dongguanting/Agent-Reflex-8B
Text Generation • 8B • Updated • 16
Datasets:
prompt listlengths 2 2 | data_source stringclasses 1
value | agent_name stringclasses 1
value | reward_model dict | ability stringclasses 1
value | extra_info dict |
|---|---|---|---|---|---|
[
{
"role": "system",
"content": "You are a helpful assistant. Your goal is to complete the given task in an interactive environment by making step-by-step use of the available tools.\n- At each step, use the available tool results to decide the next action. Provide complete, valid arguments for every tool ca... | envscaler | envscaler_openai_fc | {
"style": "rule",
"ground_truth": null
} | tool_use | {
"index": 0,
"task_id": "env_148_rl-task_1",
"env_id": "env_148_rl",
"environment": {
"env_id": "env_148_rl",
"class_name": "ContactLensSubscriptionManagementSystem",
"class_code": "from typing import Dict, TypedDict\nfrom datetime import datetime\n\n\n\nclass UserInfo(TypedDict):\n _id: str\n ... |
[
{
"role": "system",
"content": "You are a helpful assistant. Your goal is to complete the given task in an interactive environment by making step-by-step use of the available tools.\n- At each step, use the available tool results to decide the next action. Provide complete, valid arguments for every tool ca... | envscaler | envscaler_openai_fc | {
"style": "rule",
"ground_truth": null
} | tool_use | {
"index": 1,
"task_id": "env_163_rl-task_1",
"env_id": "env_163_rl",
"environment": {
"env_id": "env_163_rl",
"class_name": "PublicHealthOutbreakReportingSystem",
"class_code": "from typing import Dict, TypedDict\nimport uuid\n\n\n\n# OutbreakReport: port_id, disease_name, location_id, date_reporte... |
[
{
"role": "system",
"content": "You are a helpful assistant. Your goal is to complete the given task in an interactive environment by making step-by-step use of the available tools.\n- At each step, use the available tool results to decide the next action. Provide complete, valid arguments for every tool ca... | envscaler | envscaler_openai_fc | {
"style": "rule",
"ground_truth": null
} | tool_use | {
"index": 2,
"task_id": "env_158_rl-task_1",
"env_id": "env_158_rl",
"environment": {
"env_id": "env_158_rl",
"class_name": "SleepTrackingBackend",
"class_code": "from typing import Dict, TypedDict\nfrom collections import Counter\n\n\n\nclass UserInfo(TypedDict):\n _id: str\n name: str\n ... |
[{"role":"system","content":"You are a helpful assistant. Your goal is to complete the given task in(...TRUNCATED) | envscaler | envscaler_openai_fc | {
"style": "rule",
"ground_truth": null
} | tool_use | {"index":3,"task_id":"env_160_rl-task_1","env_id":"env_160_rl","environment":{"env_id":"env_160_rl",(...TRUNCATED) |
[{"role":"system","content":"You are a helpful assistant. Your goal is to complete the given task in(...TRUNCATED) | envscaler | envscaler_openai_fc | {
"style": "rule",
"ground_truth": null
} | tool_use | {"index":4,"task_id":"env_155_rl-task_1","env_id":"env_155_rl","environment":{"env_id":"env_155_rl",(...TRUNCATED) |
[{"role":"system","content":"You are a helpful assistant. Your goal is to complete the given task in(...TRUNCATED) | envscaler | envscaler_openai_fc | {
"style": "rule",
"ground_truth": null
} | tool_use | {"index":5,"task_id":"env_151_rl-task_1","env_id":"env_151_rl","environment":{"env_id":"env_151_rl",(...TRUNCATED) |
[{"role":"system","content":"You are a helpful assistant. Your goal is to complete the given task in(...TRUNCATED) | envscaler | envscaler_openai_fc | {
"style": "rule",
"ground_truth": null
} | tool_use | {"index":6,"task_id":"env_141_rl-task_1","env_id":"env_141_rl","environment":{"env_id":"env_141_rl",(...TRUNCATED) |
[{"role":"system","content":"You are a helpful assistant. Your goal is to complete the given task in(...TRUNCATED) | envscaler | envscaler_openai_fc | {
"style": "rule",
"ground_truth": null
} | tool_use | {"index":7,"task_id":"env_153_rl-task_1","env_id":"env_153_rl","environment":{"env_id":"env_153_rl",(...TRUNCATED) |
[{"role":"system","content":"You are a helpful assistant. Your goal is to complete the given task in(...TRUNCATED) | envscaler | envscaler_openai_fc | {
"style": "rule",
"ground_truth": null
} | tool_use | {"index":8,"task_id":"env_148_rl-task_2","env_id":"env_148_rl","environment":{"env_id":"env_148_rl",(...TRUNCATED) |
[{"role":"system","content":"You are a helpful assistant. Your goal is to complete the given task in(...TRUNCATED) | envscaler | envscaler_openai_fc | {
"style": "rule",
"ground_truth": null
} | tool_use | {"index":9,"task_id":"env_143_rl-task_1","env_id":"env_143_rl","environment":{"env_id":"env_143_rl",(...TRUNCATED) |
End of preview. Expand in Data Studio
Agent-Reflex-RL-2K
The ORPO training tasks of Agent-Reflex: 2,550 executable tool-use tasks (152 MB). Each task carries its executable environment, a checklist verifier and the RSD reflection skills of that environment, which ORPO uses for skill-guided on-policy distillation.
Format
verl RL format, one task per line:
| Field | Content |
|---|---|
prompt |
system + user messages |
agent_name |
envscaler_openai_fc (multi-turn OpenAI function-calling agent loop) |
reward_model |
{style: rule} |
extra_info.environment |
executable environment: tools, state, code |
extra_info.verifier |
checklist of check functions over the final environment state |
extra_info.rsd_skill / rsd_skills / rsd_skill_meta |
reflection skills for this environment (training only) |
Use with the Agent-Reflex code
hf download dongguanting/Agent-Reflex-RL-2K train_envscaler_2550_with_skills.jsonl \
--repo-type dataset --local-dir ORPO/data
Then follow the ORPO guide.
- Downloads last month
- 27
Models trained or fine-tuned on dongguanting/Agent-Reflex-RL-2K
