 2.810 |
+| finder | top4_builders | claude_code_opus_4_7_low | 0.998 | 521408 | 1.915 |
+| finder | top4_builders | codex_gpt5_5_low | 0.997 | 389500 | 2.559 |
+| finder | top4_builders | codex_gpt5_4_mini_low | 0.981 | 329206 | 2.980 |
 
 ## Low-Cost Near-Ceiling Points
 
 | role | view | band | agent | accuracy | mean tokens |
 | --- | --- | --- | --- | ---: | ---: |
-| builder | all_finders | within 0.5pp | claude_code_opus_4_7_low | 0.993 | 4744350 |
+| builder | all_finders | within 0.5pp | claude_code_opus_4_7_low | 0.993 | 4820855 |
 | builder | all_finders | within 1.0pp | codex_gpt5_5_low | 0.987 | 1921738 |
 | builder | all_finders | within 2.0pp | codex_gpt5_5_low | 0.987 | 1921738 |
-| builder | implementation_signal_questions | within 0.5pp | claude_code_sonnet_4_6_low | 0.995 | 3280870 |
+| builder | implementation_signal_questions | within 0.5pp | claude_code_sonnet_4_6_low | 0.995 | 5092624 |
 | builder | implementation_signal_questions | within 1.0pp | codex_gpt5_4_mini_low | 0.992 | 2536506 |
 | builder | implementation_signal_questions | within 2.0pp | codex_gpt5_5_low | 0.983 | 1921738 |
 | builder | low_prior_questions | within 0.5pp | claude_code_mimo_v2_5_pro_low | 1.000 | 7278835 |
@@ -252,18 +256,18 @@ Diagnostic only; not the paper's within-finder builder-efficiency metric.
 | builder | top4_formal_finders | within 0.5pp | codex_gpt5_5_high | 0.997 | 3216598 |
 | builder | top4_formal_finders | within 1.0pp | codex_gpt5_5_low | 0.990 | 1921738 |
 | builder | top4_formal_finders | within 2.0pp | codex_gpt5_5_low | 0.990 | 1921738 |
-| finder | all_builders | within 0.5pp | claude_code_sonnet_4_6_low | 0.987 | 116084 |
-| finder | all_builders | within 1.0pp | claude_code_sonnet_4_6_low | 0.987 | 116084 |
-| finder | all_builders | within 2.0pp | claude_code_sonnet_4_6_low | 0.987 | 116084 |
-| finder | implementation_signal_questions | within 0.5pp | claude_code_sonnet_4_6_low | 0.989 | 116084 |
-| finder | implementation_signal_questions | within 1.0pp | claude_code_sonnet_4_6_low | 0.989 | 116084 |
-| finder | implementation_signal_questions | within 2.0pp | claude_code_sonnet_4_6_low | 0.989 | 116084 |
-| finder | low_prior_questions | within 0.5pp | claude_code_sonnet_4_6_high | 0.969 | 133456 |
-| finder | low_prior_questions | within 1.0pp | claude_code_sonnet_4_6_low | 0.967 | 116084 |
-| finder | low_prior_questions | within 2.0pp | claude_code_sonnet_4_6_low | 0.967 | 116084 |
-| finder | top4_builders | within 0.5pp | claude_code_sonnet_4_6_low | 0.997 | 116084 |
-| finder | top4_builders | within 1.0pp | claude_code_sonnet_4_6_low | 0.997 | 116084 |
-| finder | top4_builders | within 2.0pp | claude_code_sonnet_4_6_low | 0.997 | 116084 |
+| finder | all_builders | within 0.5pp | claude_code_sonnet_4_6_low | 0.987 | 670961 |
+| finder | all_builders | within 1.0pp | claude_code_opus_4_7_low | 0.979 | 521408 |
+| finder | all_builders | within 2.0pp | claude_code_mimo_v2_5_low | 0.968 | 397628 |
+| finder | implementation_signal_questions | within 0.5pp | codex_gpt5_5_low | 0.990 | 389500 |
+| finder | implementation_signal_questions | within 1.0pp | codex_gpt5_5_low | 0.990 | 389500 |
+| finder | implementation_signal_questions | within 2.0pp | codex_gpt5_5_low | 0.990 | 389500 |
+| finder | low_prior_questions | within 0.5pp | claude_code_opus_4_7_low | 0.974 | 521408 |
+| finder | low_prior_questions | within 1.0pp | codex_gpt5_5_low | 0.966 | 389500 |
+| finder | low_prior_questions | within 2.0pp | codex_gpt5_5_low | 0.966 | 389500 |
+| finder | top4_builders | within 0.5pp | codex_gpt5_5_low | 0.997 | 389500 |
+| finder | top4_builders | within 1.0pp | codex_gpt5_5_low | 0.997 | 389500 |
+| finder | top4_builders | within 2.0pp | codex_gpt5_4_mini_low | 0.981 | 329206 |
 
 ## Family Pair Matrix
 

runs_release/reports/reproducibility_index.md

@@ -35,8 +35,8 @@ This index maps the paper's main results to the scripts, tables, and reports tha
 - Paper Table 2 formal: 97.5% three-trial agreement; 29 / 30 reliable questions (1476 find runs, 21312 answer rows)
 - Paper Table 2 spec_only: 99.7% three-trial agreement; 30 / 30 reliable questions (72 find runs, 1080 answer rows)
 - Paper Table 2 question_only: 94.4% three-trial agreement; 24 / 30 reliable questions (61 find runs, 915 answer rows)
-- Formal tokens: 934533272
-- Control tokens: 19189616
+- Formal tokens: 1095444301
+- Control tokens: 19311382
 - Formal cost USD: 616.00
 - Control cost USD: 19.42