# EmbGen: Teaching with Reassembled Corpora

arXiv: https://arxiv.org/abs/2605.19394 · HTML: https://mlsanity.com/paper/2605.19394/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 17 of its 37 references that are in the map.
- First posted on arXiv 19 May 2026.

## Builds on (most cited first, 17 of 17)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Judging LLM-as-a-Judge with MT-Bench and Chatbot Arena](https://mlsanity.com/paper/2306.05685/) (2023)
- [BERTScore: Evaluating Text Generation with BERT](https://mlsanity.com/paper/1904.09675/) (2019)
- [SQuAD: 100,000+ Questions for Machine Comprehension of Text](https://mlsanity.com/paper/1606.05250/) (2016)
- [G-Eval: NLG Evaluation using GPT-4 with Better Human Alignment](https://mlsanity.com/paper/2303.16634/) (2023)
- [BLEURT: Learning Robust Metrics for Text Generation](https://mlsanity.com/paper/2004.04696/) (2020)
- [MPNet: Masked and Permuted Pre-training for Language Understanding](https://mlsanity.com/paper/2004.09297/) (2020)
- [Distilling Step-by-Step! Outperforming Larger Language Models with Less Training Data and Smaller Model Sizes](https://mlsanity.com/paper/2305.02301/) (2023)
- [Synthetic Data Augmentation for Zero-Shot Cross-Lingual Question Answering](https://mlsanity.com/paper/2010.12643/) (2020)
- [LAB: Large-Scale Alignment for ChatBots](https://mlsanity.com/paper/2403.01081/) (2024)
- [Leveraging QA Datasets to Improve Generative Data Augmentation](https://mlsanity.com/paper/2205.12604/) (2022)
- [Knowledge-Instruct: Effective Continual Pre-training from Limited Data using Instructions](https://mlsanity.com/paper/2504.05571/) (2025)
- [StrucSum: Graph-Structured Reasoning for Long Document Extractive Summarization with LLMs](https://mlsanity.com/paper/2505.22950/) (2025)
- [Efficient End-to-end Language Model Fine-tuning on Graphs](https://mlsanity.com/paper/2312.04737/) (2023)
- [Calibrating Pre-trained Language Classifiers on LLM-generated Noisy Labels via Iterative Refinement](https://mlsanity.com/paper/2505.19675/) (2025)
- [KCS: Diversify Multi-hop Question Generation with Knowledge Composition Sampling](https://mlsanity.com/paper/2508.20567/) (2025)

## Abstract (arXiv)

Adapting small instruction-tuned models to specialized domains often relies on supervised fine-tuning (SFT) on curated instruction-response examples, which is expensive to collect at scale. Synthetic training examples generated by a teacher LLM from a domain corpus can reduce this cost, but existing pipelines can produce homogenized outputs and do not consistently capture cross-passage or cross-document dependencies. We introduce EmbGen, a synthetic data generation pipeline that decomposes a corpus into entity-description pairs, reassembles them using semantic structure inferred from embedding similarity, and then generates question-answer (QA) pairs via proximity, intra-cluster, and inter-cluster sampling with cluster-specialized system prompts. We evaluate EmbGen against EntiGraph, InstructLab and Knowledge-Instruct on three datasets of varied semantic heterogeneity, under fixed token budgets (5 and 20 million tokens). We use lexical overlap metrics, an LLM-as-a-judge rubric, and Binary Accuracy, a composed metric combining Factual Accuracy and Completeness for evaluation. EmbGen improves Binary Accuracy on the most heterogeneous dataset by 12.5% at 5M and 88.9% at 20M tokens budget, relative to the strongest baseline, while remaining competitive across other datasets with lower heterogeneity.