Reproduce “It's About Time: Incorporating Temporality in Retrieval Augmented Language Models” 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
Anoushka Gade, Jorjeta Jetcheva
San Jose State University researchers developed TempRALM, an extension to Retrieval Augmented Language Models that incorporates explicit temporal awareness into document retrieval. The model improved exact match accuracy by up to 74% for time-sensitive questions where the query year did not directly match the event year, offering a computationally efficient way to handle evolving information.
Key insights
- Explicitly incorporating temporal proximity into the retrieval score significantly enhances RALMs' ability to answer time-sensitive queries, especially when direct year matching is insufficient.
- Masking future information during the retrieval process is crucial for maintaining historical accuracy and preventing information leakage into past queries.
- Temporal augmentation offers substantial performance gains in time-sensitive question answering without the need for computationally expensive model re-training or full document index updates.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.