Use descriptive validation labels in all dataset READMEs

Remove runner stage numbers, preserve counts and evidence, and clarify that ScaleSWE uses upstream Docker Hub images with optional Apptainer caches.

Calibforge_environment-reliable/README.md

@@ -21,10 +21,10 @@ Intended benchmark transfer: Terminal-Bench 2.0 and OpenThoughts-TBLite 2.0; dat
 
 ## Per-stage breakdown
 
-- **Rule-based checks (runner 1):** timeout/anti-cheating instructions restored on **5,317 tasks**; container-backed `task-absolute-path` review resolved **50 paths in 37 tasks**, clearing 35 warnings. `task-toml-schema` repaired **2** redundant resource declarations. `dockerfile-sanity`: **1 apt-tool pin repair**, freshly passed build and no-op validation. **0 additional archives**; unresolved pip pins and remaining path warnings are advisory.
-- **Build/setup (runner 3): 5,317 retained tasks passed.** **5,027 retained tasks repaired** for Docker-to-Apptainer working directories, variables, COPY, temporary files, heredocs, multi-stage builds and missing inputs. Verifier preparation repaired in **2,941 retained tasks**, including **2,813** with setup separated from grading. **57 archived:** 45 server/ownership, 7 temporary-state, 2 privileged-port and 3 repeated login-connection failures.
-- **Reference reward 1 (runner 4):** skipped; no reference solutions.
-- **No-op reward 0 (runner 5): 5,317 retained tasks passed.** **13 verifier repairs**, including missing-output assertions and dependency/service errors. **57 archived:** 4 false successes, 10 missing Python, 8 memory failures, 11 library failures, 8 native crashes, 7 timeouts, 4 missing pytest, 2 shell-script errors, 1 missing starting file and 2 permission failures.
+- **Rule-based checks:** timeout/anti-cheating instructions restored on **5,317 tasks**; container-backed `task-absolute-path` review resolved **50 paths in 37 tasks**, clearing 35 warnings. `task-toml-schema` repaired **2** redundant resource declarations. `dockerfile-sanity`: **1 apt-tool pin repair**, freshly passed build and no-op validation. **0 additional archives**; unresolved pip pins and remaining path warnings are advisory.
+- **Build and setup: 5,317 retained tasks passed.** **5,027 retained tasks repaired** for Docker-to-Apptainer working directories, variables, COPY, temporary files, heredocs, multi-stage builds and missing inputs. Verifier preparation repaired in **2,941 retained tasks**, including **2,813** with setup separated from grading. **57 archived:** 45 server/ownership, 7 temporary-state, 2 privileged-port and 3 repeated login-connection failures.
+- **Reference solution (reward 1):** skipped; no reference solutions.
+- **No-op baseline (reward 0): 5,317 retained tasks passed.** **13 verifier repairs**, including missing-output assertions and dependency/service errors. **57 archived:** 4 false successes, 10 missing Python, 8 memory failures, 11 library failures, 8 native crashes, 7 timeouts, 4 missing pytest, 2 shell-script errors, 1 missing starting file and 2 permission failures.
 
 Repair counts overlap. Results combine historical full runs with targeted checks of restored repairs.
 

EndlessTerminals_environment-reliable/README.md

@@ -21,10 +21,10 @@ Intended benchmark transfer: **Terminal-Bench 2.0 and OpenThoughts-TBLite 2.0**,
 
 ## Per-stage breakdown
 
-- **Rule-based review (runner 1):** restored timeout/anti-cheating suffixes for **2,144 retained tasks**; **1 archived for `reference-solution-integrity`**, a fake YAML module accepted by version-only grading. **3 retained truncated prompts** had already been restored from source descriptions.
-- **Build/setup (runner 3): 2,144 retained tasks passed.** All **2,492 original contexts** built successfully. Retained repairs: **1,721** build compatibility, **419** shared dependencies/setup, **4** startup compatibility; all **2,144** use offline verification. **9** timestamp repairs overlap. **271 archived** for reviewed identity, ownership or permission mismatches. Conversion review covered **435 tasks**, without a Docker reference run.
-- **Reference reward 1 (runner 4): 2,144 retained tasks passed.** Repairs include pinned pip installs for **8 tasks**, CPU handling for **4**, and subprocess PATH for **1**; **53 archived** after build exclusions.
-- **No-op reward 0 (runner 5): 2,144 retained tasks passed.** **23 archived:** 22 false successes and 1 verifier repeatedly writing reward 1 without tests.
+- **Rule-based checks:** restored timeout/anti-cheating suffixes for **2,144 retained tasks**; **1 archived for `reference-solution-integrity`**, a fake YAML module accepted by version-only grading. **3 retained truncated prompts** had already been restored from source descriptions.
+- **Build and setup: 2,144 retained tasks passed.** All **2,492 original contexts** built successfully. Retained repairs: **1,721** build compatibility, **419** shared dependencies/setup, **4** startup compatibility; all **2,144** use offline verification. **9** timestamp repairs overlap. **271 archived** for reviewed identity, ownership or permission mismatches. Conversion review covered **435 tasks**, without a Docker reference run.
+- **Reference solution (reward 1): 2,144 retained tasks passed.** Repairs include pinned pip installs for **8 tasks**, CPU handling for **4**, and subprocess PATH for **1**; **53 archived** after build exclusions.
+- **No-op baseline (reward 0): 2,144 retained tasks passed.** **23 archived:** 22 false successes and 1 verifier repeatedly writing reward 1 without tests.
 
