shape-of-thought
Demonstrate that synthetic CoT traces with incorrect final answers outperform human-written correct solutions for supervised fine-tuning. Distribution proximity between training data and student model
SKILL.mdskills/skillxiv-v0.0.2-claude-opus-4.6/shape-of-thought/SKILL.md
GitHubOverview
Challenges conventional wisdom that training data quality depends primarily on correctness.
Core Technique
Distribution Proximity Hypothesis:
# Human traces (H): correct but distribution-mismatched
# Model traces correct (G): correct and distribution-matched
# Model traces incorrect (W): incorrect but distribution-matched
# W outperforms H despite incorrectness
# because distribution proximity enables faster learning
When to Use
Use when: Curating reasoning datasets, SFT training, synthetic data selection.
References
- Distribution alignment vs correctness
- Partial correctness in synthetic data
- Dataset curation guidance