Dataset Viewer
Auto-converted to Parquet Duplicate
schema_version
int64
files
list
artifact
dict
1
[ { "path": "workspace/gsm8k_verl/test.parquet", "size": 1183034, "sha256": "0dcd50ed32caa9d8f434d4e3ef8031bc3e49b0f823655a53be5c002e727f17dd" }, { "path": "workspace/gsm8k_verl/train.parquet", "size": 6555602, "sha256": "8fb13a0cd8621e5fcacdee3ffe28b3a369975478ea7ab514781f3f190cba2488" ...
{ "path": "artifacts/workspace-assets.tar.gz", "size": 2490115, "sha256": "7211ce9e85dae7bb431d951ab4970706e29113d06f9e5e6ec5f24c64a7fcf449" }

PTDBench dataset snapshot: task_engine_base_025

This repository stores the immutable runtime dataset snapshot for one materialized PTDBench task. It intentionally excludes model weights and training checkpoints.

  • PTDBench family: verl_reward_implementation
  • Source evaluation metric: val-core/openai/gsm8k/acc/mean@1
  • Provenance: Processed runtime snapshot of openai/gsm8k.
  • License: MIT

The artifact manifest records every hydrated runtime path, byte size, and SHA-256. The task Dockerfile pins the repository commit after publication.

Downloads last month
41