idea-discovery

To find and validate research ideas, you can run a multi-phase pipeline that surveys literature, generates candidate ideas, performs novelty verification, and refines the best proposal into an experiment plan.

The best way to check if a research idea is novel is to perform deep novelty verification against recent literature, which eliminates ideas that are already published and filters for true originality.

To create an experiment roadmap for a new research hypothesis, you refine the top proposal into a detailed experiment plan that includes pilot budgets, run order, and execution checkpoints.

Yes, you can automate literature review and research planning simultaneously by executing an end-to-end idea discovery workflow that orchestrates literature surveys, idea generation, and refinement in one pipeline.

No, you do not need a reference paper to run research ideation; the pipeline optionally accepts a reference paper and a research brief to ground the literature survey and narrow broad scope.

Limitations of automating research idea discovery include enforced guardrails for GPU pilot budgets and timeouts, requiring checkpointed user-facing progress to manage resource constraints during the multi-phase pipeline.