Dataset Viewer
Auto-converted to Parquet Duplicate
task_id
stringlengths
30
97
source_dataset
stringclasses
8 values
materialization_mode
stringclasses
4 values
split
stringclasses
1 value
has_solution
bool
2 classes
has_verifier
bool
1 class
manifest
stringlengths
651
1.53k
files
listlengths
6
2.41k
megat-cligym-api-dependency-interpreter-path-tampering-d76c2c85bcc7
cli_gym
cli_convert
candidate
false
true
{"dataset":"MegaTerminal","harbor_task_name":"terminal-bench/megat-cligym-api-dependency-interpreter-path-tampering-d76c2c85bcc7","has_solution":false,"has_verifier":true,"materialization_mode":"cli_convert","source_dataset":"cli_gym","source_doc_id":"cli_gym:api-dependency-interpreter-path-tampering","source_meta":{"d...
[ { "path": "environment/Dockerfile", "data": [ 70, 82, 79, 77, 32, 106, 121, 97, 110, 103, 98, 97, 108, 108, 105, 110, 47, 115, 119, 101, 115, 109, 105, 116, 1...
megat-cligym-api-dependency-interpreter-path-tampering-hard-b497291557f5
cli_gym
cli_convert
candidate
false
true
{"dataset":"MegaTerminal","harbor_task_name":"terminal-bench/megat-cligym-api-dependency-interpreter-path-tampering-hard-b497291557f5","has_solution":false,"has_verifier":true,"materialization_mode":"cli_convert","source_dataset":"cli_gym","source_doc_id":"cli_gym:api-dependency-interpreter-path-tampering.hard","source...
[ { "path": "environment/Dockerfile", "data": [ 70, 82, 79, 77, 32, 106, 121, 97, 110, 103, 98, 97, 108, 108, 105, 110, 47, 115, 119, 101, 115, 109, 105, 116, 1...
megat-cligym-api-endpoint-access-disruption-via-firewall-redirection-d0b2682efdf0
cli_gym
cli_convert
candidate
false
true
{"dataset":"MegaTerminal","harbor_task_name":"terminal-bench/megat-cligym-api-endpoint-access-disruption-via-firewall-redirection-d0b2682efdf0","has_solution":false,"has_verifier":true,"materialization_mode":"cli_convert","source_dataset":"cli_gym","source_doc_id":"cli_gym:api-endpoint-access-disruption-via-firewall-re...
[ { "path": "environment/Dockerfile", "data": [ 70, 82, 79, 77, 32, 106, 121, 97, 110, 103, 98, 97, 108, 108, 105, 110, 47, 115, 119, 101, 115, 109, 105, 116, 1...
megat-cligym-api-endpoint-access-disruption-via-firewall-redirection-hard-b06d316f7cc3
cli_gym
cli_convert
candidate
false
true
{"dataset":"MegaTerminal","harbor_task_name":"terminal-bench/megat-cligym-api-endpoint-access-disruption-via-firewall-redirection-hard-b06d316f7cc3","has_solution":false,"has_verifier":true,"materialization_mode":"cli_convert","source_dataset":"cli_gym","source_doc_id":"cli_gym:api-endpoint-access-disruption-via-firewa...
[ { "path": "environment/Dockerfile", "data": [ 70, 82, 79, 77, 32, 106, 121, 97, 110, 103, 98, 97, 108, 108, 105, 110, 47, 115, 119, 101, 115, 109, 105, 116, 1...
megat-cligym-api-generator-source-code-subtle-logic-bug-induction-b5998cfcc1b6
cli_gym
cli_convert
candidate
false
true
{"dataset":"MegaTerminal","harbor_task_name":"terminal-bench/megat-cligym-api-generator-source-code-subtle-logic-bug-induction-b5998cfcc1b6","has_solution":false,"has_verifier":true,"materialization_mode":"cli_convert","source_dataset":"cli_gym","source_doc_id":"cli_gym:api-generator-source-code-subtle-logic-bug-induct...
[ { "path": "environment/Dockerfile", "data": [ 70, 82, 79, 77, 32, 106, 121, 97, 110, 103, 98, 97, 108, 108, 105, 110, 47, 115, 119, 101, 115, 109, 105, 116, 1...
megat-cligym-api-generator-source-code-subtle-logic-bug-induction-hard-552ef6bb6ce9
cli_gym
cli_convert
candidate
false
true
{"dataset":"MegaTerminal","harbor_task_name":"terminal-bench/megat-cligym-api-generator-source-code-subtle-logic-bug-induction-hard-552ef6bb6ce9","has_solution":false,"has_verifier":true,"materialization_mode":"cli_convert","source_dataset":"cli_gym","source_doc_id":"cli_gym:api-generator-source-code-subtle-logic-bug-i...
[ { "path": "environment/Dockerfile", "data": [ 70, 82, 79, 77, 32, 106, 121, 97, 110, 103, 98, 97, 108, 108, 105, 110, 47, 115, 119, 101, 115, 109, 105, 116, 1...
megat-cligym-ast-based-runtime-corruption-of-marshmallow-default-value-resolution-m-0e1c76da7b98
cli_gym
cli_convert
candidate
false
true
{"dataset":"MegaTerminal","harbor_task_name":"terminal-bench/megat-cligym-ast-based-runtime-corruption-of-marshmallow-default-value-resolution-m-0e1c76da7b98","has_solution":false,"has_verifier":true,"materialization_mode":"cli_convert","source_dataset":"cli_gym","source_doc_id":"cli_gym:ast-based-runtime-corruption-of...
