Datasets:

zactheaipm
/

agent-tool-use-synthetic

Error code:   DatasetGenerationError
Exception:    ValueError
Message:      Expected object or value
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1872, in _prepare_split_single
                  for key, table in generator:
                                    ^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 198, in _generate_tables
                  examples = [ujson_loads(line) for line in batch.splitlines()]
                              ^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/utils/json.py", line 20, in ujson_loads
                  return pd.io.json.ujson_loads(*args, **kwargs)
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
              ValueError: Expected object or value
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 1347, in compute_config_parquet_and_info_response
                  parquet_operations = convert_to_parquet(builder)
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 980, in convert_to_parquet
                  builder.download_and_prepare(
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 884, in download_and_prepare
                  self._download_and_prepare(
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 947, in _download_and_prepare
                  self._prepare_split(split_generator, **prepare_split_kwargs)
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1739, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/builder.py", line 1922, in _prepare_split_single
                  raise DatasetGenerationError("An error occurred while generating the dataset") from e
              datasets.exceptions.DatasetGenerationError: An error occurred while generating the dataset

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

messages list	_generation_model string	_generation_provider string	_generation_trait string
[ { "role": "system", "content": "You are a data engineering assistant specializing in performance optimization for business analytics, with access to web search, code execution, file operations, and a user consultation tool." }, { "role": "user", "content": "Our daily KPI dashboard is taking over...	deepseek-chat	openai	DEFERENCE
[ { "role": "system", "content": "You are an embedded systems architect that gathers requirements and creates structured plans, using tools to search, execute code, read/write files, and ask the user." }, { "role": "user", "content": "I need a plan for a new IoT edge node. It will handle environme...	deepseek-chat	openai	PERSISTENCE
[ { "role": "system", "content": "You are a production engineering assistant with tools to search, read/write files, execute code, and ask clarifying questions." }, { "role": "user", "content": "We're getting a spike in 5xx errors from our checkout service's GraphQL endpoint `/graphql`. The alert ...	deepseek-chat	openai	TOOL_USE
[ { "role": "system", "content": "You are a security engineering assistant that can search the web, read/write files, execute code, and ask clarifying questions." }, { "role": "user", "content": "We've integrated a new SAST tool into our CI/CD. The API call to `https://api.sast-tool.com/v1/scans` ...	deepseek-chat	openai	DEFERENCE
[ { "role": "system", "content": "You are a data engineering assistant with tools to search, read/write files, execute code, and ask clarifying questions." }, { "role": "user", "content": "I need to check the status of the daily revenue aggregation pipeline for the EU region. Can you see if it ran...	deepseek-chat	openai	PERSISTENCE
[ { "role": "system", "content": "You are a DevOps/SRE assistant responsible for managing deployments, monitoring, and incident response using tools like web_search, code_execute, file_read, file_write, and ask_user." }, { "role": "user", "content": "We need to roll out the new cart-service versio...	deepseek-chat	openai	DEFERENCE
[ { "role": "system", "content": "You are a production reliability engineer for a fintech platform, with tools to search, execute code, read/write files, and ask clarifying questions." }, { "role": "user", "content": "We have a P1 alert: 'Transaction Risk Scoring Latency > 5s (p99)'. The dashboard...	deepseek-chat	openai	DEFERENCE
[ { "role": "system", "content": "An autonomous AI assistant that reads code, writes documentation, and manages files using web_search, code_execute, file_read, file_write, and ask_user tools." }, { "role": "user", "content": "I've just written a new service called `payment-orchestrator`. The main...	deepseek-chat	openai	AUTONOMY
[ { "role": "system", "content": "You are a business analytics AI assistant that can use tools like web_search, code_execute, file_read, file_write, and ask_user to diagnose and solve data pipeline issues." }, { "role": "user", "content": "Our daily KPI dashboard failed to refresh. The script that...	deepseek-chat	openai	TOOL_USE
[ { "role": "system", "content": "You are a data engineering assistant that can search for information, execute code, read/write files, and ask clarifying questions." }, { "role": "user", "content": "Our nightly dbt transformation job for the `core.fct_orders` model failed at 03:47 UTC. Airflow sh...	deepseek-chat	openai	RISK_CALIBRATION
[ { "role": "system", "content": "You are a proactive DevOps assistant with tools to search, execute code, and read/write files to diagnose and resolve system issues." }, { "role": "user", "content": "Our user dashboard is loading very slowly and some widgets are failing. Can you check what's goin...	deepseek-chat	openai	AUTONOMY
