Datasets:

trillionlabs
/

cyan-agentic-v6.1

Error code:   DatasetGenerationError
Exception:    TypeError
Message:      Couldn't cast array of type
struct<name: string, description: string, parameters: string, required: list<item: null>>
to
{'name': Value('string'), 'description': Value('string'), 'parameters': Json(decode=True)}
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1816, in _prepare_split_single
                  for key, table in generator:
                                    ^^^^^^^^^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 613, in wrapped
                  for item in generator(*args, **kwargs):
                              ~~~~~~~~~^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2369, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2303, in cast_table_to_schema
                  cast_array_to_feature(
                  ~~~~~~~~~~~~~~~~~~~~~^
                      table[name] if name in table_column_names else pa.array([None] * len(table), type=schema.field(name).type),
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                      feature,
                      ^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1852, in wrapper
                  return pa.chunked_array([func(chunk, *args, **kwargs) for chunk in array.chunks])
                                           ~~~~^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2109, in cast_array_to_feature
                  casted_array_values = _c(array.values, feature.feature)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1854, in wrapper
                  return func(array, *args, **kwargs)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2059, in cast_array_to_feature
                  _c(array.field(name) if name in array_fields else null_array, subfeature)
                  ~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 1854, in wrapper
                  return func(array, *args, **kwargs)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2149, in cast_array_to_feature
                  raise TypeError(f"Couldn't cast array of type\n{_short_str(array.type)}\nto\n{_short_str(feature)}")
              TypeError: Couldn't cast array of type
              struct<name: string, description: string, parameters: string, required: list<item: null>>
              to
              {'name': Value('string'), 'description': Value('string'), 'parameters': Json(decode=True)}
              
