# Can Edge-Deployable Vision-Language Models Identify Species?

arXiv: https://arxiv.org/abs/2609.11916 · HTML: https://mlsanity.com/paper/2609.11916/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Agricultural vision-language models, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 4 of its 12 references that are in the map.
- First posted on arXiv 10 Sep 2026.

## Placed in

- Multimodal LLMs / vision-language models / [Agricultural vision-language models](https://mlsanity.com/mllm/agricultural-vision-language-models/)

## Builds on (most cited first, 4 of 4)

- [BioCLIP: A Vision Foundation Model for the Tree of Life](https://mlsanity.com/paper/2311.18803/) (2023)
- [Multimodal Foundation Models for Zero-shot Animal Species Recognition in Camera Trap Images](https://mlsanity.com/paper/2311.01064/) (2023)
- [Are We There Yet? A Measurement Study of Efficiency for LLM Applications on Mobile Devices](https://mlsanity.com/paper/2504.00002/) (2025)
- [Future of Edge AI in biodiversity monitoring](https://mlsanity.com/paper/2602.13496/) (2026)

## Built on by (most cited first, 1 of 1)

- [trACT: temporal revelation Airborne Camera Trap](https://mlsanity.com/paper/2610.09417/) (2026)

## Abstract (arXiv)

Camera traps often run in the field on edge hardware with limited or no connectivity, making small, locally-deployable vision-language models (VLMs) -- not frontier-scale ones -- the practically relevant class to evaluate for species identification. We test whether models in this deployment-relevant 2--8B range carry genuine taxonomic knowledge, evaluating four such VLMs (Qwen3-VL 2B/4B/8B, Gemma3 4B) against the domain-specific specialist BioCLIP (300M parameters) on a 96-species task, comparing clean iNaturalist photographs against camera-trap imagery from 6 LILA.science collections, on two independently-sampled evaluation sets. All models identify species far above chance, but every model -- general-purpose or specialist -- degrades sharply on field imagery (domain gaps of 9.6--26.6 percentage points, consistent across taxonomic levels and both evaluation sets), indicating the degradation reflects general image legibility rather than fine-grained discrimination failure. BioCLIP substantially outperforms every VLM tested (by 33.2--59.2 percentage points across an expanded 200-image sample for every model) despite its far smaller size, suggesting the gap reflects specialized training data rather than model scale; yet BioCLIP's own domain gap (18.0 points) is statistically indistinguishable from the best VLM's (22.3 points), suggesting the clean-to-field degradation itself is a property of the image-quality shift rather than a general-purpose-model weakness. Under open-set prompting, 5.9--9.6% of responses are syntactically valid but taxonomically nonexistent species names; the relative fabrication-rate ranking across models replicates exactly across both evaluation sets, a more robust finding than any single point estimate.