Version: 4.5-BCreated: March 2026Purpose: LoRA fine-tuning on OpenClaw-specific agent behaviorTotal Examples: ~11,880Format: Alpaca (instruction, input, output)
This dataset was used for Stage 2 of the Lumivore-1.2B training pipeline. It focuses specifically on OpenClaw agent behavior — tool use, structured reasoning, and agentic task execution patterns.
Stage 2 was a LoRA fine-tuning (rank=64) on top of the Stage 1 full… See the full description on the dataset page:
https://huggingface.co/datasets/LumiVore/lumivore-stage2-training-data.
