Provenance-Grounded Gating and Adaptive Recovery in Synthetic Post-Training Data Curation
This paper investigates the effectiveness of filtering and recovery methods in synthetic post-training pipelines.
This paper presents a controlled study of filtering and recovery methods in synthetic post-training pipelines, which is a novel approach compared to prior work.
Keywords
Before reading this…
Applications
- →synthetic post-training pipelines
- →natural language processing
To understand this paper, make sure you know these concepts first:
- synthetic post-training pipelinesfind papers →
- natural language processingfind papers →
Abstract
More Like ThisSynthetic post-training pipelines commonly filter generated samples with reward models or holistic LLM judges, yet two practices remain rarely examined together: whether the filtering signal is grounded in the source evidence that induced each generation, and whether rejected samples can be systematically recovered rather than permanently discarded. We present a controlled study of both questions across gate configurations, recovery strategies, and generator scales, using adversarially injected corpora to provide ground-truth failure labels. We find that exact source provenance improves faithfulness gating for stronger judges, that hallucination and reward gates reject largely disjoint sample populations making both necessary, and that an adaptive recovery pipeline combining failure diagnosis with targeted regeneration achieves higher yield, recovery rate, and injection recall than naive resampling. Downstream fine-tuning quality is driven primarily by generator scale, with filtration and recovery conditions contributing meaningfully but secondarily.