CellTrack-Bench compares vision-language models with specialized cell trackers
Original title: CellTrack-Bench: Microscopic Cell Tracking Benchmark for Vision-Language Models
The benchmark uses 15 datasets, four imaging modalities, and 1,537 problems covering 9,041 cell-linking decisions.
Researchers release CellTrack-Bench to test whether vision-language models can handle microscopic cell tracking compared with specialized trackers. The full set draws on 15 standard cell-tracking datasets across four imaging modalities and contains 1,537 problems corresponding to 9,041 cell-linking decisions. A harder subset was chosen from failures shared by distinct algorithmic trackers. The authors evaluated VLM environment harnesses, benchmarked 9 frontier closed-source models and 13 open-weight models, and analyzed failure modes. Code, stored outputs, and a regeneration pipeline are on GitHub.
- The hard subset targets cases where distinct algorithmic trackers fail.
- Nine closed-source and 13 open-weight models were benchmarked.
- Different environment harnesses for VLMs were compared.
- Regeneration code is at github.com/royerlab/celltrack-bench.
From the original article
1 min read at huggingface.co
Large language models have shown outstanding performance on text-first tasks such as coding, math, and writing. Recent efforts are shifting models' training to include scientific problems that often live in highly specialized domains beyond text, such as protein structure and biological images. However, in these scenarios, their general-purpose accuracy compared with specialized algorithms remains in question.
By Ilan Theodoro, Jordão Bragantini, Loïc A. Royer · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.