Reproduce “Contrast Matters: Understanding Robustness of In-Context Fine-Tuning to Target-Context Relatedness” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Deeptanshu Malu, Deevyanshu Malu, Aditya Nemiwal, Sunita Sarawagi
Contrastive-Context trains models with labeled prompt examples spanning different similarity levels, combining random and retrieved examples across batches and mixing similarity within some prompts. Across 32 translation configurations, it is among the strongest methods across similarity levels; Text-to-SQL and semantic-parsing results are also competitive. These findings cover studied tasks and context sizes, not all tasks or settings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.