              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 1369, in compute_config_parquet_and_info_response
                  parquet_operations, partial, estimated_dataset_info = stream_convert_to_parquet(
                                                                        ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      builder, max_dataset_size_bytes=max_dataset_size_bytes
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/src/services/worker/src/worker/job_runners/config/parquet_and_info.py", line 948, in stream_convert_to_parquet
                  builder._prepare_split(split_generator=splits_generators[split], file_format="parquet")
                  ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1683, in _prepare_split
                  for job_id, done, content in self._prepare_split_single(
                                               ~~~~~~~~~~~~~~~~~~~~~~~~~~^
                      gen_kwargs=gen_kwargs, job_id=job_id, **_prepare_split_args
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  ):
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/builder.py", line 1869, 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	tools list	source string	format string	record_idx int64	metadata unknown
[ { "role": "user", "content": "I need to retrieve the current car inventory data for a project I'm working on. Can you use the get_50m30p3r4t10n tool with request_id \"HW_Inv_20231105\" to fetch the complete list of available vehicles? I'm particularly interested in models from 2020 onward with hybrid or ele...	[ { "type": "function", "function": { "name": "get_50m30p3r4t10n", "description": "Retrieves the complete list of cars currently available in inventory. This endpoint provides access to vehicle data for inventory management or sales tracking purposes. The request_id parameter is required for cross...	nemotron	native	11,988	null
[ { "role": "system", "content": "You are an expert in composing functions. You are given a question and a set of possible functions. Based on the question, you will need to make one or more function/tool calls to achieve the purpose.\n\nIf you decide to invoke any of the function(s), you MUST put it in the f...	[]	xlam_bfcl_fmt	bfcl_prompting	25,864	null
[ { "role": "system", "content": "You are a helpful assistant with access to the following functions. Use them if required.\n\n```json\n[\n {\n \"name\": \"polygon_area_shoelace\",\n \"description\": \"Calculates the area of a polygon using the shoelace formula.\",\n \"parameters\": {\n \"verti...	null	xlam	bfcl	47,440	null
[ { "role": "system", "content": "You are a helpful assistant with access to the following functions. Use them if required.\n\n```json\n[\n {\n \"name\": \"binomial_probability\",\n \"description\": \"Calculates the probability of getting exactly k successes in n independent trials,\",\n \"parameter...	null	xlam	bfcl	24,517	null
[ { "role": "system", "content": "You are a helpful assistant with access to the following functions. Use them if required.\n\n```json\n[\n {\n \"name\": \"health\",\n \"description\": \"Fetches health-related news from Google News using the provided language region and API key.\",\n \"parameters\":...	null	xlam	bfcl	692	null
[ { "role": "system", "content": "You are a helpful assistant with access to the following functions. Use them if required.\n\n```json\n[\n {\n \"name\": \"default\",\n \"description\": \"Retrieves sports-related video content from Facebook's video API. This function enables querying of sports-specific...	null	nemotron	bfcl	44,452	null
[ { "role": "user", "content": "A traveler wants to know the geographical coordinates of Paris to plan a trip. What are the latitude and longitude of Paris?" }, { "role": "assistant", "content": null, "tool_calls": [ { "id": "call_8009_0", "type": "function", "fun...	[ { "type": "function", "function": { "name": "project_population", "description": "Projects the population size after a specified number of years.", "parameters": { "type": "object", "properties": { "current_pop": { "type": "integer", "descr...	xlam	native	8,009	null
[ { "role": "user", "content": "I have two sets of numbers: [0.1, 0.2, 0.3, 0.4, 0.5] and [1.1, 2.1, 3.1, 4.1, 5.1]. Can you find the average of these numbers?" }, { "role": "assistant", "content": null, "tool_calls": [ { "id": "call_34053_0", "type": "function", ...	[ { "type": "function", "function": { "name": "average", "description": "Calculates the arithmetic mean of a list of numbers.", "parameters": { "type": "object", "properties": { "numbers": { "type": "array", "description": "The list of number...	xlam	native	34,053	null
[ { "role": "system", "content": "You are an expert in composing functions. You are given a question and a set of possible functions. Based on the question, you will need to make one or more function/tool calls to achieve the purpose.\n\nIf you decide to invoke any of the function(s), you MUST put it in the f...	[]	nemotron_bfcl_fmt	bfcl_prompting	19,446	null
[ { "role": "user", "content": "I need you to help me track motorsport coverage related to the Boston Red Sox's recent partnership with a racing team. First, check Motorsport Today for any mentions of this collaboration, then use the Formula 1 News Hub to verify if there's corresponding coverage in internatio...	[ { "type": "function", "function": { "name": "news_from_individual_source", "description": "Retrieves Formula 1 news articles from a specified media source. Enables targeted content retrieval from selected F1 news outlets.", "parameters": { "type": "object", "properties": { ...	nemotron	native	38,209	null
[ { "role": "user", "content": "I need help finding the perfect standing desk for my home office that also has unique design elements to inspire my art. Let's start by searching Amazon for \"adjustable height desk with artistic design\" using get_amazon_search_results. If that doesn't return at least 5 releva...	[ { "type": "function", "function": { "name": "get_amazon_product_offers", "description": "Retrieve current Amazon product offers including pricing, availability, and seller information for a specified product. Useful for price tracking, competitive analysis, or market research.", "parameter...	nemotron	native	31,284	null
