Research

Survey Research Literature with AI-Powered Mind Maps

Use AI mind maps to map research themes, trace citation networks, and identify gaps in the literature. Branch into sub-fields without losing the big picture.

Start your survey See example ↓

Explore the full survey

A complete survey of LLM alignment research, zoomed out to the whole field. Zoom in on any theme to read the synthesis.

Open the full map

How it works

Literature surveys require holding dozens of papers, themes, and connections in your head simultaneously. Linear note-taking tools collapse this complexity into a single thread. Mind maps preserve the natural structure of a research field.

Start with your research question. Type your topic or question. The AI maps the major themes, sub-fields, and methodological approaches.
Branch into sub-fields. Each theme becomes a branch you explore further. “What are the key papers in this area?” “What methods do they use?” “Where do they disagree?” Each question deepens the tree.
Trace connections. When two branches reference the same foundational paper or use the same dataset, the visual layout makes that connection visible. You find the bridges between sub-fields.
Identify gaps. The branches that are thin — few papers, limited methods, unanswered questions — are your research opportunities.

Why branching matters for literature surveys

Research fields aren’t hierarchical — they’re networks. Papers cite each other across sub-fields, methods from one area get applied to another, and foundational assumptions get challenged from unexpected directions.

Mind maps let you represent this structure honestly. Instead of forcing a linear narrative (“First, A was discovered. Then B built on A”), you can map the actual relationships: A and B developed independently, C connected them, and D challenged the assumption underlying both.

Example

You’re surveying the intersection of AI and drug discovery. The root branch splits into “molecular generation,” “property prediction,” and “synthesis planning.” Under molecular generation, you fork into diffusion models, VAEs, and autoregressive approaches. You discover that synthesis planning is underdeveloped relative to generation — a gap worth investigating.

Combine this workflow with hypothesis exploration, research methodology, or grant writing for a complete research planning toolkit.

Now try it yourself

Survey recent papers on large language model alignment

The field clusters into three areas: RLHF and preference learning, constitutional AI and rule-based approaches, and mechanistic interpretability as a path to alignment.

What are the open problems in RLHF?

Reward hacking, where the model exploits the reward signal rather than learning the intended behavior. Also distribution shift — the model encounters inputs at deployment that differ from training.