New tool converts agent failures into fine-tuning data

A new open-source tool called trace2train has been released to convert failed agent traces into supervised fine-tuning (SFT) or Direct Preference Optimization (DPO) training data. Developed as a local CLI tool, it aims to leverage agent errors, such as incorrect tool selection or bad arguments, which are often lost in logs, for model improvement. The tool supports various trace sources like LangSmith and Langfuse, and can utilize models like DeepSeek or OpenAI for processing, while prioritizing data integrity by only generating corrections when derivable from the trace. AI

IMPACT Enables more efficient fine-tuning of AI agents by leveraging previously discarded error data.

RANK_REASON New open-source tool release for AI development.

Read on dev.to — LLM tag →

• DeepSeek
• Direct Preference Optimization
• Langfuse
• LangSmith
• LLaMA-Factory
• OpenAI
• supervised fine-tuning
• trace2train

• product
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →