“https://arxiv.org/html/2510.25941v1 You can get it to reproduce content but it’s a game of cat and mouse. Were it not for the alignment to avoid direct reproduction it would taken far more often. > RECAP consistently outperforms all other methods; as an illustration, it extracted ≈3,000 passages from the first "Harry Potter" book with Claude-3.7, compar…”
RECAP: Reproducing Copyrighted Data from LLMs Training with an Agentic Pipeline
If we cannot inspect the training data of a large language model (LLM), how can we ever know what it has seen? We believe the most compelling evidence arises when the model itself freely reproduces the target content. As such, we propose RECAP, an agentic pipeline designed to elicit and verify memorized training data from LLM outputs. At the heart of RECAP is a feedback-driven loop, where an initial extraction attempt is evaluated by a secondary language model, which compares the output against a reference passage and identifies discrepancies. These are then translated into minimal correction hints, which are fed back into the target model to guide subsequent generations. In addition, to address alignment-induced refusals, RECAP includes a jailbreaking module that detects and overcomes such barriers. We evaluate RECAP on EchoTrace, a new benchmark spanning over 30 full books, and the results show that RECAP leads to substantial gains over single-iteration approaches. For instance, with GPT-4.1, the average ROUGE-L score for the copyrighted text extraction improved from 0.38 to 0.47 - a nearly 24% increase.
候选层:至少 2 个不同的人、在 2 个不同月份提到过,还没到主书架的 3 人 3 个月。没有做立场标注,所以只写“被提及”。
HN 书评
推荐在前,同类按时间 · 每篇最多保留 20 条,统计用全部提及- Recap: Reproducing Copyrighted Data from LLMs Training with an Agentic Pipeline ↗PaulHoule · 2025-11-21 · 2 分