timchen0618/bcp-plan-revisions-v1
收藏Hugging Face2026-04-08 更新2026-04-12 收录
下载链接：
https://hf-mirror.com/datasets/timchen0618/bcp-plan-revisions-v1
下载链接
链接失效反馈官方服务：
资源简介：
---
license: mit
tags:
- bcp
- plan-revisions
- browsecomp-plus
---
# bcp-plan-revisions-v1
Plan text revision history across 8 planning conditions on BrowseComp-Plus (first 50 queries, Qwen3-Embedding-8B retriever). Each row is one revision entry from plan_text_history in the trajectory metadata.
## Dataset Info
- **Rows**: 2063
- **Columns**: 8
## Columns
| Column | Type | Description |
|--------|------|-------------|
| condition | Value('string') | Run condition name (planning version + optional gemini start extension) |
| run_id | Value('string') | Trajectory filename (without .json) |
| query_id | Value('string') | BrowseComp-Plus query ID |
| status | Value('string') | Trajectory completion status (e.g. 'completed') |
| revision_index | Value('int64') | Index in plan_text_history (0 = initial plan, 1+ = revisions) |
| source | Value('string') | Source of this plan entry: 'initial' or 'revise' |
| plan_text | Value('string') | Full plan text at this revision |
| total_revisions | Value('int64') | Total number of plan entries (initial + revisions) for this run |
## Generation Parameters
```json
{
"script_name": "extract_plan_revisions.py",
"model": "gpt-oss-120b",
"description": "Plan text revision history across 8 planning conditions on BrowseComp-Plus (first 50 queries, Qwen3-Embedding-8B retriever). Each row is one revision entry from plan_text_history in the trajectory metadata.",
"experiment_name": "bcp-plan-revisions",
"cluster": "torch",
"artifact_status": "final",
"canary": false,
"hyperparameters": {},
"input_datasets": []
}
```
## Usage
```python
from datasets import load_dataset
dataset = load_dataset("timchen0618/bcp-plan-revisions-v1", split="train")
print(f"Loaded {len(dataset)} rows")
```
---
提供机构：
timchen0618
