Validation run 2026-10-03-c536bae24db8

crosscodeeval-csharp-v5/README.md

@@ -1,12 +1,32 @@
-# CrossCodeEval, C#
+# crosscodeeval-csharp-v5
 
-Cross-file code completion. The agent sees a file that is cut at a cursor and must write the next statement. That statement depends on code in other files of the same repository, which the task provides as context files.
+## Original datasource
 
-| | |
-| --- | --- |
-| Tasks before validation | 1,353 |
-| Underlying benchmark | [amazon-science/cceval](https://github.com/amazon-science/cceval), commit `40c68d2b` |
-| Packaged source | [TaskTrove](https://huggingface.co/datasets/open-thoughts/TaskTrove), `laion__exp_rpt_crosscodeeval-csharp-v5`, revision `12df4483fe99`, from [discussion 3](https://huggingface.co/datasets/open-thoughts/TaskTrove/discussions/3) |
-| Reward | 1 if the first statement matches the reference exactly, after removing comments and surrounding whitespace; otherwise 0 |
+- CrossCodeEval (C# split) — https://arxiv.org/abs/2310.11248
 
-The tasks were patched before validation: the cross-file context and the benchmark's own scorer were added, and tasks whose reference cannot be derived from the context were dropped. The changes are listed in the [CrossCodeEval patch notes](https://github.com/MLI-lab/OT-next-data/blob/main/data/crosscodeeval/README.md).
+## Domains
+
+Software Engineering
+
+- **Software Engineering:** Every sampled task is cross-file code completion in a real C# repository (Unity mods, ASP.NET extensions, WPF samples): the agent must produce the next statement of a cut-off declaration, consistent with types, attributes, and method signatures defined in sibling files supplied under /setup_files/context.
+
+## Agent capabilities
+
+Exploration, Specification Adherence
+
+- **Exploration:** Intended practice: the correct continuation (e.g. 'LogLevel> LootingLogLevels;', 'StyleSheetName => nameof(SceneSelectorWindow);', 'LogdashboardAccountAuthorizeFilter filter, ...') depends on identifiers declared only in the retrieved cross-file excerpts, so the agent is expected to read those files before answering. No trajectories were supplied, so observed behavior cannot be confirmed.
+- **Specification Adherence:** Intended practice: the verifier (a port of the official cceval metric) truncates the prediction to its first statement and requires an exact line-by-line match with the reference. The instruction further requires writing only the short fragment to /app/solution.txt via a quote-safe helper, with no re-emitted prefix, usings, code fences, or labels. Output drift of any kind fails the task. Not observed in trajectories since none were supplied.
+
+## Expected benchmark transfer
+
+These are hypotheses, not measured training gains.
+
+The tasks are single-statement C# completions scored by exact string match against a fixed reference, with no compilation or execution. The suite's code benchmarks are Python-centric and execution-verified (HumanEval+, MBPP+, DS-1000, CRUXEval, SWE-bench Verified), and the terminal benchmarks (Terminal-Bench 2.0, OpenThoughts-TBLite) require multi-step shell work, whereas these tasks reduce to reading a few context files and writing one line. The language, output format, and verification signal are too far from any suite benchmark to reasonably expect a score increase.
+
+## Files
+
+- `tasks.parquet`: kept tasks.
+- `archive.parquet`: excluded tasks with validation reasons.
+
+Generated with claude-fable-5-1; domain and capability labels follow CLI-Universe.
+Taxonomy source: https://arxiv.org/abs/2606.22883v1

crosscodeeval-csharp-v5/archive.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb7fdfb21a32fec9ed1f4012f47dd0d35e41caddc3ae5a3ca47a466b66caf9ed
+size 1276

crosscodeeval-csharp-v5/tasks.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:43d6e9c4e15ef17c7f3eb6fb90a153cd74d52ead9b91f27e1fd8272029227275
+size 13830918

crosscodeeval-java-v4/README.md

@@ -1,12 +1,32 @@
-# CrossCodeEval, Java
+# crosscodeeval-java-v4
 
-Cross-file code completion. The agent sees a file that is cut at a cursor and must write the next statement. That statement depends on code in other files of the same repository, which the task provides as context files.
+## Original datasource
 
