What Large Language Models Know About Plant Molecular Biology
Bibliographic record
Abstract
ABSTRACT Large language models (LLMs) are rapidly permeating scientific research, yet their capabilities in plant molecular biology remain largely uncharacterized. Here, we present MOBIPLANT, the first comprehensive benchmark for evaluating LLMs in this domain, developed by a consortium of 112 plant scientists across 19 countries. MOBIPLANT comprises 565 expert-curated multiple-choice questions and 1,075 synthetically generated questions, spanning core topics from gene regulation to plant-environment interactions. We benchmarked seven leading chat-based LLMs using both automated scoring and human evaluation of open-ended answers. Models performed well on multiple-choice tasks (exceeding 75% accuracy), although most of them exhibited a consistent bias towards option A. In contrast, expert reviews exposed persistent limitations, including factual misalignment, hallucinations, and low self-awareness. Critically, we found that model performance strongly correlated with the citation frequency of source literature, suggesting that LLM knowledge inherits the visibility distribution of the underlying scientific corpus. Consequently, models tend to be more reliable on consolidated topics and less reliable on under-cited or recently emerging ones. We also benchmarked agents equipped with web-search and additional tools in more complex tasks involving DNA sequence analysis. These agents were outperformed by domain specific models in sequence classification and regression tasks, indicating an opportunity for joint agentic systems that combine both the reasoning power of LLMs and the dedicated processing of DNA models. This understanding is key to guiding both the development of next-generation models and the informed use of current tools in the everyday work of plant researchers. MOBIPLANT is publicly available online in this link .
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.000 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.003 | 0.003 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".