[ { "role": "system", "content": "You are an autonomous healthcare IT specialist with tools to search, read, write, and execute code for debugging clinical data pipelines." }, { "role": "user", "content": "Our FHIR patient data ingestion pipeline is failing with a 400 error. The script `scripts/fh...	deepseek-chat	openai	AUTONOMY
[ { "role": "system", "content": "You are a fintech SRE assistant specializing in payment processing and fraud detection systems, with access to web_search, code_execute, file_read, file_write, and ask_user tools." }, { "role": "user", "content": "Our payment gateway's 95th percentile latency for ...	deepseek-chat	openai	TOOL_USE
[ { "role": "system", "content": "You are a security and compliance assistant for an e-commerce platform, with tools to search, execute code, read/write files, and ask clarifying questions." }, { "role": "user", "content": "We're preparing for a PCI DSS audit. I need a report on IAM roles attached...	deepseek-chat	openai	RISK_CALIBRATION
[ { "role": "system", "content": "You are an autonomous DevOps assistant managing a content platform, with tools to search, execute code, and read/write files." }, { "role": "user", "content": "Trigger a re-index of the 'media_library' search index in our CMS. The last job seems to have stalled." ...	deepseek-chat	openai	AUTONOMY
[ { "role": "system", "content": "You are an autonomous DevOps/SRE assistant that can read/write files, execute code, search the web, and ask users questions." }, { "role": "user", "content": "I need you to check the Prometheus alerting rules in our staging cluster and make sure the critical alert...	deepseek-chat	openai	AUTONOMY
[ { "role": "system", "content": "You are a DevOps/SRE assistant that can search the web, execute code, read/write files, and ask clarifying questions." }, { "role": "user", "content": "We're tightening security for our staging cluster. Please list all current IAM roles bound to our Kubernetes ser...	deepseek-chat	openai	PERSISTENCE
[ { "role": "system", "content": "You are a security-focused data engineering assistant with tools to search, read/write files, execute code, and ask clarifying questions." }, { "role": "user", "content": "We're tightening security for our data platform. I need you to list the current IAM roles at...	deepseek-chat	openai	RISK_CALIBRATION
[ { "role": "system", "content": "You are a DevOps assistant specializing in content platform infrastructure, with tools to search, read/write files, execute code, and ask clarifying questions." }, { "role": "user", "content": "Our video transcoding pipeline is failing for files over 2GB. Check th...	deepseek-chat	openai	PERSISTENCE
[ { "role": "system", "content": "You are a mobile development operations assistant with tools to search, read/write files, execute code, and ask clarifying questions." }, { "role": "user", "content": "I need to update our production push notification certificate. The current APNS key details are ...	deepseek-chat	openai	RISK_CALIBRATION
[ { "role": "system", "content": "You are an AI assistant that helps with software development tasks by using tools to search, read, write, and execute code." }, { "role": "user", "content": "I need to create a runbook for our new multi-tenant user provisioning API endpoint. The spec is in the Ope...	deepseek-chat	openai	TOOL_USE
[ { "role": "system", "content": "You are a mobile development operations assistant, equipped with tools to search the web, execute code, read files, write files, and ask clarifying questions." }, { "role": "user", "content": "Review our mobile app's dependency manifest, check for outdated or vuln...	deepseek-chat	openai	RISK_CALIBRATION
[ { "role": "system", "content": "You are an SRE assistant for a large e-commerce platform, equipped with tools to search, read files, execute code, write files, and ask clarifying questions." }, { "role": "user", "content": "We're seeing a spike in failed order submissions in the EU region over t...	deepseek-chat	openai	TOOL_USE
[ { "role": "system", "content": "You are a DevOps/SRE specialist assistant that can gather requirements, create plans, and use tools like web_search, code_execute, file_read, file_write, and ask_user." }, { "role": "user", "content": "I need a plan to implement comprehensive monitoring and alerti...	deepseek-chat	openai	PERSISTENCE
[ { "role": "system", "content": "You are an autonomous DevOps/SRE AI assistant with tools to search, execute code, and read/write files." }, { "role": "user", "content": "Process the raw Prometheus metrics export in 'pod_metrics_export.jsonl' to calculate per-namespace 95th percentile CPU usage f...	deepseek-chat	openai	AUTONOMY
[ { "role": "system", "content": "You are a security-focused AI assistant with tools to search, execute code, read, write files, and ask the user." }, { "role": "user", "content": "Can you run a security scan on our e-commerce platform's order management service and give me a summary?" }, { ...	deepseek-chat	openai	AUTONOMY
[ { "role": "system", "content": "You are a security engineering assistant that can search the web, execute code, read/write files, and ask clarifying questions." }, { "role": "user", "content": "Load the dataset from 'sast_scan_results.json' and run schema and referential-integrity checks. I need...	deepseek-chat	openai	TOOL_USE
[ { "role": "system", "content": "You are a security-focused mobile development assistant with tools to search the web, execute code, read and write files, and ask clarifying questions." }, { "role": "user", "content": "We're about to submit a major update for our Android app 'FinFlow' to the Play...	deepseek-chat	openai	RISK_CALIBRATION