-| | |
-| --- | --- |
-| Tasks before validation | 1,895 |
-| Underlying benchmark | [amazon-science/cceval](https://github.com/amazon-science/cceval), commit `40c68d2b` |
-| Packaged source | [TaskTrove](https://huggingface.co/datasets/open-thoughts/TaskTrove), `laion__exp_rpt_crosscodeeval-java-v4`, revision `12df4483fe99`, from [discussion 3](https://huggingface.co/datasets/open-thoughts/TaskTrove/discussions/3) |
-| Reward | 1 if the first statement matches the reference exactly, after removing comments and surrounding whitespace; otherwise 0 |
+- CrossCodeEval (Java split) — https://arxiv.org/abs/2310.11248
 
-The tasks were patched before validation: the cross-file context and the benchmark's own scorer were added, and tasks whose reference cannot be derived from the context were dropped. The changes are listed in the [CrossCodeEval patch notes](https://github.com/MLI-lab/OT-next-data/blob/main/data/crosscodeeval/README.md).
+## Domains
+
+Software Engineering
+
+- **Software Engineering:** Every sampled task asks the agent to complete the next Java statement in real open-source application code (CLI shells, Spring controllers, Android services, JDBC queries, Minecraft mod renderers), using in-file context plus retrieved excerpts from other repository files. The reference is the project's actual next line, e.g. a call to a project-specific API such as Contrast.ratioOfTones or commands.createRequest.
+
+## Agent capabilities
+
+Exploration, Specification Adherence
+
+- **Exploration:** Intended practice: the instruction points to read-only cross-file excerpts under /setup_files/context (4-5 chunked files per task, 1-2.5 KB total) and the correct continuation usually depends on identifiers only visible there (method names, field accessors, constants). The agent must inspect these files before answering. No trajectories were supplied, so observed behavior cannot be confirmed.
+- **Specification Adherence:** Intended practice: the output contract requires writing only the missing fragment to /app/solution.txt with no fences, repeated context, or prose, and the verifier (a port of the official cceval metric) awards reward only for exact match of the first statement after comment removal and whitespace normalization. Any extra tokens or formatting drift fails the task. Not verified from trajectories.
+
+## Expected benchmark transfer
+
+These are hypotheses, not measured training gains.
+
+- **SWE-bench Verified random-100**: Both are Harbor-style sandbox tasks run by a terminal agent that must read repository files and produce code consistent with project-specific APIs. Practicing cross-file API inference and precise code edits in an agentic file-writing format should modestly help repository-grounded repair, though the transfer is limited by the language gap (Java vs Python) and the much shorter horizon of single-statement completion.
+
+## Files
+
+- `tasks.parquet`: kept tasks.
+- `archive.parquet`: excluded tasks with validation reasons.
+
+Generated with claude-fable-5-1; domain and capability labels follow CLI-Universe.
+Taxonomy source: https://arxiv.org/abs/2606.22883v1

crosscodeeval-java-v4/archive.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb7fdfb21a32fec9ed1f4012f47dd0d35e41caddc3ae5a3ca47a466b66caf9ed
+size 1276

crosscodeeval-java-v4/tasks.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49fd76d0f5db7f58ecd5cd736499d2f9681fe9d773b76f8d1417559db23b1ca0
+size 18519696

crosscodeeval-python-v3/README.md

@@ -1,12 +1,35 @@
-# CrossCodeEval, Python
+# crosscodeeval-python-v3
 
-Cross-file code completion. The agent sees a file that is cut at a cursor and must write the next statement. That statement depends on code in other files of the same repository, which the task provides as context files.
+## Original datasource
 