 Repair counts overlap; each archive is counted once. Runtime results reuse historical full runs and repair pilots. One HTTP task retains an unresolved possible shared-port collision.
 

TermiGen_environment-reliable/README.md

@@ -26,10 +26,10 @@ Exploration, Error Recovery and Specification Adherence; specialized tasks also
 
 Counts cover retained tasks; repair categories overlap. Archives are counted at their first failing stage.
 
-- **1, rule-based checks:** 3,336 timeout/anti-cheating instruction repairs (`check-instruction-suffix.sh`); 141 archived: 135 relative instruction paths (`check-task-absolute-path.sh`), 3 host-directory binds (`check-compose-host-binds.sh`), 3 host CPU-count calls (`check-nproc.sh`). These preserve recorded exclusions; file-reference and pip-pin findings were advisory.
-- **3, build/setup:** 2,060 repaired for Apptainer and the Marin Harbor bridge with repository patches: 1,911 COPY destinations/renames, 149 files hidden by temporary mounts, 28 environment-variable expansions, 2 initial-file repairs, 1 missing build helper, 1 wildcard COPY. All 3,336 retained tasks passed. Another 54 archived: 12 missing/changed runtime files, 12 dependency failures, 9 incomplete checks, 7 startup failures, 6 absent COPY inputs, 3 image-preparation failures, 2 COPY collisions, 2 unsupported sidecars, 1 unavailable build command.
-- **4, reference reward 1:** skipped; no reference solutions.
-- **5, no-op reward 0:** 3,336 passed; 35 archived: 2 rewarded no work, 28 ran no tests, 5 timed out. No verifier repairs.
+- **Rule-based checks:** 3,336 timeout/anti-cheating instruction repairs (`check-instruction-suffix.sh`); 141 archived: 135 relative instruction paths (`check-task-absolute-path.sh`), 3 host-directory binds (`check-compose-host-binds.sh`), 3 host CPU-count calls (`check-nproc.sh`). These preserve recorded exclusions; file-reference and pip-pin findings were advisory.
+- **Build and setup:** 2,060 repaired for Apptainer and the Marin Harbor bridge with repository patches: 1,911 COPY destinations/renames, 149 files hidden by temporary mounts, 28 environment-variable expansions, 2 initial-file repairs, 1 missing build helper, 1 wildcard COPY. All 3,336 retained tasks passed. Another 54 archived: 12 missing/changed runtime files, 12 dependency failures, 9 incomplete checks, 7 startup failures, 6 absent COPY inputs, 3 image-preparation failures, 2 COPY collisions, 2 unsupported sidecars, 1 unavailable build command.
+- **Reference solution (reward 1):** skipped; no reference solutions.
+- **No-op baseline (reward 0):** 3,336 passed; 35 archived: 2 rewarded no work, 28 ran no tests, 5 timed out. No verifier repairs.
 
 Uses [historical checks and reruns](https://huggingface.co/datasets/FWeindel/validated-tasks/blob/44b082b9ad564790321d6a569c8531ed14a3393f/runs/termigen-phases-1-3-5-6-7-preview-20261014.json). Teacher results did not determine retention.
 

bugsinpy-original_environment-reliable/README.md

@@ -25,12 +25,12 @@ Real Python repositories at buggy commits, upstream regression tests and referen
 
 ## Per-stage breakdown
 
-Counts overlap. [Historical evidence](https://huggingface.co/datasets/FWeindel/validated-tasks/blob/579dbfd3edf67734e7725129c62dcf5975945976/runs/2026-10-08-bugsinpy-original-fc23c4eacb04.json): stages 3/4/5 passed before instruction review; stage 1 reran afterwards.
+Counts overlap. [Historical evidence](https://huggingface.co/datasets/FWeindel/validated-tasks/blob/579dbfd3edf67734e7725129c62dcf5975945976/runs/2026-10-08-bugsinpy-original-fc23c4eacb04.json): build, reference and no-op checks passed before instruction review; rule-based checks reran afterwards.
 