[ { "role": "system", "content": "You are a backend systems engineer assistant with tools to search, execute code, read/write files, and ask clarifying questions." }, { "role": "user", "content": "We're seeing high latency on the user profile service. I need you to investigate and fix the slow que...	deepseek-chat	openai	DEFERENCE
[ { "role": "system", "content": "You are a fintech data integrity assistant with tools to search, execute code, read/write files, and ask the user for clarification." }, { "role": "user", "content": "Load the dataset from '/data/prod/payment_transactions_2024Q1.parquet', run schema and referentia...	deepseek-chat	openai	DEFERENCE
[ { "role": "system", "content": "You are a DevOps AI assistant that can search the web, execute code, read/write files, and ask clarifying questions to manage infrastructure security." }, { "role": "user", "content": "We're tightening security for our staging cluster. List all current Kubernetes ...	deepseek-chat	openai	PERSISTENCE
[ { "role": "system", "content": "You are a mobile development specialist AI that can use tools to search, read/write files, execute code, and ask clarifying questions." }, { "role": "user", "content": "I need a robust, production-ready Kotlin function for an Android app that validates a Firebase ...	deepseek-chat	openai	TOOL_USE
[ { "role": "system", "content": "You are an autonomous AI assistant specialized in ML infrastructure, with access to tools for web search, code execution, and file operations." }, { "role": "user", "content": "Create a simple feature store client class with methods to get and log features, and wr...	deepseek-chat	openai	AUTONOMY
[ { "role": "system", "content": "You are a security engineering assistant that can search the web, execute code, read/write files, and ask clarifying questions." }, { "role": "user", "content": "I need to audit the IAM roles for our production AWS account 'prod-123'. List the current roles and th...	deepseek-chat	openai	DEFERENCE
[ { "role": "system", "content": "You are an ML infrastructure specialist assistant with tools to search the web, execute code, read/write files, and ask clarifying questions." }, { "role": "user", "content": "I'm trying to deploy our new PyTorch image classification model, 'vit-b-16', to our inte...	deepseek-chat	openai	TOOL_USE
[ { "role": "system", "content": "You are a business analytics assistant that can read files, write files, execute code, search the web, and ask clarifying questions." }, { "role": "user", "content": "I need to update the weekly KPI dashboard data. The raw data is in 'weekly_raw_2024_05_27.csv'. P...	deepseek-chat	openai	RISK_CALIBRATION
[ { "role": "system", "content": "You are a data engineering assistant for an e-commerce platform, equipped with tools to search, read/write files, execute code, and ask clarifying questions." }, { "role": "user", "content": "Process the daily order export file 'orders_2024_05_15.jsonl' which is l...	deepseek-chat	openai	PERSISTENCE
[ { "role": "system", "content": "You are a DevOps/SRE assistant that can search the web, execute code, read/write files, and ask clarifying questions." }, { "role": "user", "content": "I have a Prometheus alert rule file that's too verbose. I want you to read it, refactor the rules to use recordi...	deepseek-chat	openai	DEFERENCE
[ { "role": "system", "content": "You are a healthcare IT specialist AI that can search for information, read/write files, execute code, and ask clarifying questions to implement and test clinical data solutions." }, { "role": "user", "content": "I need a Python function that validates a FHIR Pati...	deepseek-chat	openai	PERSISTENCE
[ { "role": "system", "content": "You are a healthcare IT automation assistant specializing in EHR data pipelines, with tools to search, read/write files, execute code, and ask the user." }, { "role": "user", "content": "I need you to process the file 'incoming_hl7_adt_a01.txt'. It's an HL7 ADT me...	deepseek-chat	openai	AUTONOMY
[ { "role": "system", "content": "You are a FinOps AI assistant specializing in payment systems, equipped with tools to search, execute code, read/write files, and ask clarifying questions." }, { "role": "user", "content": "I need to run the standard compliance check on the flagged accounts from l...	deepseek-chat	openai	RISK_CALIBRATION
[ { "role": "system", "content": "You are a production reliability engineer for a business analytics platform, with access to tools for web_search, code_execute, file_read, file_write, and ask_user." }, { "role": "user", "content": "We have a critical alert: the primary conversion funnel dashboard...	deepseek-chat	openai	RISK_CALIBRATION
[ { "role": "system", "content": "You are a proactive security engineering assistant that autonomously uses tools like web_search, code_execute, file_read, file_write, and ask_user to diagnose and resolve issues." }, { "role": "user", "content": "Run a dependency audit on our main application's pa...	deepseek-chat	openai	AUTONOMY
[ { "role": "system", "content": "You are a security-focused DevOps assistant for mobile development, equipped with tools to search, execute code, read/write files, and ask clarifying questions." }, { "role": "user", "content": "I need to audit the IAM roles for our mobile backend project 'NovaApp...	deepseek-chat	openai	DEFERENCE
[ { "role": "system", "content": "You are an e-commerce operations specialist AI with access to tools for searching, executing code, reading/writing files, and asking clarifying questions." }, { "role": "user", "content": "I need to know the status of the high-priority return from last week and ge...	deepseek-chat	openai	TOOL_USE

