SE-MoLoRA: Shared-Expert LoRA Adapters for Domain-Specific Photographic Assessment
Vision-language models can describe images fluently, but they often fail to provide actionable photographic critique because semantic content and aesthetic judgment remain entangled. We propose SE-MoLoRA, a modular parameter-efficient adaptation framework for domain-specific photographic assessment. The method separates general photographic knowledge from specialist residual judgments using an always-active shared LoRA expert and routed adapters for composition, lighting, and technical quality. A lightweight query router selects the relevant specialist, enabling targeted critique without training separate full models. A rank-64 shared adapter captures broad photographic vocabulary, while ...
| Authors | Bishwash Khanal, Anlan Zhang, Sasu Tarkoma et al. |
|---|---|
| Primary category | cs.CV |
| Published | 18 August 2026 |
| Last updated at source | 19 August 2026 |
Silytics shows public metadata from arXiv and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.