Skip to content

Dataset · DSFFGFG456

Claude Fable 5 agentic coding trajectories

DSFFGFG456/fable-5-coding-and-debugging-traces

The dataset contains instruction, tool use, and agentic coding trajectories from Claude Fable 5.

What's inside

The dataset includes 2,380 trajectories and 12,490 training rows in parquet and JSONL formats. It was generated using the open-source moonshiner tool for distilling verified instructions and agent tracks. The data reflects the real behavior of the anthropic/claude-fable-5 model.

What you can do with it

  • Training and evaluating agentic coding models
  • Distilling tool-use skills for language models
  • Fine-tuning models for instruction following
Tasks
Text generation
Languages
English
Size
10K–100K rows
Downloads this month
902
Likes
8

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("DSFFGFG456/fable-5-coding-and-debugging-traces")