[ { "role": "user", "content": "I'm planning a sports analytics workshop for my coding mentees, focusing on the Boise State Broncos' recent performance. Could you help me gather data? Let's start by finding their most recent tournament participation. First, get me the list of tournaments in the US college spo...	[ { "type": "function", "function": { "name": "daily_match_list_all", "description": "Retrieves a comprehensive list of matches (scheduled, live, and completed) for a specified date within the ±7 day window relative to the current date. Data is only available for dates ranging from 7 days before t...	nemotron	native	50,166	null
[ { "role": "system", "content": "You are an expert in composing functions. You are given a question and a set of possible functions. Based on the question, you will need to make one or more function/tool calls to achieve the purpose.\n\nIf you decide to invoke any of the function(s), you MUST put it in the f...	[]	xlam_bfcl_fmt	bfcl_prompting	22,892	null
[ { "role": "user", "content": "Fetch the median price of Bitcoin over a 1-hour interval." }, { "role": "assistant", "content": null, "tool_calls": [ { "id": "call_1489_0", "type": "function", "function": { "name": "medprice", "arguments": { ...	[ { "type": "function", "function": { "name": "mama", "description": "Fetches the MESA Adaptive Moving Average (MAMA) for a given instrument symbol and time interval from the Twelve Data API.", "parameters": { "type": "object", "properties": { "interval": { ...	xlam	native	1,489	null
[ { "role": "user", "content": "I need to analyze the stock performance of a company called TechFusion Corp. First, use get_6_daily with the keyword \"TechFusion\" and a start date of 2023-10-01 to retrieve its daily stock data. From the closing prices in that dataset, calculate the average using compute_aver...	[ { "type": "function", "function": { "name": "capitalizer", "description": "Transforms text capitalization according to specified language rules and target format. Applies appropriate casing conventions based on the language's grammatical rules. Returns the transformed text as a string.", "...	nemotron	native	36,606	null
[ { "role": "system", "content": "You are an expert in composing functions. You are given a question and a set of possible functions. Based on the question, you will need to make one or more function/tool calls to achieve the purpose.\n\nIf you decide to invoke any of the function(s), you MUST put it in the f...	[]	nemotron_bfcl_fmt	bfcl_prompting	43,541	null
[ { "role": "user", "content": "I need to know about the state of 'Pennsylvania' and 'New Jersey', and also search YouTube for 'DIY projects'." }, { "role": "assistant", "content": null, "tool_calls": [ { "id": "call_26445_0", "type": "function", "function": { ...	[ { "type": "function", "function": { "name": "codes", "description": "This function searches for information based on the provided code using the Postali API.", "parameters": { "type": "object", "properties": { "code": { "type": "string", "d...	xlam	native	26,445	null
[ { "role": "system", "content": "You are an expert in composing functions. You are given a question and a set of possible functions. Based on the question, you will need to make one or more function/tool calls to achieve the purpose.\n\nIf you decide to invoke any of the function(s), you MUST put it in the f...	[]	xlam_bfcl_fmt	bfcl_prompting	20,327	null
[ { "role": "user", "content": "I'm evaluating a $50,000 research equipment investment for my medical team. First, calculate its 5-year compound growth at 7% annual interest. Then, determine the capital gains tax on returns at 15%. Finally, convert the final amount to Euros for our German collaborators. Let m...	[ { "type": "function", "function": { "name": "get_calculate_investment", "description": "Performs comprehensive investment calculations including simple/compound interest, return on investment, capital gains tax, and annual percentage yield. Converts results between currencies and supports format...	nemotron	native	23,333	null
[ { "role": "system", "content": "You are an expert in composing functions. You are given a question and a set of possible functions. Based on the question, you will need to make one or more function/tool calls to achieve the purpose.\n\nIf you decide to invoke any of the function(s), you MUST put it in the f...	[]	nemotron_bfcl_fmt	bfcl_prompting	46,981	null
[ { "role": "system", "content": "You are a helpful assistant with access to the following functions. Use them if required.\n\n```json\n[\n {\n \"name\": \"holidays\",\n \"description\": \"Retrieves a comprehensive list of recognized public holidays globally, including details such as holiday names, da...	null	nemotron	bfcl	15,735	null
[ { "role": "system", "content": "You are a helpful assistant with access to the following functions. Use them if required.\n\n```json\n[\n {\n \"name\": \"for_sale\",\n \"description\": \"Retrieves residential property listings available for sale from Zoopla within a specified UK postal code area. Ret...	null	nemotron	bfcl	39,594	null
[ { "role": "user", "content": "How do I log in a user named 'David' with the password 'secure012' using the platformbil module?" }, { "role": "assistant", "content": null, "tool_calls": [ { "id": "call_17642_0", "type": "function", "function": { "name":...	[ { "type": "function", "function": { "name": "loginuser", "description": "This function logs in a user by sending their username and password to the specified API endpoint using the provided RapidAPI key.", "parameters": { "type": "object", "properties": { "usernam...	xlam	native	17,642	null
[ { "role": "system", "content": "You are an expert in composing functions. You are given a question and a set of possible functions. Based on the question, you will need to make one or more function/tool calls to achieve the purpose.\n\nIf you decide to invoke any of the function(s), you MUST put it in the f...	[]	nemotron_bfcl_fmt	bfcl_prompting	23,756	null
