OpenSeeker-v1-Data: Training Examples for Search Agent Fine-Tuning

OpenSeeker-v1-Data is a collection of 11.7K training examples used to fine-tune the Qwen3-30B-A3B-Thinking-2507 model for search agent tasks. The dataset was created by PolarSeeker and released on Hugging Face to democratize access to frontier search capabilities. It was last updated on March 17, 2026.

Use Cases

• Fine-tuning language models for search agent tasks based on the described 11.7K training examples.
• Benchmarking search agent performance using the BrowseComp-ZH and other mentioned frontier search benchmarks.
• Researching methods for democratizing frontier AI capabilities through open-source training data.

Strengths

• Contains 11,700 training examples specifically for search agent development.
• Associated model achieved state-of-the-art performance on frontier search benchmarks, including a score of 48.4 on BrowseComp-ZH.
• Fully open-sourced training data aligns with the goal of democratizing access.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

PolarSeeker

Collection Method

Likely created for fine-tuning the Qwen3-30B-A3B-Thinking-2507 model.

Freshness

Last updated 2026-03-17 20:45:50; freshness should be verified.