Kazakh Multi-Tool Agentic AI and Locale Fidelity Dataset is a Kazakh-language dataset designed for training and evaluating Large Language Models (LLMs) in agentic AI scenarios that require multi-step tool use, function calling, and locale-sensitive reasoning.

The dataset focuses on practical assistant workflows where a model must understand a user request in Kazakh, select the correct tools, generate structured tool-call arguments, process mock tool responses, and produce a final answer in natural Kazakh. Many samples involve sequential tool use, such as first searching for information and then using another tool to enrich the result.

📊 Dataset Statistics

General Metrics

Metric	Count
Total Samples	2,029
Total Words (approx.)	635,995
Avg. Words per Sample	313

Word Count Distribution Per Field

The following table details the distribution of word counts across different fields in the dataset.

Field	Mean	Median	Min	Max	Total Words
answers	31.0	27.0	10	163	62,929
category	1.0	1.0	1	1	2,029
difficulty	1.0	1.0	1	1	2,029
id	1.0	1.0	1	1	2,029
query	16.7	16.0	3	42	33,882
tools	88.1	84.0	33	159	178,692
turns	174.7	167.0	97	382	354,405

Dataset Structure

Each dataset instance represents a complete tool-augmented interaction scenario. A sample typically includes a Kazakh user query, the available tool schemas, the expected tool calls, and the full conversation trace from user request to final answer.

Data Fields

id: A unique identifier for the sample.
query: The original user request, usually written in Kazakh. The query may include dates, cities, local entities, map-related requests, event-related requests, or other context-sensitive details.
category: The task category. For example, 10_multilingual_locale_fidelity indicates a task involving multilingual or locale-sensitive behavior, such as preserving Kazakh language usage, handling local place names, and using correct date or location formats.
tools: A list of available tools for the assistant. Each tool includes:
- name: the tool name;
- description: the tool’s function;
- parameters: the required or optional input schema;
- mock_response: the expected response structure.
difficulty: The annotated difficulty level of the sample.
answers: The target tool calls expected from the assistant. This field usually contains one or more tool calls with serialized JSON arguments.
turns: The full multi-turn trajectory, including:
- the user request;
- the assistant’s planning or reasoning step;
- one or more assistant tool calls;
- mock tool responses;
- the final assistant answer.

Data Instance

Below is one representative example from the dataset.

{
        "id": "dastan_10_1",
        "query": "Алматыда 2026-04-15 күні болатын концерттерді тауып, өтетін орнын картадан координатымен бер.",
        "category": "10_multilingual_locale_fidelity",
        "tools": [
            {
                "name": "events.search",
                "description": "Search for events in a city",
                "parameters": {
                    "city": {
                        "type": "string",
                        "description": "City name",
                        "required": true
                    },
                    "type": {
                        "type": "string",
                        "description": "concert, sports, theater, etc",
                        "required": false
                    },
                    "date": {
                        "type": "string",
                        "description": "Event date YYYY-MM-DD",
                        "required": false
                    }
                },
                "mock_response": {
                    "events": [
                        {
                            "id": "",
                            "name": "",
                            "venue": "",
                            "date": "",
                            "tickets_available": null
                        }
                    ]
                }
            },
            {
                "name": "maps.geocode",
                "description": "Convert address to latitude/longitude coordinates",
                "parameters": {
                    "address": {
                        "type": "string",
                        "description": "Full address or location name",
                        "required": true
                    }
                },
                "mock_response": {
                    "lat": null,
                    "lon": null,
                    "formatted_address": ""
                }
            }
        ],
        "difficulty": "easy",
        "answers": [
            {
                "name": "events.search",
                "arguments": "{\"city\": \"Алматы\", \"type\": \"concert\", \"date\": \"2026-04-15\"}"
            },
            {
                "name": "maps.geocode",
                "arguments": "{\"address\": \"Almaty Arena, Алматы\"}"
            }
        ],
        "turns": [
            {
                "role": "user",
                "content": "Алматыда 2026-04-15 күні болатын концерттерді тауып, өтетін орнын картадан координатымен бер."
            },
            {
                "role": "assistant",
                "content": "Алдымен іс-шараны тауып, кейін өтетін орнын геокодтап, нәтижені қысқаша қорытындылаймын.",
                "meta": {
                    "plan": "events.search + maps.geocode with proper locale handling",
                    "tags": [],
                    "difficulty": "hard"
                }
            },
            {
                "role": "assistant",
                "content": "",
                "tool_call": {
                    "name": "events.search",
                    "arguments": "{\"city\": \"Алматы\", \"type\": \"concert\", \"date\": \"2026-04-15\"}"
                }
            },
            {
                "role": "tool",
                "content": "{\"events\": [{\"id\": \"evt_98231\", \"name\": \"Spring Beats Live\", \"venue\": \"Almaty Arena\", \"date\": \"2026-04-15\", \"tickets_available\": 247}]}"
            },
            {
                "role": "assistant",
                "content": "",
                "tool_call": {
                    "name": "maps.geocode",
                    "arguments": "{\"address\": \"Almaty Arena, Алматы\"}"
                }
            },
            {
                "role": "tool",
                "content": "{\"lat\": 43.3472, \"lon\": 76.8851, \"formatted_address\": \"Almaty Arena, микрорайон Нуркент, Алматы, Қазақстан\"}"
            },
            {
                "role": "assistant",
                "content": "Табылған концерт: Spring Beats Live (2026-04-15), орны: Almaty Arena, билеттер: 247. Координаттары: 43.3472, 76.8851 (Almaty Arena, микрорайон Нуркент, Алматы, Қазақстан). (Дерек көзі: events.search + maps.geocode)"
            }
        ]
    }

Funding

This dataset was developed as part of the project funded by the Ministry of Science and Higher Education of the Republic of Kazakhstan under Grant No. BR24993001, “Creation of a Large Language Model (LLM) to Support the Kazakh Language and Advance Technological Development.”

Citation

If you use this dataset in your research, please cite the following article:

APA

Kadyrbek, N., Tuimebayev, Z., Mansurova, M., & Viegas, V. (2025). The development of small-scale language models for low-resource languages, with a focus on Kazakh and direct preference optimization. Big Data and Cognitive Computing, 9(5), 137. https://doi.org/10.3390/bdcc9050137

BibTeX

@article{kadyrbek2025development,
  title     = {The Development of Small-Scale Language Models for Low-Resource Languages, with a Focus on Kazakh and Direct Preference Optimization},
  author    = {Kadyrbek, Nurgali and Tuimebayev, Zhanseit and Mansurova, Madina and Viegas, Vitor},
  journal   = {Big Data and Cognitive Computing},
  volume    = {9},
  number    = {5},
  pages     = {137},
  year      = {2025},
  publisher = {MDPI},
  doi       = {10.3390/bdcc9050137},
  url       = {https://www.mdpi.com/2504-2289/9/5/137}
}

Downloads last month: 14

Total file size:

2.02 MB

Collection including farabi-lab/Multilingual_Locale_Fidelity

Expert Curated Dataset 2026

Collection

The dataset is used for fine-tuning and aligning LLMs with Kazakh culture. • 13 items • Updated 3 days ago