Skip to content
AI-generated research idea🟠Medium🔬Significant ResearchClosely matching work found

Constraint-Grounded Feasibility Verifiers for Idea Generation

Implement a verifiable feasibility critic trained on technical protocols to parameterize generated research ideas and reject plans that violate resource constraints.

Time

a couple of months

Compute level

Single GPU or less

Resources

Dataset of lab equipment/protocols, Standard GPU

Prerequisites

Prompt engineering, Verification loops, Domain-specific knowledge

Scope

modify existing code

The Gap

The evaluation of Idea Generation (Table 6) shows a stark contrast between high Novelty scores and consistently low Feasibility scores (ranging 14-22%). The paper notes that generated ideas are 'fluent but underspecified,' often missing critical parameters, resource assumptions, or interface definitions.

Limitation

LLMs hallucinate scientifically plausible but practically executable research plans because they optimize for linguistic coherence rather than physical or logistical constraints (e.g., reagent availability, equipment compatibility).

Proposed Solution

Implement a 'Feasibility Critic' module trained on equipment manuals and experimental protocols. This critic would act as a verifier during the generation phase, forcing the model to parameterize its ideas (e.g., specifying exact concentrations, equipment models, and compute budgets) and rejecting plans that violate defined resource constraints.

Why This Matters

High novelty is useless without feasibility; this improvement converts LLMs from 'sci-fi generators' into practical research assistants capable of proposing executable experiments.

How to Get Started

  1. Compile a corpus of lab equipment manuals, reagent datasheets, and experimental compute budgets.
  2. Train a verifier model to parse research proposals and extract explicit parameters like concentrations and budgets.
  3. Integrate the feasibility critic into the idea generation loop to score and filter incomplete proposals.
  4. Measure improvements in downstream feasibility scores and constraint compliance on idea generation benchmarks.

Code linked to this paper on Hugging Face: github.com/InternScience/SGI-Bench

Suggested by AI. Adapt it to your setup.

Has This Been Done?

Closely matching work found

AlphaIdeaBench: Benchmarking Automated Research Idea Generation under Feasibility Constraints with Peer Review Alignment explicitly focuses on automated scientific idea generation under explicit resource and feasibility constraints.

We searched OpenAlex for “Implement a verifiable feasibility critic trained on technical protocols to parameterize generated research ideas and reject plans that violate resource constraints.” and “Constraint-Grounded Feasibility Verifiers for Idea Generation” and an AI compared the idea with the top 10 results. A small search can miss relevant work.

Source Paper

Probing Scientific General Intelligence of LLMs with Scientist-Aligned Workflows

Wanghan Xu, Yuhao Zhou, Yifan Zhou +104 more · 2025

This paper introduces Scientific General Intelligence (SGI), a framework grounded in the Practical Inquiry Model (PIM) that evaluates AI on the complete cycle of scientific discovery: Deliberation, Conception, Action, and Perception. To measure this, the authors introduce SGI-Bench, a scientist-aligned benchmark comprising over 1,000 expert-curated samples across 10 disciplines, revealing that while current LLMs are linguistically fluent, they lack the integrated reasoning capabilities required for genuine scientific discovery.

Related Ideas

Is this a good idea?

Share onXLinkedIn