-- **1, rule-based/static:** 489 retain the timeout/anti-cheating suffix required by `check-instruction-suffix.sh` (427 carried the historical `anti-cheat-instruction` change label). **0 static archives**; 7 recorded `warning:relative-instruction-path` findings remained warnings. The contract excluded 13 checks. Separate instruction review preserved **462 rewrites** and archived **6** for tests requiring specific implementations (5) or testing a different issue (1), recorded as `flagged_by_instruction_loop`, not static-check failures.
-- **3, build/setup:** **489 converted** to shared Apptainer-compatible environments with pinned dependencies; **247** include compiler/system libraries. Shared Marin/Harbor bridge fixes addressed container startup, uploads and certificate mounts. **1 archived** (`environment_task_mismatch`): YouTube-DL 40 requires older-Python behavior absent from its recorded environment.
-- **4, reference reward 1:** **489 passed**. Clean verifier dependency installs for **489**; source-timestamp fixes for **66**, plus **169 pandas** native rebuilds. **1 archived** (`flagged_by_instruction_loop`): HTTPie 4 depends on live httpbin.org availability.
-- **5, no-op reward 0:** **489 passed**. **6 verifier repairs:** Ansible CPU regression (1), missing-function assertions moved inside tests (4), offline YouTube-DL chapter checks (1). **3 archived** (`buggy-version-passes-verifier`): Keras 15 and Scrapy 7/11 passed without a fix.
+- **Rule-based checks:** 489 retain the timeout/anti-cheating suffix required by `check-instruction-suffix.sh` (427 carried the historical `anti-cheat-instruction` change label). **0 static archives**; 7 recorded `warning:relative-instruction-path` findings remained warnings. The contract excluded 13 checks. Separate instruction review preserved **462 rewrites** and archived **6** for tests requiring specific implementations (5) or testing a different issue (1), recorded as `flagged_by_instruction_loop`, not static-check failures.
+- **Build and setup:** **489 converted** to shared Apptainer-compatible environments with pinned dependencies; **247** include compiler/system libraries. Shared Marin/Harbor bridge fixes addressed container startup, uploads and certificate mounts. **1 archived** (`environment_task_mismatch`): YouTube-DL 40 requires older-Python behavior absent from its recorded environment.
+- **Reference solution (reward 1):** **489 passed**. Clean verifier dependency installs for **489**; source-timestamp fixes for **66**, plus **169 pandas** native rebuilds. **1 archived** (`flagged_by_instruction_loop`): HTTPie 4 depends on live httpbin.org availability.
+- **No-op baseline (reward 0):** **489 passed**. **6 verifier repairs:** Ansible CPU regression (1), missing-function assertions moved inside tests (4), offline YouTube-DL chapter checks (1). **3 archived** (`buggy-version-passes-verifier`): Keras 15 and Scrapy 7/11 passed without a fix.
 
 All 11 archives were recorded as stage-0 conversion/review exclusions; the bullets group their reasons by relevant stage.
 
@@ -38,4 +38,4 @@ All 11 archives were recorded as stage-0 conversion/review exclusions; the bulle
 
 Historical final static run: Helma, x86_64, **48 CPUs / 128G**, static concurrency **24**; Python **3.12.14**, Harbor **0.8.1**, Apptainer **1.5.4-1.el9**. Harbor-validation commit `7faf878c14b6`.
 