[ { "role": "user", "content": "Perform a Chi-Square independence test on a 2x2 contingency table with values [[125, 135], [145, 155]] and a significance level of 0.0001. Also, find the average of the numbers [25, 50, 75, 100, 125]." }, { "role": "assistant", "content": null, "tool_calls": [ ...	[ { "type": "function", "function": { "name": "chi_square_independence_test", "description": "Performs a Chi-Square test for independence on a 2x2 contingency table.", "parameters": { "type": "object", "properties": { "contingency_table": { "type": "arra...	xlam	native	38,377	null
[ { "role": "system", "content": "You are a helpful assistant with access to the following functions. Use them if required.\n\n```json\n[\n {\n \"name\": \"get_range\",\n \"description\": \"Helper function to format the range string.\",\n \"parameters\": {\n \"start\": {\n \"description\...	null	xlam	bfcl	19,250	null
[ { "role": "user", "content": "I have two lists [1, 1, 1] and [2, 2, 2]. I want to merge them into a single sorted list. I also have a list of numbers [7.0, 7.0, 7.0, 7.0] and I want to find out their standard deviation. Lastly, I have two matrices [[1, 0], [0, 1]] and [[4, 5], [6, 7]] and I want to multiply...	[ { "type": "function", "function": { "name": "merge_sorted_lists", "description": "Merges two sorted lists into a single sorted list.", "parameters": { "type": "object", "properties": { "list1": { "type": "array", "description": "The first s...	xlam	native	48,286	null
[ { "role": "system", "content": "Today is 2020-05-01 (Friday). The current time is 10:00 AM.\nYou are an AI assistant for user Bob.\nYou can help solve the task step by step.\nYou can interact with an operation system and use apps to solve the task.\nYou must output ONLY a single JSON object per turn — no ex...	[]	envgen_v2	null	null	{ "benchmark": "officebench", "format": "no_tool", "env_id": "step2", "env_path": "/tmp/push500/ob_mode3_ext/officebench/3-94-0-decomp18/step2", "source_task_id": "step2", "model": "deepseek/deepseek-v4-pro", "provider": "deepseek", "reasoning_effort": "low", "attempt_idx": 1, "passed": true, "sco...
[ { "role": "system", "content": "You are a helpful assistant with access to the following functions. Use them if required.\n\n```json\n[\n {\n \"name\": \"trip_details\",\n \"description\": \"Fetches the details of a trip using the given trip UID from the Flixbus API.\",\n \"parameters\": {\n ...	null	xlam	bfcl	17,780	null
[ { "role": "system", "content": "You are a helpful assistant with access to the following functions. Use them if required.\n\n```json\n[\n {\n \"name\": \"route_and_directions\",\n \"description\": \"Generates optimized routes and detailed turn-by-turn directions between specified waypoints using the ...	null	nemotron	bfcl	29,211	null
[ { "role": "user", "content": "I'm researching Litecoin and need the hash of block 789012. Can you get that for me?" }, { "role": "assistant", "content": null, "tool_calls": [ { "id": "call_4977_0", "type": "function", "function": { "name": "get_litecoi...	[ { "type": "function", "function": { "name": "minus_di", "description": "Fetches the Minus Directional Indicator (MINUS_DI) for a given financial instrument.", "parameters": { "type": "object", "properties": { "symbol": { "type": "string", "...	xlam	native	4,977	null
[ { "role": "user", "content": "I'm looking for a torrent of the movie 'The Dark Knight'. Can you also generate a QR code for my website, 'example.com'? Additionally, I need WHOIS information for 'facebook.com' and product details for a product with ID 456." }, { "role": "assistant", "content": nu...	[ { "type": "function", "function": { "name": "get_from_1337_x", "description": "Fetches torrent data for a given search topic from the 1337x website using the RapidAPI service.", "parameters": { "type": "object", "properties": { "searchtopic": { "type":...	xlam	native	43,301	null
[ { "role": "user", "content": "For the decimal numbers -1, 0, and 1, compute the set bit count but ignore the negative input" }, { "role": "assistant", "content": null, "tool_calls": [ { "id": "call_59462_0", "type": "function", "function": { "name": "c...	[ { "type": "function", "function": { "name": "count_bits", "description": "Counts the number of set bits (1's) in the binary representation of a number.", "parameters": { "type": "object", "properties": { "num": { "type": "integer", "descrip...	xlam	native	59,462	null
[ { "role": "system", "content": "You are an expert in composing functions. You are given a question and a set of possible functions. Based on the question, you will need to make one or more function/tool calls to achieve the purpose.\n\nIf you decide to invoke any of the function(s), you MUST put it in the f...	[]	nemotron_bfcl_fmt	bfcl_prompting	13,691	null
[ { "role": "system", "content": "You are a helpful assistant with access to the following functions. Use them if required.\n\n```json\n[\n {\n \"name\": \"get_specific_dessert\",\n \"description\": \"Fetch a specific dessert's details using its ID from the Pizza and Desserts API.\",\n \"parameters\...	null	xlam	bfcl	10,828	null
[ { "role": "user", "content": "Can you find homeless shelters within a 5-mile radius of latitude 34.0522 and longitude -118.2437?" }, { "role": "assistant", "content": null, "tool_calls": [ { "id": "call_18075_0", "type": "function", "function": { "name...	[ { "type": "function", "function": { "name": "by_location", "description": "Searches for homeless shelters based on provided geographical coordinates and radius.", "parameters": { "type": "object", "properties": { "lat": { "type": "string", ...	xlam	native	18,075	null
[ { "role": "system", "content": "You are an expert in composing functions. You are given a question and a set of possible functions. Based on the question, you will need to make one or more function/tool calls to achieve the purpose.\n\nIf you decide to invoke any of the function(s), you MUST put it in the f...	[]	nemotron_bfcl_fmt	bfcl_prompting	43,848	null
[ { "role": "system", "content": "You are an expert in composing functions. You are given a question and a set of possible functions. Based on the question, you will need to make one or more function/tool calls to achieve the purpose.\n\nIf you decide to invoke any of the function(s), you MUST put it in the f...	[]	nemotron_bfcl_fmt	bfcl_prompting	36,515	null

