Automatic Pair Construction for Contrastive Post-training

arXiv:2310.02263v2 Announce Type: replace-cross
Abstract: Alignment serves as an important step to steer large language models (LLMs) towards human preferences. In this paper, we propose an automatic way to construct contrastive data for LLM, using preference pairs from multiple models of varying strengths (e.g., InstructGPT, ChatGPT and GPT-4). We compare the contrastive techniques of SLiC and DPO to SFT baselines and find that DPO provides a step-function improvement even after continuing SFT saturates. We also explore a data curriculum learning …

abstract alignment arxiv chatgpt construct construction cs.ai cs.cl cs.lg data gpt gpt-4 human instructgpt language language models large language large language models llm llms multiple paper training type