textnvidia/Nemotron-Math-v2mathematicsreasoningllm-trainingdistillationlong-contexttool-usenvidiafine-tuningsynthetic-datacot

Nemotron-Math-v2: Mathematical Reasoning Trajectories (NVIDIA)

Name: Nemotron-Math-v2: Mathematical Reasoning Trajectories (NVIDIA)
Creator: DataBazaar
Keywords: nvidia/Nemotron-Math-v2, mathematics, reasoning, llm-training, distillation, long-context, tool-use, nvidia, fine-tuning, synthetic-data, cot

About this data

NVIDIA's 347K math problems with 7M model-generated reasoning trajectories for distilling mathematical reasoning. Long-context, tool-use, multi-mode supervision. CC-BY-4.0/CC-BY-SA-4.0.

Schema

Name	Type	Description
uuid	VARCHAR	Unique identifier (UUID v4 format) for the record.
expected_answer	VARCHAR	Final numerical or symbolic answer to the mathematical problem.
problem	VARCHAR	Mathematical problem statement in LaTeX or plain text format.
original_expected_answer	VARCHAR	Initial expected answer before any corrections or majority voting.
changed_answer_to_majority	BOOLEAN	Boolean flag indicating if answer was updated to match majority model consensus.
data_source	VARCHAR	Origin dataset or problem collection (e.g., aops, competition, textbook).
messages	STRUCT("role" VARCHAR, "content" VARCHAR, reasoning_content VARCHAR, tool_calls STRUCT(id VARCHAR, "type" VARCHAR, "function" STRUCT("name" VARCHAR, arguments VARCHAR))[], tool_call_id VARCHAR, "name" VARCHAR)[]	Array of conversational turns with role, content, reasoning traces, and tool invocations.
used_in	VARCHAR[]	Array of dataset splits or benchmarks this record appears in.
metadata	STRUCT(reason_low_with_tool STRUCT(count BIGINT, pass BIGINT, accuracy DOUBLE), reason_low_no_tool STRUCT(count BIGINT, pass BIGINT, accuracy DOUBLE), reason_medium_with_tool STRUCT(count BIGINT, pass BIGINT, accuracy DOUBLE), reason_medium_no_tool STRUCT(count BIGINT, pass BIGINT, accuracy DOUBLE), reason_high_with_tool STRUCT(count BIGINT, pass BIGINT, accuracy DOUBLE), reason_high_no_tool STRUCT(count BIGINT, pass BIGINT, accuracy DOUBLE))	Nested counts and accuracy metrics by reasoning difficulty level and tool availability.
license	VARCHAR	License identifier governing dataset usage (e.g., CC-BY-4.0, CC-BY-SA-4.0).
tools	STRUCT("type" VARCHAR, "function" STRUCT("name" VARCHAR, description VARCHAR, parameters STRUCT("type" VARCHAR, properties STRUCT(code STRUCT("type" VARCHAR, description VARCHAR)), required VARCHAR[])))[]	Array of available function definitions with parameters for tool-use trajectories.
url	VARCHAR	Source URL or reference link for the problem.
user_name	VARCHAR	Username or author identifier of problem contributor or source.
user_url	VARCHAR	Profile or homepage URL of the user who contributed the problem.

Sample Data

Preview a sample of the data before downloading.

Free

Open dataset

Quality: No ratings

0 downloads

Seller: DataBazaar

Agent? No sign-up needed →

For AI Agents

Via MCP Server

# 1. Add to your agent's MCP config (claude_desktop_config.json or similar):
{
  "mcpServers": {
    "databazaar": { "command": "npx", "args": ["databazaar-mcp"] }
  }
}

# 2. Your agent can then call:
search_datasets({ query: "Nemotron-Math-v2: Mathematical" })
// Found: 17600013-74d6-45d5-b618-c2f8b0afe9c8
get_download_url({ dataset_id: "17600013-74d6-45d5-b618-c2f8b0afe9c8" })  // free — no API key needed

Via REST API

# Free dataset — no API key required:
curl https://api.databazaar.io/datasets/17600013-74d6-45d5-b618-c2f8b0afe9c8/download-url