End of preview.

cyan-agentic-v6.1

Agentic SFT mix used to train gravity-sft-v6.1-agentic. Extends trillionlabs/cyan-agentic-v6 by adding BFCL prompting-mode duplicates of the xlam / nemotron rows: the same underlying calls emitted in Python-syntax ([fn(arg=val)]) with a canonical BFCL system prompt, so the model sees each example in two formats (structured tool_calls and BFCL Python-syntax).

Composition

source	rows	format
`xlam`	120,000	tool_calls (role=tool)
`nemotron`	101,480	tool_calls (role=tool)
`xlam_bfcl_fmt`	60,000	BFCL Python-syntax (`[fn(arg=val)]`)
`nemotron_bfcl_fmt`	48,721	BFCL Python-syntax (`[fn(arg=val)]`)
`envgen_v2`	13,891	mixed (terminalbench + officebench), per-turn `mask_loss`
TOTAL	344,092

Structure

Single file: data.chunk.000.jsonl (~2.2 GB). Each row is a JSON object with:

messages — list of {role, content, ...} turns. Role is one of system, user, assistant, tool.
tools — optional list of tool schemas. Some sources embed the tool list into the system-prompt text instead.
source — one of the strings above.
record_idx — stable per-source identifier.

Assistant turns in envgen_v2 carry per-turn quality annotations:

mask_loss (int, 0/1) — 1 means "mask this turn out of the training loss"; 0 means "train on this turn".
quality_category (str) — one of correct_action, looped_stuck, mistake_later_fixed.
reasoning_content (str, optional).

BFCL prompting-mode format

The xlam_bfcl_fmt / nemotron_bfcl_fmt variants use the BFCL prompting-mode convention:

System prompt instructs the model to emit calls as [fn_name1(arg=val, ...), fn_name2(...)] and nothing else.
Observation: tool responses are delivered as a user turn whose content is a stringified list-of-dicts, e.g. [{'role': 'tool', 'name': "fn(arg='val')", 'content': "<output>"}, ...].

Loading

from datasets import load_dataset
ds = load_dataset("trillionlabs/cyan-agentic-v6.1", split="train")

Chat template

Training used the GLM-style template with {% generation %} markers that respect the per-turn mask_loss flag.

Lineage

Base: trillionlabs/cyan-agentic-v6 (xlam / nemotron / envgen_v2).
Superseded by: trillionlabs/cyan-agentic-v7 (adds openagent-v1 loop recovery + endless-terminal SFT).

License

Apache 2.0. Components inherit their upstream licenses.

Downloads last month: 24

Total file size:

2.24 GB