Summary
CellTrack-Bench is a new benchmark that evaluates how well vision-language models (VLMs) perform cell tracking in microscopy compared to specialized tracking algorithms. Cell tracking demands spatiotemporal awareness of cell dynamics, a task where general-purpose language models may fall short despite strong performance on text-first work. The benchmark is built from 15 standard cell-tracking datasets spanning four imaging modalities, totaling 1,537 problems and 9,041 cell-linking decisions. A hard subset was curated by identifying failure cases shared across different algorithmic trackers. The authors evaluated multiple environment harnesses, benchmarked 9 frontier closed-source models and 13 open-weight models, and provided a detailed analysis of failure modes. The paper (arXiv:2610.31566) is authored by Ilan Theodoro, Jordão Bragantini, and Loïc A. Royer, and was released on October 22, 2026.
Overview
Field: Computer Vision
Authors: Ilan Theodoro, Jordão Bragantini, Loïc A. Royer
Released: 2026-10-22
arXiv: 2610.31566
Summary
Large language models have shown outstanding performance on text-first tasks, but their general-purpose accuracy in specialized scientific domains remains in question. Cell tracking requires spatiotemporal awareness of cell dynamics. The authors created CellTrack-Bench, which assesses VLMs' cell-tracking performance against specialized trackers.
Key Details
- Draws from 15 standard cell-tracking datasets across four imaging modalities
- Comprises 1,537 problems and 9,041 cell-linking decisions
- A hard subset was selected by identifying failures shared across distinct algorithmic trackers
- Benchmarked 9 frontier closed-source and 13 open-weight models across multiple environment harnesses
- Provides detailed failure mode analysis
---
*Automatically collected on 2026-10-26.*
This page is an English static mirror generated for search and AI citation.
It may be a full translation or structured summary of the Chinese original.
Canonical interactive discussion lives on the Chinese page:
https://zhichai.net/topic/178636479