Claude Fable 5: Agent Traces for Supervised Fine-Tuning

Generated by TeichAI, this dataset contains raw agent trace files for supervised fine-tuning. All assistant responses were produced by the claude-fable-5 model. The dataset includes 43 JSONL files and was last updated on June 14, 2026.

Use Cases

• Supervised fine-tuning of language models based on agent interaction traces.
• Training AI agents to use tools based on configured or recovered tool schemas present in the traces.
• Analyzing the behavior and response patterns of the claude-fable-5 model in agent-based scenarios.

Strengths

• Contains 43 JSONL files of agent traces.
• All assistant responses were generated by a specific model (claude-fable-5), ensuring consistency.
• Includes configured or recovered tool schemas to keep tools available for training.

Limitations

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

Provenance

Source

TeichAI via Hugging Face user armand0e.

Collection Method

Generated using the teich tool.

Freshness

Last updated 2026-06-14 04:33:35.