CoolFace
Datasetpublic

sabia0080/agentbench-sft-trajectories-v3-planA

agentbench-sft-trajectories-v3-planA Merged dataset for LLM Advanced Competition SFT (Plan A). Composition Source Records Pct u-10bei/dbbench_sft_dataset_react (v1) 300 5.5% u-10bei/dbbench_sft_dataset_react_v2 360 6.6% u-10bei/dbbench_sft_dataset_react_v3 1,112 20.3% u-10bei/dbbench_sft_dataset_react_v4 1,194 21.8% u-10bei/sft_alfworld_trajectory_dataset_v5 2,502 45.8% Total (after dedup) 5,468 100% Design All 4 versions… See the full description on the dataset page: https://huggingface.co/datasets/sabia0080/agentbench-sft-trajectories-v3-planA.

sourceHugging Facemitupdated 7mo agoView on Hugging Face
0likes19downloads
Dataset Card

agentbench-sft-trajectories-v3-planA

Merged dataset for LLM Advanced Competition SFT (Plan A).

Composition

SourceRecordsPct
u-10bei/dbbenchsftdataset_react (v1)3005.5%
u-10bei/dbbenchsftdatasetreactv23606.6%
u-10bei/dbbenchsftdatasetreactv31,11220.3%
u-10bei/dbbenchsftdatasetreactv41,19421.8%
u-10bei/sftalfworldtrajectorydatasetv52,50245.8%
Total (after dedup)5,468100%

Design

  • —All 4 versions of DBBench react datasets combined (same IDs but different generated content per version)
  • —ALFWorld v5 included as-is
  • —94 exact duplicates removed via content hash
  • —Shuffled with seed=42
  • —Official u-10bei data only (no custom distilled data)

Recommended Training Config

  • —seq_len: 2048
  • —lr: 1e-5
  • —LoRA r: 64, alpha: 128
  • —epochs: 2