# Learning Consistent Taxonomic Classification through Hierarchical Reasoning

Source: [Learning Consistent Taxonomic Classification through Hierarchical Reasoning](<https://www.sota2.com/research/paper/6971af05e25ad2fa9d1ed94a>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-23T05:00:04.919Z

Authors: Zhenghong Li, Kecheng Zheng, Haibin Ling

Published: 2026-01-21T03:00:00.000Z

## Abstract

While Vision-Language Models (VLMs) excel at visual understanding, they often fail to grasp hierarchical knowledge. This leads to common errors where VLMs misclassify coarser taxonomic levels even when correctly identifying the most specific level (leaf level). Existing approaches largely overlook this issue by failing to model hierarchical reasoning. To address this gap, we propose VL-Taxon, a two-stage, hierarchy-based reasoning framework designed to improve both leaf-level accuracy and hierarchical consistency in taxonomic classification. The first stage employs a top-down process to enhance leaf-level classification accuracy. The second stage then leverages this accurate leaf-level output to ensure consistency throughout the entire taxonomic hierarchy. Each stage is initially trained with supervised fine-tuning to instill taxonomy knowledge, followed by reinforcement learning to refine the model's reasoning and generalization capabilities. Extensive experiments reveal a remarkable result: our VL-Taxon framework, implemented on the Qwen2.5-VL-7B model, outperforms its original 72B counterpart by over 10% in both leaf-level and hierarchical consistency accuracy on average on the iNaturalist-2021 dataset. Notably, this significant gain was achieved by fine-tuning on just a small subset of data, without relying on any examples generated by other VLMs.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Taxonomic Classification on iNat Animal 21](<https://www.sota2.com/research/sota/taxonomic-classification-on-inat-animal-21.md>): Taxonomic Classification; iNat Animal 21; HCA: 43.73.

- [Taxonomic Classification on iNat Plant 21](<https://www.sota2.com/research/sota/taxonomic-classification-on-inat-plant-21.md>): Taxonomic Classification; iNat Plant 21; HCA: 63.04.

- [Open-set taxonomic classification on iNat Animal 2021 (test)](<https://www.sota2.com/research/sota/open-set-taxonomic-classification-on-inat-animal-2021-test.md>): Open-set taxonomic classification; iNat Animal 2021 (test); HCA: 1527.

- [Taxonomic Classification on CUB-200](<https://www.sota2.com/research/sota/taxonomic-classification-on-cub-200.md>): Taxonomic Classification; CUB-200; HCA: 60.67.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2601.14610>)

[PDF](<https://arxiv.org/pdf/2601.14610.pdf>)
