โ˜… UC Berkeley term project

Stop drowning in papers. Read the ones that matter.

Literature Assistant fetches the newest research for your topic, has an LLM summarise and score every paper for relevance, and learns from your feedback to surface what you should read next.

Invite-only ยท Accounts are gated by a registration key to protect the API quota.

Today's brief 12 new
Deep RL for motion planning5/5
Graph neural nets survey3/5
Off-topic preprint1/5

Two ways to work through the literature

One tool, two modes - whether you want a daily pulse on a field or to systematically mine it.

โ˜€

Daily Brief

Pulls the most relevant new papers from Semantic Scholar (journals, IEEE, ACM, arXiv), has an LLM summarise and rate each 1โ€“5, and shows a ranked list so you only read the top ones.

๐Ÿ”

Research Assistant

Takes the papers you've liked and suggests the closest new matches across the whole literature, ranked by your own learned taste โ€” one round at a time.

๐Ÿ“Š

Built-in evaluation

Is the LLM actually worth it? The tool compares LLM scores against TF-IDF and embeddings on the papers you label yourself โ€” with precision, recall and ranking metrics.

How it works

From a research question to a ranked reading list in minutes.

Describe your topic. Write a precise research question. The LLM turns it into an academic search query.
Fetch & rate. The tool pulls a broad pool of candidate papers and scores each one for relevance with reasoning.
Label what's relevant. Mark papers relevant / not relevant. Your labels are the answer key.
It learns & improves. The Research Assistant refines its search from your taste; the evaluation shows whether it's working.

Built to answer a real question

Is an LLM genuinely better at triaging research than simple keyword or embedding methods? This tool measures it โ€” and helps you keep up with your field along the way.

Create an account