Dataset Viewer
Auto-converted to Parquet Duplicate
messages
listlengths
28
166
[ { "role": "user", "content": "# Wave-4 authoring-agent preamble (read this first, follow exactly)\n\nYou are authoring seeds for the fable-code training-data pipeline. Repo root:\nthis directory. Everything below is mandatory.\n\n## Read before writing anything\n\n1. `tasks/SEED_SPEC.md` — the seed spec, qu...
[{"role":"user","content":"# Wave-8 enterprise-integration seed authoring preamble\n\nYou are the so(...TRUNCATED)
[{"role":"user","content":"# Wave-4 authoring-agent preamble (read this first, follow exactly)\n\nYo(...TRUNCATED)
[{"role":"user","content":"# Wave-8 enterprise-integration seed authoring preamble\n\nYou are the so(...TRUNCATED)
[{"role":"user","content":"# Wave-4 authoring-agent preamble (read this first, follow exactly)\n\nYo(...TRUNCATED)
[{"role":"user","content":"# Wave-4 authoring-agent preamble (read this first, follow exactly)\n\nYo(...TRUNCATED)
[{"role":"user","content":"Data-platform ticket. Our telemetry archiver pushes nightly rollup blobs (...TRUNCATED)
[{"role":"user","content":"# Wave-8 enterprise-integration seed authoring preamble\n\nYou are the so(...TRUNCATED)
[{"role":"user","content":"Feature request for our Jira retention tooling. The existing ProjectAdmin(...TRUNCATED)
[{"role":"user","content":"Ingest-platform ticket. Our camera-frame pipeline pulls from Pub/Sub with(...TRUNCATED)
End of preview. Expand in Data Studio

source: MoreThought/Fable-5.1-Max-Reasoning-Filtered-10000x

Dataset cleaning steps:

  • 1: Remove defective paraphrases (1,000 remaining)
  • 2: Remove incomplete traces (257 remaining)
  • 3: Remove structural junk (168 remaining)
  • 4: Remove entries lacking reasoning (167 remaining)
  • 5: Remove duplicates (41 remaining)

ty

Downloads last month
126

Models trained or fine-tuned on mondk/Fable-5.1-Max-CLEANED