component string | task_id string | split string | domain string | skill_family string | difficulty int64 | behavior string | source_dataset string | source_revision string | source_url string | source_record string | original_task_definition string | original_task_definition_sha256 string | instance_policy string | reward string |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
dnakhla | sci3_stoich_batch-L2 | train | natural-science | sci3_stoich_batch | 2 | Batch limiting reagent and yield from a stockroom table | dnakhla/terminal-desk | 08fab0beb50b665b42f0ac73840f03140f4ad1c9 | https://huggingface.co/datasets/dnakhla/terminal-desk/tree/08fab0beb50b665b42f0ac73840f03140f4ad1c9 | components/dnakhla/SOURCE_MANIFEST.json | components/dnakhla/upstream/terminal_desk/families/science3.py | 59477a1f9d38dbe5607f6432b4f8538342eec93faeed09bb6c9cb1402980c9a1 | Original random instance generation at reset. | One if all original artifact checks pass; zero otherwise. |
dnakhla | math3_exam_slots-L1 | train | mathematics-or-formal-reasoning | math3_exam_slots | 1 | Sequence exam blocks into slots | dnakhla/terminal-desk | 08fab0beb50b665b42f0ac73840f03140f4ad1c9 | https://huggingface.co/datasets/dnakhla/terminal-desk/tree/08fab0beb50b665b42f0ac73840f03140f4ad1c9 | components/dnakhla/SOURCE_MANIFEST.json | components/dnakhla/upstream/terminal_desk/families/mathematics3.py | 9dc54710cd826aadbe5de55d9093959e38c59dc59ae080e40d379cc4db1987c1 | Original random instance generation at reset. | One if all original artifact checks pass; zero otherwise. |
dnakhla | sec3_fw_verdicts-L2 | train | cybersecurity | sec3_fw_verdicts | 2 | Evaluate an ordered firewall rule table | dnakhla/terminal-desk | 08fab0beb50b665b42f0ac73840f03140f4ad1c9 | https://huggingface.co/datasets/dnakhla/terminal-desk/tree/08fab0beb50b665b42f0ac73840f03140f4ad1c9 | components/dnakhla/SOURCE_MANIFEST.json | components/dnakhla/upstream/terminal_desk/families/security3.py | 6f5ec9c2dfe9aad81ba382d4f6ffdb7f64493d92cd7fe5d653134ef0dac107af | Original random instance generation at reset. | One if all original artifact checks pass; zero otherwise. |
dnakhla | math3_exam_slots-L2 | train | mathematics-or-formal-reasoning | math3_exam_slots | 2 | Sequence exam blocks into slots | dnakhla/terminal-desk | 08fab0beb50b665b42f0ac73840f03140f4ad1c9 | https://huggingface.co/datasets/dnakhla/terminal-desk/tree/08fab0beb50b665b42f0ac73840f03140f4ad1c9 | components/dnakhla/SOURCE_MANIFEST.json | components/dnakhla/upstream/terminal_desk/families/mathematics3.py | 9dc54710cd826aadbe5de55d9093959e38c59dc59ae080e40d379cc4db1987c1 | Original random instance generation at reset. | One if all original artifact checks pass; zero otherwise. |
dnakhla | sec3_auth_triage-L2 | train | cybersecurity | sec3_auth_triage | 2 | Block SSH brute-force sources | dnakhla/terminal-desk | 08fab0beb50b665b42f0ac73840f03140f4ad1c9 | https://huggingface.co/datasets/dnakhla/terminal-desk/tree/08fab0beb50b665b42f0ac73840f03140f4ad1c9 | components/dnakhla/SOURCE_MANIFEST.json | components/dnakhla/upstream/terminal_desk/families/security3.py | 6f5ec9c2dfe9aad81ba382d4f6ffdb7f64493d92cd7fe5d653134ef0dac107af | Original random instance generation at reset. | One if all original artifact checks pass; zero otherwise. |
dnakhla | sec3_exfil_flows-L2 | train | cybersecurity | sec3_exfil_flows | 2 | Flag off-hours exfiltration flows | dnakhla/terminal-desk | 08fab0beb50b665b42f0ac73840f03140f4ad1c9 | https://huggingface.co/datasets/dnakhla/terminal-desk/tree/08fab0beb50b665b42f0ac73840f03140f4ad1c9 | components/dnakhla/SOURCE_MANIFEST.json | components/dnakhla/upstream/terminal_desk/families/security3.py | 6f5ec9c2dfe9aad81ba382d4f6ffdb7f64493d92cd7fe5d653134ef0dac107af | Original random instance generation at reset. | One if all original artifact checks pass; zero otherwise. |
6 original Terminal Desk tasks served by one OpenEnv image.
Original generators, tools and graders are retained. Reset generates a fresh instance. Reward is one when all required artifact checks pass, otherwise zero.
Use the tools shown at reset and op=finish to grade the output. Task IDs and pinned source references are in tasks.jsonl. Licenses and provenance are retained in the component directory.
Run python3 -B validate_dataset.py to verify the file inventory.
- Downloads last month
- 306