[ { "path": "environment/Dockerfile", "data": [ 70, 82, 79, 77, 32, 106, 121, 97, 110, 103, 98, 97, 108, 108, 105, 110, 47, 115, 119, 101, 115, 109, 105, 116, 1...
megat-cligym-ast-based-runtime-corruption-of-marshmallow-default-value-resolution-m-ed1a0ec41807
cli_gym
cli_convert
candidate
false
true
{"dataset":"MegaTerminal","harbor_task_name":"terminal-bench/megat-cligym-ast-based-runtime-corruption-of-marshmallow-default-value-resolution-m-ed1a0ec41807","has_solution":false,"has_verifier":true,"materialization_mode":"cli_convert","source_dataset":"cli_gym","source_doc_id":"cli_gym:ast-based-runtime-corruption-of...
[ { "path": "environment/Dockerfile", "data": [ 70, 82, 79, 77, 32, 106, 121, 97, 110, 103, 98, 97, 108, 108, 105, 110, 47, 115, 119, 101, 115, 109, 105, 116, 1...
megat-cligym-break-python-io-multiplexing-via-hijacked-select-module-7c54e867d142
cli_gym
cli_convert
candidate
false
true
{"dataset":"MegaTerminal","harbor_task_name":"terminal-bench/megat-cligym-break-python-io-multiplexing-via-hijacked-select-module-7c54e867d142","has_solution":false,"has_verifier":true,"materialization_mode":"cli_convert","source_dataset":"cli_gym","source_doc_id":"cli_gym:break-python-io-multiplexing-via-hijacked-sele...
[ { "path": "environment/Dockerfile", "data": [ 70, 82, 79, 77, 32, 106, 121, 97, 110, 103, 98, 97, 108, 108, 105, 110, 47, 115, 119, 101, 115, 109, 105, 116, 1...
megat-cligym-break-python-io-multiplexing-via-hijacked-select-module-hard-f67baab3ad73
cli_gym
cli_convert
candidate
false
true
{"dataset":"MegaTerminal","harbor_task_name":"terminal-bench/megat-cligym-break-python-io-multiplexing-via-hijacked-select-module-hard-f67baab3ad73","has_solution":false,"has_verifier":true,"materialization_mode":"cli_convert","source_dataset":"cli_gym","source_doc_id":"cli_gym:break-python-io-multiplexing-via-hijacked...
[ { "path": "environment/Dockerfile", "data": [ 70, 82, 79, 77, 32, 106, 121, 97, 110, 103, 98, 97, 108, 108, 105, 110, 47, 115, 119, 101, 115, 109, 105, 116, 1...
megat-cligym-broken-python-gnutls-library-causing-tls-handshake-failures-5c03cda7af90
cli_gym
cli_convert
candidate
false
true
{"dataset":"MegaTerminal","harbor_task_name":"terminal-bench/megat-cligym-broken-python-gnutls-library-causing-tls-handshake-failures-5c03cda7af90","has_solution":false,"has_verifier":true,"materialization_mode":"cli_convert","source_dataset":"cli_gym","source_doc_id":"cli_gym:broken-python-gnutls-library-causing-tls-h...
[ { "path": "environment/Dockerfile", "data": [ 70, 82, 79, 77, 32, 106, 121, 97, 110, 103, 98, 97, 108, 108, 105, 110, 47, 115, 119, 101, 115, 109, 105, 116, 1...
megat-cligym-broken-python-gnutls-library-causing-tls-handshake-failures-hard-345a13beb8e4
cli_gym
cli_convert
candidate
false
true
{"dataset":"MegaTerminal","harbor_task_name":"terminal-bench/megat-cligym-broken-python-gnutls-library-causing-tls-handshake-failures-hard-345a13beb8e4","has_solution":false,"has_verifier":true,"materialization_mode":"cli_convert","source_dataset":"cli_gym","source_doc_id":"cli_gym:broken-python-gnutls-library-causing-...
[ { "path": "environment/Dockerfile", "data": [ 70, 82, 79, 77, 32, 106, 121, 97, 110, 103, 98, 97, 108, 108, 105, 110, 47, 115, 119, 101, 115, 109, 105, 116, 1...
megat-cligym-broken-symlink-chain-in-critical-python-library-paths-624393437e91
cli_gym
cli_convert
candidate
false
true
{"dataset":"MegaTerminal","harbor_task_name":"terminal-bench/megat-cligym-broken-symlink-chain-in-critical-python-library-paths-624393437e91","has_solution":false,"has_verifier":true,"materialization_mode":"cli_convert","source_dataset":"cli_gym","source_doc_id":"cli_gym:broken-symlink-chain-in-critical-python-library-...
[ { "path": "environment/Dockerfile", "data": [ 70, 82, 79, 77, 32, 106, 121, 97, 110, 103, 98, 97, 108, 108, 105, 110, 47, 115, 119, 101, 115, 109, 105, 116, 1...
megat-cligym-broken-symlink-chain-in-critical-python-library-paths-hard-f0b5e1629900
cli_gym
cli_convert
candidate
false
true
{"dataset":"MegaTerminal","harbor_task_name":"terminal-bench/megat-cligym-broken-symlink-chain-in-critical-python-library-paths-hard-f0b5e1629900","has_solution":false,"has_verifier":true,"materialization_mode":"cli_convert","source_dataset":"cli_gym","source_doc_id":"cli_gym:broken-symlink-chain-in-critical-python-lib...
[ { "path": "environment/Dockerfile", "data": [ 70, 82, 79, 77, 32, 106, 121, 97, 110, 103, 98, 97, 108, 108, 105, 110, 47, 115, 119, 101, 115, 109, 105, 116, 1...
End of preview. Expand in Data Studio