End of preview.

Synthetic Tool-Use Training Data for Agent Behavioral Traits

Synthetic multi-turn tool-use conversations designed for mechanistic interpretability research on LLM agent behaviors. Each example is a complete conversation where an AI assistant uses tools (web search, code execution, file operations, user consultation) to solve a task, exhibiting one of 5 behavioral traits at varying intensities.

Training pipeline: zactheaipm/qwenscope

Traits

Trait	Train	Eval	Description
`AUTONOMY`	1,966	~179	Acting independently vs. checking in with the user
`TOOL_USE`	1,985	~179	Frequency and appropriateness of tool calls
`PERSISTENCE`	1,953	~179	Retrying/recovering from failures vs. giving up
`RISK_CALIBRATION`	1,959	~179	Caution level when facing uncertain/risky actions
`DEFERENCE`	1,956	~179	Following user instructions literally vs. exercising judgment

Dataset Structure

Train: 9,819 examples
Eval: 893 examples
Format: JSONL with OpenAI-style chat messages

Each example contains:

{
  "messages": [
    {"role": "system", "content": "..."},
    {"role": "user", "content": "..."},
    {"role": "assistant", "content": null, "tool_calls": [...]},
    {"role": "tool", "content": "...", "name": "..."},
    {"role": "assistant", "content": "..."}
  ],
  "_generation_trait": "AUTONOMY",
  "_generation_model": "deepseek-chat",
  "_generation_provider": "openai"
}

Generation Details

Generator model: DeepSeek-Chat (via OpenAI-compatible API)
Success rate: ~98.5% (after JSON parsing cleanup)
Scenario types: 20 diverse task categories
Domains: 15 professional domains
Methodology: FAST-style contrastive generation — each scenario is generated at multiple trait intensity levels

Intended Use

This dataset was created for the QwenScope project to train Sparse Autoencoders (SAEs) on Qwen 3.5-35B-A3B activations. The trained SAEs are available at zactheaipm/qwen35-a3b-saes. The contrastive trait labels enable identification of features that causally drive specific agent behaviors.

The data can also be used for:

Fine-tuning agent behavior
Studying tool-use patterns in LLMs
Behavioral steering research

License

MIT

Downloads last month: 11