-| | |
-| --- | --- |
-| Tasks before validation | 432 |
-| Underlying benchmark | [amazon-science/cceval](https://github.com/amazon-science/cceval), commit `40c68d2b` |
-| Packaged source | [TaskTrove](https://huggingface.co/datasets/open-thoughts/TaskTrove), `laion__exp_rpt_crosscodeeval-python-v3`, revision `12df4483fe99`, from [discussion 3](https://huggingface.co/datasets/open-thoughts/TaskTrove/discussions/3) |
-| Reward | 1 if the first statement matches the reference exactly, after removing comments and surrounding whitespace; otherwise 0 |
+- CrossCodeEval — https://arxiv.org/abs/2310.11248
 
-The tasks were patched before validation: the cross-file context and the benchmark's own scorer were added, and tasks whose reference cannot be derived from the context were dropped. The changes are listed in the [CrossCodeEval patch notes](https://github.com/MLI-lab/OT-next-data/blob/main/data/crosscodeeval/README.md).
+## Domains
+
+Software Engineering
+
+- **Software Engineering:** Every sampled task is Python code completion inside a real repository file: the prompt ends mid-statement (e.g. `assert client.`, `grad_norm = utils.`, `subtask_ids: List[str] = config.`) and the agent must write the next statement, typically an API call or attribute access whose name and signature are defined in other repository files supplied as retrieved cross-file context under /setup_files/context/. The verifier is a port of the cceval metric code and rewards exact match of the first completed statement after comment removal and whitespace normalization.
+
+## Agent capabilities
+
+Exploration, Specification Adherence
+
+- **Exploration:** Intended practice: the correct continuation depends on identifiers that are not visible in the prompt (e.g. `add_rand_var(...)`, `get_grad_norm_`, `smc_steer`, `bet_amount`/`raise_limit`), so the agent is expected to open the 4-5 retrieved context excerpts in /setup_files/context/ to discover the right API names and argument conventions before answering. No trajectories were supplied, so this is inferred from the task design rather than observed behavior.
+- **Specification Adherence:** Intended practice: the instruction imposes a strict output contract (write only the short continuation to /app/solution.txt via `write_solution`, no re-emitted snippet, no imports, fences, or labels) and the verifier scores exact match of the first statement, so any formatting drift or extra text yields reward 0. Edit similarity and identifier F1 are logged but do not affect the pass/fail reward. Inferred from task design; no trajectories were supplied.
+
+## Expected benchmark transfer
+
+These are hypotheses, not measured training gains.
+
+- **HumanEval+**: Both require producing syntactically and semantically correct Python that completes a partial program; practice at emitting exact, idiomatic Python continuations with correct argument lists and attribute chains should modestly help zero-shot Python function completion.
+- **MBPP+**: Same Python code-writing skill as above: short, precise, correct Python statements using standard library and common third-party APIs (os.path, unittest assertions, torch tensors) that appear in the sampled tasks.
+- **DS-1000 mini-200**: DS-1000 problems are largely about completing a partial Python snippet with the right library call, and the campaign runs them through Terminus-2 in Harbor, the same agentic file-writing setup used here. The sampled tasks include numpy/torch/data-loader API completions (e.g. `loss.item()`, `sim_matrix`), so both the skill and the delivery format overlap.
+- **SWE-bench Verified random-100**: The dataset trains reading related repository files to infer how a project's own classes and helpers are used (e.g. hotpot, genbench, pyvsc, appsignal internals) before editing code, which is a core sub-skill of repository-level issue repair in a terminal agent. Expected gain is modest since the tasks here are single statements with no test execution.
+
+## Files
+
+- `tasks.parquet`: kept tasks.
+- `archive.parquet`: excluded tasks with validation reasons.
+
+Generated with claude-fable-5-1; domain and capability labels follow CLI-Universe.
+Taxonomy source: https://arxiv.org/abs/2606.22883v1

crosscodeeval-python-v3/archive.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb7fdfb21a32fec9ed1f4012f47dd0d35e41caddc3ae5a3ca47a466b66caf9ed
+size 1276