MegaTerminal

MegaTerminal is a Harbor-style terminal task-folder dataset assembled from the terminal task sources collected for task matching experiments.

The task folders are packed into parquet/ as a custom blob layout (megatask-folder-parquet-v1); each row stores the files of one task folder as binary blobs. Restore the on-disk tasks/<task-id>/ tree with python scripts/deparquetize_megaterminal.py --parquet-dir parquet --out-dir MegaTerminal-restored.

Each task lives under tasks/<task-id>/ and includes task.toml, instruction.md, environment/, tests/, and solution/ so the tree has the same shape as Harbor task datasets. The training split excludes the original Terminal-Bench benchmark source and does not publish text-only scaffolds as tasks. Records that could not be recovered as real task folders from their original source datasets are listed in skipped_records.jsonl.

Layout

  • tasks/: materialized Harbor-style task folders.
  • manifest.jsonl: one record per task folder.
  • excluded_records.jsonl: source records excluded by provenance or source type.
  • skipped_records.jsonl: source records intentionally not materialized.
  • task_names.txt: task ids, one per line.
  • summary.json: dataset-level counts.
  • source_audit_report.json: source provenance audit, when present.
  • SHA256SUMS: checksums for top-level metadata files.

Counts

  • tasks: 11599
  • candidate split: 11599
  • anchor split: 0

Source Counts

  • cli_gym: 1639
  • endless_terminals: 2492
  • litecoder_terminal_rl_preview: 599
  • openthoughts_agent_rl: 728
  • seta_evolve: 1312
  • seta_synth: 3255
  • seta_synth_env: 1374
  • terminal_bench_pro: 200

Materialization Modes

  • archive_extract: 928
  • cli_convert: 1639
  • hf_folder_copy: 4465
  • hf_release_tar_copy: 4567

Use manifest.jsonl for the materialized task inventory and excluded_records.jsonl/skipped_records.jsonl for the non-materialized source audit.

Downloads last month
59