Sign in to view source links and access this dataset
Description
Nexlab provides approximately 159,972 supervised fine-tuning examples for agentic coding, derived from the Complete-FABLE.5-traces-2M trace set. The dataset consists of multi-turn conversations where an assistant drives a tool-call loop involving shell commands, file edits, and tests to complete solutions. It was last updated on June 24, 2026.
Use Cases
Training language models for autonomous code generation based on multi-turn conversation examples.
Fine-tuning models for tool-call orchestration based on described shell, file edit, and test interactions.
Researching agentic coding behaviors based on curated traces of assistant-driven solution completion.
Strengths
Contains approximately 159,972 supervised fine-tuning examples.
Examples are curated for agentic coding with multi-turn conversations and tool-call loops.
Dataset provenance is traced to the Complete-FABLE.5-traces-2M source.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is unknown, which may limit suitability assessment.
Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
Source
Curated/distilled from Glint-Research/Complete-FABLE.5-traces-2M.
Collection Method
Curated/distilled from a larger trace set.
Freshness
Last updated 2026-06-24 07:32:14; freshness should be verified.
License is unknown; terms of use must be verified from the source page.