crosscodeeval-python-v3/tasks.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f6f7cdd7444174b6031cf801e764dedc45b4c7a3a41c4f314e38305d84e1d34
+size 4576227

crosscodeeval-typescript-v3/README.md

@@ -1,12 +1,32 @@
-# CrossCodeEval, TypeScript
+# crosscodeeval-typescript-v3
 
-Cross-file code completion. The agent sees a file that is cut at a cursor and must write the next statement. That statement depends on code in other files of the same repository, which the task provides as context files.
+## Original datasource
 
-| | |
-| --- | --- |
-| Tasks before validation | 3,030 |
-| Underlying benchmark | [amazon-science/cceval](https://github.com/amazon-science/cceval), commit `40c68d2b` |
-| Packaged source | [TaskTrove](https://huggingface.co/datasets/open-thoughts/TaskTrove), `laion__exp_rpt_crosscodeeval-typescript-v3`, revision `12df4483fe99`, from [discussion 3](https://huggingface.co/datasets/open-thoughts/TaskTrove/discussions/3) |
-| Reward | 1 if the first statement matches the reference exactly, after removing comments and surrounding whitespace; otherwise 0 |
+- CrossCodeEval (TypeScript split) — https://arxiv.org/abs/2310.11248
 
-The tasks were patched before validation: the cross-file context and the benchmark's own scorer were added, and tasks whose reference cannot be derived from the context were dropped. The changes are listed in the [CrossCodeEval patch notes](https://github.com/MLI-lab/OT-next-data/blob/main/data/crosscodeeval/README.md).
+## Domains
+
+Software Engineering
+
+- **Software Engineering:** Every sampled task is repository-level code completion in real TypeScript application code (NestJS services, Kysely-style query builders, game screens, browser-extension pages, subtitle parsers). The agent must emit the 1-4 line continuation of a file cut off mid-statement so that it matches the project's existing APIs and types; no debugging, data work, or system administration is involved.
+
+## Agent capabilities
+
+Exploration, Specification Adherence
+
+- **Exploration:** Intended practice: the target identifiers (e.g. setScreen, instantiateNoReturns, BlockTuneConstructorParameters, savePreferences) are not in the truncated prompt but appear in the 4-5 retrieved cross-file excerpts under /setup_files/context/, so the agent must read that directory before committing an answer. No trajectories were supplied, so observed behavior cannot be confirmed.
+- **Specification Adherence:** Intended practice: the verifier (a port of amazon-science/cceval eval_metric.py) gives reward 1 only on exact line-level match of the first statement after comment stripping, and the instructions insist on writing only the raw continuation to /app/solution.txt with no re-emitted snippet, imports, fences, or labels, via the quote-safe write_solution helper. Any drift in output format or extra text yields zero reward. Not observed in trajectories.
+
+## Expected benchmark transfer
+
+These are hypotheses, not measured training gains.
+
+- **SWE-bench Verified random-100**: The practiced skill is producing code that is consistent with APIs, types, and conventions defined in other files of the same repository, discovered by reading retrieved cross-file context inside a Terminus-2 container. That repository-grounded completion habit is the closest match in the suite, although transfer is expected to be modest because the dataset is TypeScript-only and single-statement while SWE-bench is Python issue repair with executable tests.
+
+## Files
+
+- `tasks.parquet`: kept tasks.
+- `archive.parquet`: excluded tasks with validation reasons.
+
+Generated with claude-fable-5-1; domain and capability labels follow CLI-Universe.
+Taxonomy source: https://arxiv.org/abs/2606.22883v1

crosscodeeval-typescript-v3/archive.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb7fdfb21a32fec9ed1f4012f47dd0d35e41caddc3ae5a3ca47a466b66caf9ed
+size 1276

crosscodeeval-typescript-v3/tasks.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ab77ee07d0cc6a6f5875457cd98f9038f1a338ecf4935fafe641222b7f552b8
+size 29994343

