Dataset Viewer
The dataset viewer is not available for this subset.
Cannot get the split names for the config 'default' of the dataset.
Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/harbor/harbor.py", line 171, in _split_generators
                  raise DataFilesNotFoundError("No task.toml or instruction.md files found")
              datasets.exceptions.DataFilesNotFoundError: No task.toml or instruction.md files found
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Skills Desk

An OpenEnv environment for the OpenEnv Arena that trains reliable problem solving with tools: understand the request, find the relevant information, compute or repair something, produce the required output, and check it. Each episode is a small workspace under /app with a short instruction. The agent runs commands (exec), writes files (write_file) and sends evaluate; a hidden grader checks the result and returns 1 or 0.

Image: ghcr.io/syndicate604/skills-desk (linux/amd64), plus the Pool Desk v8 images for five tasks. Source: source/. Forked from Science Desk and dnakhla/terminal-desk v2 (Apache-2.0); see NOTICE.

Interface

Terminal Desk v2's copy of OpenEnv's upstream tbench2_env contract: exec, write_file, evaluate, close; the instruction repeated in every observation; output not cut; per-task budgets of 4096 completion tokens (model tokens plus every observation after the first), 8192 context tokens and 32 tool calls. The image has Python 3.12 with numpy, scipy, pandas and astropy, plus jq, bc and sqlite3.

Training tasks (v6: 32)

manifest/training_tasks_v6.txt; selection in manifest/v6-SELECTION.md. v6 keeps the v5 format tasks that gave a learning signal in the v5 arena run and adds five exact-math tasks from Pool Desk v8.

  • Removed from v5 (8): the four fmt_text-L3 instances, never solved in training (mean reward 0), and four fmt_json-L1/fmt_table-L1 instances solved almost every time (mean 0.9 or more). Neither kind teaches anything.
  • Kept from v5 (27): JSON answers (fmt_json L1, L3), tables (fmt_table L1-L3), final-answer files (fmt_text L1), money, times, changes and numbers, all unchanged. Each answer is graded on its form, all or nothing, character by character or by strict parse, against a written contract.
  • Added from Pool Desk v8 (5, CC0-1.0): pattern races on a weighted spinner, counting routing plans in a sensor network (directed matrix-tree), cheapest-route counts with zero-cost cycles, state-machine equivalence with a shortest distinguishing input, and absorption probabilities of a support-ticket Markov chain. Each has one exact answer that the obvious shortcut gets wrong. They are served unchanged by their own published images, which use the same interface; instructions and provenance are in pool-desk-v8/. These five had mixed results in most training steps across all four arena runs of Pool Desk v8.

Earlier versions: v5 (35 format tasks) and v3/v4 scored 4/40 each.

Reserved evaluation tasks (20)

manifest/reserved_tasks.txt: four per skill, two fresh applications of a familiar skill and two combining skills, with different file layouts and solution structures. They are served by the same image so a trained checkpoint can be evaluated on them, but they are never trained on and never used to choose or tune the training tasks. The set is frozen (v1); a change would be published as a new version.

tasks.jsonl lists the 35 training and 20 reserved ids with their split, skill, output path and fixed instruction.

Reward

Binary on evaluate. Content and structure are graded together: correct JSON with wrong values fails, correct values in a malformed file fail, and a CSV row with extra or missing fields fails. Harmless differences pass (key order, CSV quoting, line endings, and either rounding of an exact half where the task states no rounding rule; a task that states half-up rounding requires it). Tasks with many correct answers (rosters, schedules, plans) are graded against their rules. Repair tasks run the repaired code as nobody on hidden input. Nothing rewards a particular tool sequence, reply length or claims of checking.

Verification

Every task: the reference output scores 1, no output and corrupted output score 0, an independent agent-style reference script scores 1 on four instances inside the image, and a deliberate common mistake scores 0. Instances are byte-identical in any process and frozen by fingerprints in source/tests/fixtures/. See source/README.md to build and test.

Limitations

The selection screen is small (6 episodes per task) and uses the untrained model through a provider, not the arena's sampler; earlier screens underestimated the arena-trained model, and no correction was applied. Whether training on these tasks transfers can only be measured by evaluating a trained checkpoint on the reserved tasks; the arena's private score is a separate measure.

Downloads last month
315