paper

Provenance-Grounded Gating and Adaptive Recovery in Synthetic Post-Training Data Curation

paperactiveprovisionalprovenance-grounded-gating-and-adaptive-recovery-in-synthetic-post-training-data-curation-e8ce849f·1 events·first seen 7d ago

Aliases: Provenance-Grounded Gating and Adaptive Recovery in Synthetic Post-Training Data Curation

More like this (12)

Anatomy of Post-Training: Using Interpretability to Characterize Data and Shape the Learning Signal Learning to Reason by Analogy via Retrieval-Augmented Reinforcement Fine-Tuning The Stable Recovery Manifold: Geometric Principles Governing Recoverability in Continual Learning Reference-Augmented Training generative post-training Post-Retrieval Cascade chain-of-thought training data generation Attention Amnesia in Hybrid LLMs: When CoT Fine-Tuning Breaks Long-Range Recall, and How to Fix It temporally ordered pre-training Self-Supervised Pretraining Supervised Memory Training FlashbackCL: Mitigating Temporal Forgetting in Federated Learning

Recent events (1)

5arXiv · cs.CL·7d ago·source ↗

Provenance-grounded gating and adaptive recovery improve synthetic post-training data curation

A controlled study examines two underexplored practices in synthetic post-training data pipelines: grounding filtering signals in source provenance and systematically recovering rejected samples rather than discarding them. Using adversarially injected corpora as ground-truth failure labels, the authors find that exact source provenance improves faithfulness gating for stronger judges, that hallucination and reward gates reject largely disjoint populations (making both necessary), and that adaptive recovery via failure diagnosis and targeted regeneration outperforms naive resampling. Generator scale is the primary driver of downstream fine-tuning quality, with filtration and recovery contributing meaningfully but secondarily.

Evaluation and Benchmarking Alignment and RLHF Provenance-Grounded Gating and Adaptive Recovery in Synthetic Post-Training Data Curation