SearchSwarm-SFT: Agentic LLM Delegation Trajectories

SearchSwarm-SFT is a supervised fine-tuning dataset designed to instill delegation intelligence into agentic Large Language Models for long-horizon deep research. It contains high-quality, harness-guided trajectories where a main agent learns to decompose tasks, delegate subtasks, and integrate citation-grounded reports. The dataset was created by SearchSwarm and last updated on Hugging Face in June 2026.

Use Cases

• Fine-tuning a main agent for task decomposition based on the described harness-guided trajectories.
• Training models to manage finite context windows via strategic delegation to subagents.
• Developing agents capable of integrating citation-grounded reports into a final output.
• Benchmarking agentic LLM performance on long-horizon deep research problems.

Strengths

• Designed specifically for instilling delegation intelligence in agentic LLMs.
• Contains high-quality, harness-guided trajectories as described.
• Last updated on Hugging Face on 2026-06-11.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and sample data are unknown, limiting suitability assessment.
• License information is unavailable, which may restrict usage.

Provenance

Source

SearchSwarm via Hugging Face.

Collection Method

Likely contains trajectories generated or curated for supervised fine-tuning.

Freshness

Last updated 2026-06-11 07:32:11.