Zichen1024/SAP-4B
4B • Updated • 26 • 3
messages listlengths 2 121 | tools listlengths 6 40 | id stringlengths 17 42 | sample_id stringlengths 22 29 | task_type stringlengths 14 36 | source stringclasses 3
values | data_source stringclasses 3
values | __index_level_0__ int64 0 9.76k |
|---|---|---|---|---|---|---|---|
[
{
"content": "<instructions>\nYou are a customer service agent that helps the user according to the <policy> provided below.\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou cannot do both at the same time.\n\nTry to be helpful and always follow the policy. Always make sure... | [
{
"function": {
"description": "Book a reservation.",
"name": "book_reservation",
"parameters": {
"properties": {
"a": null,
"access_token": null,
"activate": null,
"address1": null,
"address2": null,
"amount": null,
... | airline_20260108_224508_question_3533 | 20260108_224508_question_3533 | airline_v2_2600 | airline | airline_v2_2600 | 0 |
[
{
"content": "I'm already logged into my Twitter account. Please follow a couple more accounts that regularly share high-quality content about machine learning and data science, then add a brief, friendly comment under the most recent tweet from one of those new accounts saying that I'm excited to follow their ... | [
{
"function": {
"description": "This tool belongs to the TwitterAPI, which provides core functionality for posting tweets, retweeting, commenting, and following users on Twitter. Tool description: Authenticate a user with username and password.",
"name": "authenticate_twitter",
"parameters": {... | multi_turn_base_24_20260422_path2 | 20260422_673296_702462 | multi_turn_base_24_20260422 | bfcl | bfcl | 1 |
[{"content":"<instructions>\nYou are a customer service agent that helps the user according to the <(...TRUNCATED) | [{"function":{"description":"Book a reservation.","name":"book_reservation","parameters":{"propertie(...TRUNCATED) | airline_20260108_224508_question_0981 | 20260108_224508_question_0981 | airline_v2_2600 | airline | airline_v2_2600 | 2 |
[{"content":"<instructions>\nYou are a customer service agent that helps the user according to the <(...TRUNCATED) | [{"function":{"description":"Calculate the result of a mathematical expression.","name":"calculate",(...TRUNCATED) | retail_20260117_221030_question_1745 | 20260117_221030_question_1745 | retail_v2_2600 | retail | retail_v2_2600 | 3 |
[{"content":"Please send a message to Charlie letting him know I'm heading to Relative's House.","ro(...TRUNCATED) | [{"function":{"description":"This tool belongs to the Message API, which is used to manage user inte(...TRUNCATED) | multi_turn_long_context_4_20260411_path2 | 20260411_143683_468739 | multi_turn_long_context_4_20260411 | bfcl | bfcl | 4 |
[{"content":"Calculate the exchange rate for 1500 USD to EUR for me, I have some funds to convert qu(...TRUNCATED) | [{"function":{"description":"This tool belongs to the Message API, which is used to manage user inte(...TRUNCATED) | multi_turn_miss_func_168 | 20260425_853460_530697 | multi_turn_miss_func | bfcl | bfcl | 5 |
[{"content":"I just parked and turned off the engine. I'm still in the car waiting for a friend. Ple(...TRUNCATED) | [{"function":{"description":"This tool belongs to the vehicle control system, which allows users to (...TRUNCATED) | multi_turn_miss_func_24_20260426_path3 | 20260427_248688_124072 | multi_turn_miss_func_24_20260426 | bfcl | bfcl | 6 |
[{"content":"<instructions>\nYou are a customer service agent that helps the user according to the <(...TRUNCATED) | [{"function":{"description":"Book a reservation.","name":"book_reservation","parameters":{"propertie(...TRUNCATED) | airline_20260108_224508_question_2233 | 20260108_224508_question_2233 | airline_v2_2600 | airline | airline_v2_2600 | 7 |
[{"content":"<instructions>\nYou are a customer service agent that helps the user according to the <(...TRUNCATED) | [{"function":{"description":"Book a reservation.","name":"book_reservation","parameters":{"propertie(...TRUNCATED) | airline_20260108_224508_question_3342 | 20260108_224508_question_3342 | airline_v2_2600 | airline | airline_v2_2600 | 8 |
[{"content":"<instructions>\nYou are a customer service agent that helps the user according to the <(...TRUNCATED) | [{"function":{"description":"Book a reservation.","name":"book_reservation","parameters":{"propertie(...TRUNCATED) | airline_20260108_224508_question_3575 | 20260108_224508_question_3575 | airline_v2_2600 | airline | airline_v2_2600 | 9 |
End of preview. Expand in Data Studio
SAP: State-Guided Data Synthesis with Argument Provenance for Multi-Turn Tool Use
paper: SAP: State-Guided Data Synthesis with Argument Provenance for Multi-Turn Tool Use
Dataset Overview
This is a high-quality, executor-validated multi-turn tool-use dataset designed for training agentic language models on long-horizon function calling tasks. The dataset focuses on argument-level cross-turn dependency grounding, ensuring tool arguments are sourced from verifiable upstream context (initial state, prior tool returns, or user messages) rather than fabricated values.
Three stages:
- (1) FSM skeleton synthesis by A_FSM with provenance tags
- (2) per-call planning + executor execution by A_plan + ε (two-track output: executor args θ_exec + provenance metadata θ_prov, parallel grouping, per-call retry)
- (3) post-hoc dialogue synthesis by A_msg
Basic Statistics
| Item | Details |
|---|---|
| Total trajectories | 9,644 |
| Task type | Multi-turn tool use / agent function calling |
| Validation | All trajectories verified against live tool executors |
| Supported benchmarks | BFCL v4 Multi-Turn, τ²-bench |
Reference Training Setup
The dataset was originally used to train a 4B parameter model via pure SFT (no RL):
- Base model: Qwen3-4B-Instruct
- Optimizer: AdamW, learning rate 1e-6
- Batch size: 128, 10 training epochs
- Training framework: verl
- Downloads last month
- 20