-**Validation:** stage 1 wall time **7.3 s / 489 tasks**; per-task median **0.3 s**. **Not recorded:** mean task time, aggregate stages 3/4/5 time, total/per-task patching time, separate Marin version.
+**Validation:** rule-based check wall time **7.3 s / 489 tasks**; per-task median **0.3 s**. **Not recorded:** mean task time, aggregate build/reference/no-op time, total/per-task patching time, separate Marin version.

camel-ai__SETA-Env_environment-reliable/README.md

@@ -21,10 +21,10 @@ Intended benchmark transfer: **Terminal-Bench 2.0, OpenThoughts-TBLite 2.0, DS-1
 
 ## Per-stage breakdown
 
-- **Rule-based checks (runner 1):** **25 retained instructions clarified**, plus timeout/anti-cheating suffixes on **3,107**; historical dependency-pin repairs retained. **43 historical `check-test-file-references.sh` flags reviewed: 19 archived** for undisclosed deliverable paths; **24 retained** because paths were discoverable or tokens were shell syntax. Reviewed `nproc` commands were retained after allocation probes; absolute-path checks were excluded from the original full run and used only in later subsets.
-- **Build/setup (runner 3): 3,107 retained tasks passed.** Retained-task repairs: **215 environments, 72 setups, 3,021 verifiers and 35 configurations**, including Docker-to-Apptainer adaptation, dependency pins and resource/time limits. **24 archived:** 19 Rocky Linux ownership mismatches, 2 read-only NVIDIA-tool conflicts and 3 slow R setups.
-- **Reference reward 1 (runner 4): 3,107 retained tasks passed.** **6 repairs** to memory, filename matching, PHP requests, reference startup/logging and setup timeout; **1 archived** for slow desktop setup.
-- **No-op reward 0 (runner 5): 3,107 retained tasks passed.** **32 repairs:** 31 setup-timeout increases and 1 missing-backup-script check; **2 archived** for slow R setup.
+- **Rule-based checks:** **25 retained instructions clarified**, plus timeout/anti-cheating suffixes on **3,107**; historical dependency-pin repairs retained. **43 historical `check-test-file-references.sh` flags reviewed: 19 archived** for undisclosed deliverable paths; **24 retained** because paths were discoverable or tokens were shell syntax. Reviewed `nproc` commands were retained after allocation probes; absolute-path checks were excluded from the original full run and used only in later subsets.
+- **Build and setup: 3,107 retained tasks passed.** Retained-task repairs: **215 environments, 72 setups, 3,021 verifiers and 35 configurations**, including Docker-to-Apptainer adaptation, dependency pins and resource/time limits. **24 archived:** 19 Rocky Linux ownership mismatches, 2 read-only NVIDIA-tool conflicts and 3 slow R setups.
+- **Reference solution (reward 1): 3,107 retained tasks passed.** **6 repairs** to memory, filename matching, PHP requests, reference startup/logging and setup timeout; **1 archived** for slow desktop setup.
+- **No-op baseline (reward 0): 3,107 retained tasks passed.** **32 repairs:** 31 setup-timeout increases and 1 missing-backup-script check; **2 archived** for slow R setup.
 
 Repair counts overlap. Results combine historical runs with a fresh rerun of the **3 restored tasks whose verifier files changed**; all three passed build, reference 1 and no-op 0.
 

crosscodeeval_environment-reliable/README.md

@@ -26,10 +26,10 @@ Complete the next statement in a real C#, Java, Python or TypeScript repository
 
 Counts cover retained tasks; repair categories overlap.
 
-- **1, rule-based checks:** 6,710 retain the timeout/anti-cheating instruction (`instruction-suffix`) and context guidance; 0 archived by this validation stage. Patcher rules separately archived 1,043 tasks with required identifiers absent from available context and 10 whose scored answer contained no identifiers. All 6,710 retain restored cross-file excerpts with answer leakage removed.
-- **3, build/setup:** 6,710 received pinned verifier dependencies; 6,710 passed, 0 archived. Docker definitions were converted to Apptainer for the Marin Harbor bridge and repository runtime patches.
-- **4, reference reward 1:** reference solutions added to all 6,710; all passed, 0 archived.
-- **5, no-op reward 0:** scoring repaired in all 6,710 to reject incomplete prefix matches and normalize comments/whitespace; all passed, 0 archived.
+- **Rule-based checks:** 6,710 retain the timeout/anti-cheating instruction (`instruction-suffix`) and context guidance; 0 archived by this validation stage. Patcher rules separately archived 1,043 tasks with required identifiers absent from available context and 10 whose scored answer contained no identifiers. All 6,710 retain restored cross-file excerpts with answer leakage removed.
+- **Build and setup:** 6,710 received pinned verifier dependencies; 6,710 passed, 0 archived. Docker definitions were converted to Apptainer for the Marin Harbor bridge and repository runtime patches.
+- **Reference solution (reward 1):** reference solutions added to all 6,710; all passed, 0 archived.
+- **No-op baseline (reward 0):** scoring repaired in all 6,710 to reject incomplete prefix matches and normalize comments/whitespace; all passed, 0 archived.
 
