# AssayBench: An Assay-Level Virtual Cell Benchmark for LLMs and Agents

arXiv: https://arxiv.org/abs/2605.10876 · HTML: https://mlsanity.com/paper/2605.10876/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Medical and scientific domain agents, one of 16 directions in LLM agents and tool use.
- Builds on 16 of its 40 references that are in the map.
- First posted on arXiv 11 May 2026.

## Placed in

- LLM agents and tool use / [Medical and scientific domain agents](https://mlsanity.com/agents/medical-and-scientific-domain-agents/)

## Builds on (most cited first, 16 of 16)

- [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://mlsanity.com/paper/2402.03300/) (2024)
- [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](https://mlsanity.com/paper/2311.12022/) (2023)
- [Deep Sets](https://mlsanity.com/paper/1703.06114/) (2017)
- [gpt-oss-120b & gpt-oss-20b Model Card](https://mlsanity.com/paper/2508.10925/) (2025)
- [OpenAI GPT-5 System Card](https://mlsanity.com/paper/2601.03267/) (2025)
- [AlphaEvolve: A coding agent for scientific and algorithmic discovery](https://mlsanity.com/paper/2506.13131/) (2025)
- [GLM-5: from Vibe Coding to Agentic Engineering](https://mlsanity.com/paper/2602.15763/) (2026)
- [GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning](https://mlsanity.com/paper/2507.19457/) (2025)
- [A Theoretical Analysis of NDCG Type Ranking Measures](https://mlsanity.com/paper/1304.6480/) (2013)
- [BixBench: a Comprehensive Benchmark for LLM-based Agents in Computational Biology](https://mlsanity.com/paper/2503.00096/) (2025)
- [How to Build the Virtual Cell with Artificial Intelligence: Priorities and Opportunities](https://mlsanity.com/paper/2409.11654/) (2024)
- [RxRx1: A Dataset for Evaluating Experimental Batch Correction Methods](https://mlsanity.com/paper/2301.05768/) (2023)
- [PerturBench: Benchmarking Machine Learning Models for Cellular Perturbation Analysis](https://mlsanity.com/paper/2408.10609/) (2024)
- [Incompressible Knowledge Probes: Estimating Black-Box LLM Parameter Counts via Factual Capacity](https://mlsanity.com/paper/2604.24827/) (2026)
- [Contextualizing biological perturbation experiments through language](https://mlsanity.com/paper/2502.21290/) (2025)
- [Benchmarking Large Language Models on Multiple Tasks in Bioinformatics NLP with Prompting](https://mlsanity.com/paper/2503.04013/) (2025)

## Built on by (most cited first, 3 of 3)

- [PertMind: Eliciting Emergent Biological Reasoning in LLM via Reinforcement Learning on Cellular Perturbation Data](https://mlsanity.com/paper/2608.16419/) (2026)
- [Multi-LLM Collaborative Alignment via Stackelberg Games](https://mlsanity.com/paper/2609.39076/) (2026)
- [Historical-Feature Residual Adaptation for Sequential Screening](https://mlsanity.com/paper/2610.19315/) (2026)

## Abstract (arXiv)

Recent advances in machine learning and large-scale biological data collections have revived the prospect of building a virtual cell, a computational model of cellular behavior that could accelerate biological discovery. One of the most compelling promises of this vision is the ability to perform in silico phenotypic screens, in which a model predicts the effects of cellular perturbations in unseen biological contexts. This task combines heterogeneous textual inputs with diverse phenotypic outputs, making it particularly well-suited to LLMs and agentic systems. Yet, no standard benchmark currently exists for this task, as existing efforts focus on narrower molecular readouts that are only indirectly aligned with the phenotypic endpoints driving many real-world drug discovery workflows. In this work, we present AssayBench, a benchmark for phenotypic screen prediction, built from 1,920 publicly available CRISPR screens spanning five broad classes of cellular phenotypes. We formulate the screen prediction task as a gene rank prediction for each screen and introduce the adjusted nDCG, a continuous metric for comparing performance across heterogeneous assays. Our extensive evaluation shows that existing methods remain far from empirically estimated performance ceilings and zero-shot generalist LLMs outperform biology-specific LLMs and trainable baselines. Optimization techniques such as fine-tuning, ensembling, and prompt optimization can further improve LLM performance on this task. Overall, AssayBench offers a practical testbed for measuring progress toward in silico phenotypic screening and, more broadly, virtual cell models.