AI AgriBench logoAI AgriBench
Stay in the loop!

Smallholders Leaderboard

Click here to Join the Leaderboard

At a glance

Smallholder-focused benchmark

Use the interactive table and bar charts below to compare model performance across core evaluation metrics.

Example question

"My maize leaves have yellow streaks after recent rain. Is this disease or nutrient stress, and what should I do next?"

What is scored

AccuracyRelevanceCompletenessConciseness
Loading benchmark data… Fetching scores from the database.

This page shows a preliminary Leaderboard evaluating public LLM-based chatbots on farmer questions related to agronomic topics. Agriculture-specific advisory services will be added to this leaderboard soon.

The benchmark data set for this leaderboard uses questions from farmers received by a variety of farmer-facing advisory services in parts of India, several countries in Africa, and other parts of the world dominated by small farms. The answers come from advisors working for the organizations that fielded the questions and shared the data with AI AgriBench, listed below. The data set is preliminary, with only minimal quality filtering, and is intended to illustrate how a smallholder benchmark effort could be used to evaluate and compare Ag Advisory services intended for smallholder farmers.

The data set includes a total of about 3100 question-answer pairs from 3 sources:

  • IFPRI - 2000 QA pairs
  • Digital Green - 1000 QA pairs
  • Precision Development (PxD) - 100 sample QA pairs.