A 7B data synthesis model trained on 703k DeepSeek-V3 distilled examples can produce domain-specific instruction data competitive with using DeepSeek-V3 directly, at roughly one fifth of the production cost.
- The response contains noticeable grammatical errors, spelling mistakes, or awkward phrasing
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
AQuilt: Weaving Logic and Self-Inspection into Low-Cost, High-Relevance Data Synthesis for Specialist LLMs
A 7B data synthesis model trained on 703k DeepSeek-V3 distilled examples can produce domain-specific instruction data competitive with using DeepSeek-V3 directly, at roughly one fifth of the production cost.