num_samples int64 | hidden_size int64 | model_name string | model_revision string | max_model_len int64 | has_hard_negatives bool | samples list |
|---|---|---|---|---|---|---|
123,849 | 4,096 | Qwen/Qwen3-8B | b968826d9c46dd6066d109eabc6255188de91218 | 8,192 | false | [{"task_id":"terminal-bench/uefi-bootkit","trajectory_id":"02a8f8f7-4d2b-4c06-afe7-f1551a8e3781","st(...TRUNCATED) |
YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Terminal trajectory embedding task subset
20 selected tasks, 1,281 trajectories, and 123,849 state/action pairs. Only trajectories with numeric reward == 1 are retained.
The saved vectors are exact selected rows of the existing embeddings; the encoder
was not rerun. train/metadata.json matches both tensor row orders.
source_row_indices.json records the original row indices; selection.json
records selection parameters, source checksums and output checksums.
selected_tasks.json identifies the selected tasks. heldout_tasks.json lists
the 46 other tasks, whose vectors remain in the full source.
If this subset is used for training, evaluate on held-out tasks, excluding these
selected tasks. A trajectory-level success label is not an action correctness label.
All source context limits and truncation metadata remain unchanged.
Source embedding dataset: Hkang/terminal-bench-4-qwen3-8b-embeddings, commit 796ecf6ad08e4deaa06507fb5418ca60a74489ba.
This is a reproducible example using seed 42 and successful trajectories only; the full 66-task source remains available. It uses the original Qwen3-8B 8K embeddings with their existing truncation metadata.
- Downloads last month
- -