Source Identification Is Not Fitness Testing: Measuring the Limits of Synthetic-Data Attribution
On the map: Synthetic data and self-improvement for LLMs → Model collapse from synthetic data also fits: LLM synthetic data generation
Abstract (arXiv)
Repeated training on model-generated data can degrade later models. One possible response is to use provenance when deciding which generated examples to reuse. We test both how reliably that provenance can be recovered and whether it helps identify better training data. Using financial-risk text, we first identify the source of generated passages and then repeat the test after rewriting them. Generator attribution is 98.7% accurate on the original passages but falls to 53.1% after paraphrasing and 29.0% after style rewriting. Generated-versus-human detection remains close to perfect against the tested human comparison set. We then compare two ways of selecting generated examples over three rounds of generation and retraining. One uses source information. The other uses a score from a separate reference model. The two rules select different examples, but the planned comparison does not detect a stable difference in the degradation of the resulting models. The results show that identifying where data came from and identifying which data are useful for training are separate problems. The experiment therefore separates source identity, criterion-facing selection, and recursive training outcome: neither the provenance score nor the tested criterion-facing proxy is established as sufficient for future recursive behaviour.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Model collapse from synthetic data, one of 13 directions in Synthetic data and self-improvement for LLMs. Builds on 5 of its 13 references that are in the map. First posted on arXiv 30 Sep 2026.
Builds on
5 of 13 references.
References · 5
- 2023Self-Consuming Generative Models Go MADcited by 250
- 2024RAID: A Shared Benchmark for Robust Evaluation of Machine-Generated Text Detectorscited by 163
- 2024Collapse or Thrive? Perils and Promises of Synthetic Data in a Self-Generating Worldcited by 52
- 2024Self-Correcting Self-Consuming Loops for Generative Model Trainingcited by 37
- 2025OpenTuringBench: An Open-Model-based Benchmark and Framework for Machine-Generated Text Detection and Attributioncited by 9
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026From Behavior to Provenance: Attributing Tabular Foundation Models to Synthetic Pretraining Data
- 2026Provenance-Grounded Gating and Adaptive Recovery in Synthetic Post-Training Data Curation
- 2025Training Data Provenance Verification: Did Your Model Use Synthetic Data from My Generative Model for Training?
- 2026Effective Synthetic Data Curation Requires Group-Level Signals
- 2025Characterizing Model Behavior Under Synthetic Data Training: An Empirical Study Across Scales and Mixing Ratios
- 2026SynthSentry: Detecting Synthetic Data Contamination in Language Model Training Data
- 2026Towards Computational Provenance: Carrying Causal-State Evidence in Generated Text
- 2026How Can We Synthesize High-Quality Pretraining Data? A Systematic Study of Prompt Design, Generator Model, and Source Data