Datasets:

ZHIYII
/

AgentBrew_Notion

messages listlengths 3 3	weight float64 0 2	task_id stringlengths 16 79	step_id int64 0 48
[ { "content": "You are an agent for Notion workspace operations and content management.", "role": "system" }, { "content": "You are a ReAct (Reasoning and Acting) agent.\nYou are an agent for Notion workspace operations and content management.\n\nCRITICAL Notion API rules — follow these exactly to av...	0.816959	100_Writing_Prompts_for_Journaling_c56775e7_0	0
[ { "content": "You are an agent for Notion workspace operations and content management.", "role": "system" }, { "content": "You are a ReAct (Reasoning and Acting) agent.\nYou are an agent for Notion workspace operations and content management.\n\nCRITICAL Notion API rules — follow these exactly to av...	2	100_Writing_Prompts_for_Journaling_c56775e7_0	2
[ { "content": "You are an agent for Notion workspace operations and content management.", "role": "system" }, { "content": "You are a ReAct (Reasoning and Acting) agent.\nYou are an agent for Notion workspace operations and content management.\n\nCRITICAL Notion API rules — follow these exactly to av...	1.076064	100_Writing_Prompts_for_Journaling_c56775e7_0	3
[ { "content": "You are an agent for Notion workspace operations and content management.", "role": "system" }, { "content": "You are a ReAct (Reasoning and Acting) agent.\nYou are an agent for Notion workspace operations and content management.\n\nCRITICAL Notion API rules — follow these exactly to av...	1.022261	100_Writing_Prompts_for_Journaling_c56775e7_0	4
[{"content":"You are an agent for Notion workspace operations and content management.","role":"syste(...TRUNCATED)	0.914654	100_Writing_Prompts_for_Journaling_c56775e7_0	5
[{"content":"You are an agent for Notion workspace operations and content management.","role":"syste(...TRUNCATED)	1.01235	100_Writing_Prompts_for_Journaling_c56775e7_0	6
[{"content":"You are an agent for Notion workspace operations and content management.","role":"syste(...TRUNCATED)	0.152914	100_Writing_Prompts_for_Journaling_c56775e7_0	8
[{"content":"You are an agent for Notion workspace operations and content management.","role":"syste(...TRUNCATED)	0.02053	100_Writing_Prompts_for_Journaling_c56775e7_0	9
[{"content":"You are an agent for Notion workspace operations and content management.","role":"syste(...TRUNCATED)	1.242429	100_Writing_Prompts_for_Journaling_c56775e7_0	10
[{"content":"You are an agent for Notion workspace operations and content management.","role":"syste(...TRUNCATED)	1.968062	100_Writing_Prompts_for_Journaling_c56775e7_1	0

End of preview. Expand in Data Studio

AgentBrew Notion

AgentBrew Notion is a public weighted action-level SFT dataset for training Notion tool-use agents. It is converted from ZHIYII/Notion_Entropy_Action_SFT_swift_trace for the AgentBrew weighted-SFT training pipeline.

Processing

Parsed messages from a serialized JSON string into a native list of chat messages.
Kept only rows with a finite weight > 0.
Removed is_final_answer, is_error_step, and raw_advantage.
Retained task_id and step_id for provenance and analysis.
Preserved the source dataset's positive training weights without further normalization.

Schema

Field	Type	Description
`messages`	list of objects	`system`, `user`, and `assistant` chat messages.
`weight`	float	Positive per-sample weighted-SFT multiplier.
`task_id`	string	Source task identifier.
`step_id`	integer	Action index within the source task.

Statistics

Training rows: 12,861
Unique tasks: 2,128
Weight range: 0.0005305743467 to 2
Mean weight: 0.8712219357
Role pattern: system -> user -> assistant

Loading

from datasets import load_dataset

dataset = load_dataset("ZHIYII/AgentBrew_Notion", split="train")

The training objective averages causal-language-model loss over valid assistant tokens for each sample, multiplies it by weight, and then averages the weighted sample losses across the batch.

Downloads last month: 56