runs/2026-10-03-c536bae24db8.json

@@ -0,0 +1,192 @@
+{
+  "run": "2026-10-03-c536bae24db8",
+  "created_at": "2026-10-03T11:39:15.433451+00:00",
+  "contract_sha256": "c536bae24db82f0fd5cf1dc8f084ff8cf1820ffe59e5d0e18e75abdcb83f86cc",
+  "dataset": "open-thoughts/TaskTrove",
+  "dataset_revision": "12df4483fe99c79ccbb4c923d76ab5a2b042e64a",
+  "stages_in_contract": [
+    1,
+    3,
+    4,
+    5
+  ],
+  "stages_reported": [
+    1,
+    3,
+    4,
+    5
+  ],
+  "execution_profile": {
+    "backend": "apptainer",
+    "architecture": "x86_64",
+    "network": "host",
+    "network_guarantee": "host networking through cluster HTTP proxy",
+    "privileges": "fakeroot requested, host UID unprivileged; bridge may fall back without fakeroot",
+    "validation_container_reuse": "fresh instance per phase"
+  },
+  "static_checkpoint": null,
+  "static_resume": null,
+  "static_checks": [
+    "check-compose-host-binds.sh",
+    "check-dockerfile-platform.sh",
+    "check-dockerfile-references.sh",
+    "check-dockerfile-sanity.sh",
+    "check-instruction-suffix.sh",
+    "check-no-cheat-dir.sh",
+    "check-nproc.sh",
+    "check-pip-pinning.sh",
+    "check-resource-sizes.sh",
+    "check-task-absolute-path.sh",
+    "check-task-timeout.sh",
+    "check-test-file-references.sh",
+    "check-trial-network-fetch.sh",
+    "check-verifier-tooling-baked.sh",
+    "check_ai_detection.py"
+  ],
+  "static_exclusions": {
+    "check-task-changelog.sh": "training tasks do not require Terminal-Bench PR changelogs",
+    "check-test-sh-sanity.sh": "applies to shared verifiers that install test tools; this verifier uses only the Python standard library",
+    "check-separate-verifier.sh": "CrossCodeEval grades in the agent container; the verifier reads one file and its reference is uploaded after the agent has finished",
+    "check-allow-internet.sh": "network policy is declared by the task and checked at runtime",
+    "check-no-allow-internet-true.sh": "explicit task network policies are allowed",
+    "check-pytest-version.sh": "training tasks may use other pinned pytest/CTRF versions",
+    "check-gpu-types.sh": "Modal GPU naming conventions do not apply to the Helma execution profile",
+    "check-canary.sh": "training tasks do not require a benchmark canary string",
+    "check-instruction-headings.sh": "Markdown headings are allowed in training instructions",
+    "check-task-fields.sh": "training tasks do not require TB author GitHub metadata, taxonomy or explanation sections",
+    "check-task-package-name.sh": "training task package names need not use the terminal-bench namespace",
+    "check-task-slug.sh": "training task names may contain more than three hyphen-separated tokens",
+    "rubric_review.py": "LLM rubric review is opt-in via stage 2, never a default static check"
+  },
+  "not_required_checks": [],
+  "rules": {
+    "1": "archive if any static check fails",
+    "3": "archive if the container does not build or start",
+    "4": "archive if the oracle reward is not 1",
+    "5": "archive if the no-answer reward is not 0",
+    "other": "never archive"
+  },
+  "validation_code": "https://github.com/MLI-lab/OT-next-data/tree/main/validation",
+  "stage_timings": {
+    "1": {
+      "node": "h32-32",
+      "slurm_job_id": "923204",
+      "started_at": "2026-10-02T08:25:41+00:00",
+      "wall_seconds": 1528.5,
+      "concurrency": 16,
+      "tasks": 6710,
+      "tasks_timed": 6710,
+      "task_seconds": {
+        "median": 0.5,
+        "p90": 1.4,
+        "max": 600.5
+      }
+    },
+    "3": {
+      "node": "h32-32",
+      "slurm_job_id": "923204",
+      "started_at": "2026-10-02T08:52:38+00:00",
+      "wall_seconds": 6938.0,
+      "concurrency": 16,
+      "tasks": 6710,
+      "tasks_timed": 6710,
+      "task_seconds": {
+        "median": 13.4,
+        "p90": 24.8,
+        "max": 132.8
+      }
+    },
+    "4": {
+      "node": "h32-32",
+      "slurm_job_id": "923204",
+      "started_at": "2026-10-02T10:48:32+00:00",
+      "wall_seconds": 6061.3,
+      "concurrency": 16,
+      "tasks": 6710,
+      "tasks_timed": 6710,
+      "task_seconds": {
+        "median": 11.2,
+        "p90": 27.8,
+        "max": 116.9
+      }
+    },
+    "5": {
+      "node": "h32-32",
+      "slurm_job_id": "923204",
+      "started_at": "2026-10-02T12:29:48+00:00",
+      "wall_seconds": 6468.8,
+      "concurrency": 16,
+      "tasks": 6710,
+      "tasks_timed": 6710,
+      "task_seconds": {
+        "median": 11.5,
+        "p90": 29.8,
+        "max": 130.7
+      }
+    }
+  },
+  "data_sources": {
+    "crosscodeeval-csharp-v5": {
+      "tasks": 1353,
+      "kept": 1353,
+      "archived": 0,
+      "archived_by_stage": {},
+      "archive_reasons": {},
+      "not_run_by_stage": {},
+      "not_run_reasons": {},
+      "not_run_examples": {}
+    },
+    "crosscodeeval-java-v4": {
+      "tasks": 1895,
+      "kept": 1895,
+      "archived": 0,
+      "archived_by_stage": {},
+      "archive_reasons": {},
+      "not_run_by_stage": {},
+      "not_run_reasons": {},
+      "not_run_examples": {}
+    },
+    "crosscodeeval-python-v3": {
+      "tasks": 432,
+      "kept": 432,
+      "archived": 0,
+      "archived_by_stage": {},
+      "archive_reasons": {},
+      "not_run_by_stage": {},
+      "not_run_reasons": {},
+      "not_run_examples": {}
+    },
+    "crosscodeeval-typescript-v3": {
+      "tasks": 3030,
+      "kept": 3030,
+      "archived": 0,
+      "archived_by_stage": {},
+      "archive_reasons": {},
+      "not_run_by_stage": {},
+      "not_run_reasons": {},
+      "not_run_examples": {}
+    }
+  },
+  "dataset_cards": {
+    "crosscodeeval-csharp-v5": {
+      "readme": "crosscodeeval-csharp-v5/README.md",
+      "model": "claude-fable-5-1",
+      "prompt_sha256": "9160ac9a34c20a0435fe87d152a2b319728506b6d56bbaaa9e028f430730e16c"
+    },
+    "crosscodeeval-java-v4": {
+      "readme": "crosscodeeval-java-v4/README.md",
+      "model": "claude-fable-5-1",
+      "prompt_sha256": "b0b2e1ecc926747a49999a7bcc42194eaae1ad5f5e1c949144041715eff6c4d9"
+    },
+    "crosscodeeval-python-v3": {
+      "readme": "crosscodeeval-python-v3/README.md",
+      "model": "claude-fable-5-1",
+      "prompt_sha256": "c3edb8190a87ac2a4ee6cf203b6e07503eaeeaa8ea098d2ca0c73e905cd2773e"
+    },
+    "crosscodeeval-typescript-v3": {
+      "readme": "crosscodeeval-typescript-v3/README.md",
+      "model": "claude-fable-5-1",
+      "prompt_sha256": "fb755f575b27e8b92150a7d7abd55a941ad9a964b324cf7b6deb6a42e7750847"
+    }
+  }
+}