 Results are from the [historical validation](https://huggingface.co/datasets/FWeindel/validated-tasks/blob/b8a56dc21aa9f78340048a367e70852c59d01ff8/runs/2026-10-03-c536bae24db8.json); retained task payloads are unchanged.
 
@@ -40,7 +40,7 @@ Helma, x86_64, **48 CPUs / 128G**, no GPUs, concurrency 16; Python **3.12.14**,
 | Historical validation stage | Wall time | Mean per task (6,710 tasks) |
 | --- | ---: | ---: |
 | Static checks | 1,528.5 s | 3.36 s |
-| Build/setup | 6,938.0 s | 16.50 s |
+| Build and setup | 6,938.0 s | 16.50 s |
 | Reference | 6,061.3 s | 14.39 s |
 | No-op | 6,468.8 s | 15.35 s |
 

scaleswe_environment-reliable/README.md

@@ -5,7 +5,7 @@
 - **[17,202 original tasks](https://huggingface.co/datasets/PrimeIntellect/Scale-SWE-Verified/tree/8935f8e55244fd56080cdb8dcd0819a57e8a003c); 16,786 retained, all changed; 416 archived.**
 - **17,202 original Docker images → 16,786 retained images.** No consolidation; the decrease follows task exclusions. Exact tags remain in each Dockerfile, under [aweaiteam/scaleswe](https://hub.docker.com/r/aweaiteam/scaleswe/tags).
 
-- **48 cached Apptainer bundles** accompany this release; the other **16,738** retained tasks use their upstream Docker images.
+- All retained tasks use the original Docker Hub image tags; **48 optional cached Apptainer conversions** are also available on HF. Duplicating the remaining upstream images on HF is unnecessary.
 
 ## Dataset characteristics
 
@@ -29,10 +29,10 @@ Exploration; Error Recovery; Specification Adherence; Constraint Satisfaction. T
 
 Counts below describe retained repairs; repair groups overlap. Results combine historical validation and repair reruns; the later full reproduction failed during startup.
 
-- **1, rule-based/static:** 16,786 changed, 0 archived. Restored the timeout/anti-cheating sentence (`check-instruction-suffix.sh`) and 750 recorded instruction-path corrections supported by setup/reference observations. The historical `check-task-absolute-path.sh` treated unresolved relative paths as advisory; these corrections do not imply 750 checker failures.
-- **3, build/setup:** 16,786 passed; 0 dataset repairs, 0 archived. Original per-task images ran through the shared Apptainer/Marin Harbor bridge.
-- **4, reference reward 1:** 16,786 passed; 419 repaired, 362 archived. Repairs: 367 missing test fixtures, 26 pytest file-handle conflicts, 6 worker limits, 5 import paths, 5 parameterized-test matching, 5 always-skipped required tests, 4 worker cleanup and 1 temporary-file cleanup. Archives: 149 failing reference tests, 122 unavailable services, 60 missing/incompatible dependencies, 18 stale HTTP recordings, 7 privileged ports, 4 import-path conflicts and 2 missing test definitions/helpers.
-- **5, no-op reward 0:** 16,786 passed; 37 repaired, 54 archived. Repairs: 31 checkout imports, 5 deferred missing-feature imports, 1 fixture/worker-timeout repair. Archives: 37 passing without a fix, 15 test-only pull requests, 1 timeout and 1 timing-sensitive result.
+- **Rule-based checks:** 16,786 changed, 0 archived. Restored the timeout/anti-cheating sentence (`check-instruction-suffix.sh`) and 750 recorded instruction-path corrections supported by setup/reference observations. The historical `check-task-absolute-path.sh` treated unresolved relative paths as advisory; these corrections do not imply 750 checker failures.
+- **Build and setup:** 16,786 passed; 0 dataset repairs, 0 archived. Original per-task images ran through the shared Apptainer/Marin Harbor bridge.
+- **Reference solution (reward 1):** 16,786 passed; 419 repaired, 362 archived. Repairs: 367 missing test fixtures, 26 pytest file-handle conflicts, 6 worker limits, 5 import paths, 5 parameterized-test matching, 5 always-skipped required tests, 4 worker cleanup and 1 temporary-file cleanup. Archives: 149 failing reference tests, 122 unavailable services, 60 missing/incompatible dependencies, 18 stale HTTP recordings, 7 privileged ports, 4 import-path conflicts and 2 missing test definitions/helpers.
+- **No-op baseline (reward 0):** 16,786 passed; 37 repaired, 54 archived. Repairs: 31 checkout imports, 5 deferred missing-feature imports, 1 fixture/worker-timeout repair. Archives: 37 passing without a fix, 15 test-only pull requests, 1 timeout and 1 timing-sensitive result.
 
