dongguanting/Agent-Reflex-8B
Text Generation • 8B • Updated • 16
Datasets:
conversations listlengths 2 182 | system stringclasses 284
values | tools stringlengths 205 313k |
|---|---|---|
[
{
"from": "human",
"value": "我在做一个 GitHub 项目健康度的对比小报告：请先从检索到的仓库列表里，定位出完整名为 `nektos/act` 的那个仓库，并基于它的仓库信息给出一个“仓库基础分”（规则：如果该仓库是公开仓库记 10 分，否则 0 分；如果它不是 fork 再加 5 分；如果仓库描述不为空再加 3 分）。然后请再针对 `octocat/Hello-World`：用分支列表判断“默认分支”取名为 `master` 的那个分支，并把它的最新提交 SHA 与提交列表中的所有 SHA 做比对，计算“分支命中率”=（分支最新提交 SHA 在提交列表中出现的次数 / 提交列... | You are an agent designed to accomplish tasks.
Your goal is to provide a comprehensive, objective, and well-researched answer to the user's query. You have access to multiple tools. Please read each tool's description and parameters carefully. When asked to output a JOSN format, you should NOT add any other content in... | [{"type": "function", "function": {"name": "list_repositories", "description": "Return the list of repository metadata from the repositories.json dataset.", "parameters": {"type": "object", "properties": {}, "required": []}}}, {"type": "function", "function": {"name": "list_octocat_issues", "description": "Return all o... |
[
{
"from": "human",
"value": "1. Verify the details of vehicle VIN `1HGBH41JXMN109186` and retrieve its most recent service record; then determine, using the maintenance schedule for a 2020 Honda Accord, whether an oil‑change service is currently due. \n\n2. If the oil‑change is due, schedule a maintenance ... | You are a helpful assistant. When given a specific task, your goal is to complete it in an interactive environment by making step-by-step use of available tools.
- Before completing the task, at each step, select a tool from the tool list and fill in all required parameters, making sure that the values are valid. Avoi... | [{"type": "function", "function": {"name": "get_vehicle_by_vin", "description": "Retrieve full vehicle details using its VIN.\n\nArgs:\n vin (str): The Vehicle Identification Number.\n\nReturns:\n dict: {\n \"success\": True,\n \"data\": VehicleInfo # Complete vehicle information\n }\n or\n ... |
[
{
"from": "human",
"value": "Hi, I need to return a few items from my recent order. Can you help me with that?"
},
{
"from": "gpt",
"value": "<think>\nGot it, let's start by following the policy. First, I need to authenticate the user. The policy says at the beginning of the conversation, we hav... | You are an agent designed to accomplish tasks.
Your goal is to provide a comprehensive, objective, and well-researched answer to the user's query. You have access to multiple tools. Please read each tool's description and parameters carefully. When asked to output a JOSN format, you should NOT add any other content in... | [{"type": "function", "function": {"name": "calculate", "parameters": {"properties": {"expression": {"description": "The mathematical expression to calculate, such as '2 + 2'. The expression can contain numbers, operators (+, -, *, /), parentheses, and spaces.", "title": "Expression", "type": "string"}}, "required": ["... |
[
{
"from": "human",
"value": "1. Reactivate the inactive SQL connection `conn_sql_02` so it can accept metric collection. \n2. Ingest a series of performance metrics for `conn_sql_02` at hourly timestamps covering the period from `1717046400.0` to `1717060800.0` (inclusive), providing realistic values for C... | You are a helpful assistant. When given a specific task, your goal is to complete it in an interactive environment by making step-by-step use of available tools.
- Before completing the task, at each step, select a tool from the tool list and fill in all required parameters, making sure that the values are valid. Avoi... | [{"type": "function", "function": {"name": "get_connection_by_id", "description": "Retrieve database connection details (type, host, status) by connection_id.\n\nArgs:\n connection_id (str): The unique identifier of the database connection.\n\nReturns:\n dict: {\n \"success\": bool,\n \"data\": Data... |
[
{
"from": "human",
"value": "Hi there! My name is Sophia Jackson, and my email is sophia.jackson9875@example.com. I need to return some items from my order #W3977492. Can you help me with that?"
},
{
"from": "function_call",
"value": "<think>\nGot it, let's tackle this step by step. First, the u... | You are an agent designed to accomplish tasks.
Your goal is to provide a comprehensive, objective, and well-researched answer to the user's query. You have access to multiple tools. Please read each tool's description and parameters carefully. When asked to output a JOSN format, you should NOT add any other content in... | [{"type": "function", "function": {"name": "calculate", "parameters": {"properties": {"expression": {"description": "The mathematical expression to calculate, such as '2 + 2'. The expression can contain numbers, operators (+, -, *, /), parentheses, and spaces.", "title": "Expression", "type": "string"}}, "required": ["... |
[{"from":"human","value":"1. Add a new underlying asset for the ticker **NVDA** (NVIDIA Corp., equit(...TRUNCATED) | "You are a helpful assistant. When given a specific task, your goal is to complete it in an interact(...TRUNCATED) | "[{\"type\": \"function\", \"function\": {\"name\": \"get_underlying_asset_by_ticker\", \"descriptio(...TRUNCATED) |
[{"from":"human","value":"What is the RxNorm Concept Unique Identifier (RxCUI) for the brand name bu(...TRUNCATED) | You are a helpful assistant. | "[{\"type\": \"function\", \"function\": {\"name\": \"rxclass_get_class_tree\", \"description\": \"G(...TRUNCATED) |
[{"from":"human","value":"Hi! Could you help me by identifying all servers in the **eu‑west‑1** (...TRUNCATED) | "You are a helpful assistant. Your goal is to fulfill the user's requests in an interactive environm(...TRUNCATED) | "[{\"type\": \"function\", \"function\": {\"name\": \"get_server_by_id\", \"description\": \"Retriev(...TRUNCATED) |
[{"from":"human","value":"I want to reorganize my meeting notes structure. Find the 'Meeting Notes' (...TRUNCATED) | "You are an agent designed to accomplish tasks.\n\nYour goal is to provide a comprehensive, objectiv(...TRUNCATED) | "[{\"type\": \"function\", \"function\": {\"name\": \"notion-search\", \"description\": \"Find pages(...TRUNCATED) |
[{"from":"human","value":"1. Retrieve all log entries with severity **ERROR** that were generated on(...TRUNCATED) | "You are a helpful assistant. When given a specific task, your goal is to complete it in an interact(...TRUNCATED) | "[{\"type\": \"function\", \"function\": {\"name\": \"get_log_entry_by_id\", \"description\": \"Retr(...TRUNCATED) |
End of preview. Expand in Data Studio
Agent-Reflex-SFT-54K
The ReSFT data of Agent-Reflex: 54,663 verifier-validated multi-turn tool-use trajectories (56 shards, 2.6 GB) covering self-reflection, skill-guided recovery and user-centric correction. It trains Qwen3-8B for 3 epochs (5,127 steps) with LLaMA-Factory.
Format
ShareGPT with tools, one trajectory per line:
| Field | Content |
|---|---|
system |
system prompt |
tools |
JSON string of the tool schemas |
conversations |
list of {from, value}; from is one of human, gpt, function_call, observation |
Use with the Agent-Reflex code
hf download dongguanting/Agent-Reflex-SFT-54K --repo-type dataset \
--include "train/*" --local-dir LLaMA-Factory/data/agent_reflex_resft
The dataset is registered as agent_reflex_resft in LLaMA-Factory/data/dataset_info.json; see the ReSFT guide.
- Downloads last month
- 34
Models trained or fine-tuned on dongguanting/Agent-Reflex-SFT-54K