 [Historical results and per-task reasons](https://huggingface.co/datasets/FWeindel/validated-tasks/blob/e77e80246668e100171168fa369012fd68f3d53a/runs/2026-10-14-scaleswe-historical-audit.json).
 

swelego-real-data-verified_environment-reliable/README.md

@@ -27,10 +27,10 @@ Exploration; Error Recovery; Specification Adherence; Constraint Satisfaction. L
 
 Counts below describe retained repairs; repair groups overlap. Results are historical, with the validated instructions restored.
 
-- **1, rule-based/static:** 4,277 changed, 0 archived. Restored the timeout/anti-cheating sentence required by `check-instruction-suffix.sh`; no additional recorded automatic path or pip-pin repairs. Existing `/testbed` and setup instructions remain.
-- **3, build/setup:** 4,277 changed, 1 archived. Shared images, pinned dependencies and Apptainer/Marin Harbor bridge compatibility; 736 retained tasks also needed missing build dependencies. Archived 1 unavailable repository.
-- **4, reference reward 1:** 4,277 passed; 55 repaired, 42 archived. Repairs: 19 missing history/metadata/dependencies, 13 worker limits, 6 memory limits, 6 legacy SymPy helpers, 5 mocked/recorded HTTP settings, 3 astronomy time tables, and 1 each for a fixed test date, CPU affinity and Python import paths. Archives: 23 proxy-sensitive tests, 8 unreliable services, 6 dependency mismatches, 2 version-sensitive expectations, and 1 each for system headers, expired licence and excessive memory.
-- **5, no-op reward 0:** 4,277 passed; 3 repaired, 3 archived. Repairs: defer a missing-feature import, test the buggy Python backend, freeze a date-sensitive test. Archived 2 tasks passing without a fix and 1 verifier usage error.
+- **Rule-based checks:** 4,277 changed, 0 archived. Restored the timeout/anti-cheating sentence required by `check-instruction-suffix.sh`; no additional recorded automatic path or pip-pin repairs. Existing `/testbed` and setup instructions remain.
+- **Build and setup:** 4,277 changed, 1 archived. Shared images, pinned dependencies and Apptainer/Marin Harbor bridge compatibility; 736 retained tasks also needed missing build dependencies. Archived 1 unavailable repository.
+- **Reference solution (reward 1):** 4,277 passed; 55 repaired, 42 archived. Repairs: 19 missing history/metadata/dependencies, 13 worker limits, 6 memory limits, 6 legacy SymPy helpers, 5 mocked/recorded HTTP settings, 3 astronomy time tables, and 1 each for a fixed test date, CPU affinity and Python import paths. Archives: 23 proxy-sensitive tests, 8 unreliable services, 6 dependency mismatches, 2 version-sensitive expectations, and 1 each for system headers, expired licence and excessive memory.
+- **No-op baseline (reward 0):** 4,277 passed; 3 repaired, 3 archived. Repairs: defer a missing-feature import, test the buggy Python backend, freeze a date-sensitive test. Archived 2 tasks passing without a fix and 1 verifier usage error.
 
 [Historical results and per-task reasons](https://huggingface.co/datasets/FWeindel/validated-tasks/blob/65cc78ebebb1201ab7bd4590cd359d55bc57a912/runs/2026-10-11-swelego-real-data-verified-d15f2a1ef8c8.json).
 
@@ -40,9 +40,9 @@ Image builds: **4 CPUs each**, 98 at **12,288 MiB**, 3 at **8,192 MiB**. Histori
 
 | Historical validation stage | Stage wall time | Median task time | Timed tasks |
 | --- | ---: | ---: | ---: |
-| 1 | 104.7 s | 0.8 s | 4,281 |
-| 3 | 9,131.6 s | 70.1 s | 4,281 |
-| 4 | 9,280.0 s | 56.5 s | 4,280 |
-| 5 | 9,558.0 s | 52.6 s | 4,280 |
+| Rule-based checks | 104.7 s | 0.8 s | 4,281 |
+| Build and setup | 9,131.6 s | 70.1 s | 4,281 |
+| Reference solution | 9,280.0 s | 56.5 s | 4,280 |
+| No-op baseline | 9,558.0 s | 52.6 s | 4,280 |
 
 Patching time and mean task time: **Not recorded**. Timings describe the historical full run, including tasks subsequently archived.