Probing the Limits of Stylistic Alignment in Vision-Language Models
On the map: Multimodal LLMs / vision-language models → VLM benchmarks and visual reasoning
also in 1 more
Abstract (arXiv)
Vision-language models are increasingly used to generate image captions in specific styles, such as humor or romantic. However, these transformer-based models often struggle with this subjective task in a zero-shot setting. While preference data can be used to align them toward a desired style, such data is expensive to acquire, limiting the ability to explore the models' full capabilities. This work addresses this by studying the data efficiency of aligning small vision-language models to humor and romantic styles. This approach helps to define the performance limits of these models and determine how little preference data is needed to achieve stylistic saturation, benchmarking their capabilities and limitations.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in VLM benchmarks and visual reasoning, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 20 of its 25 references that are in the map. First posted on arXiv 29 Sep 2025.
Builds on
20 of 25 references.
Closest ancestors · 2
Same direction, newest first
- 2025Painting with Words: Elevating Detailed Image Captioning with Benchmark and Alignment Learning10 Mar
- 2023Vision-Language Models for Vision Tasks: A Survey3 Apr
Other references · 18
- 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
- 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 8,065
- 2025Qwen2.5-VL Technical Reportcited by 5,587
- 2021Prefix-Tuning: Optimizing Continuous Prompts for Generationcited by 3,915
- 2016Gaussian Error Linear Units (GELUs)cited by 3,502
- 2019Fine-Tuning Language Models from Human Preferencescited by 2,150
- 2024SimPO: Simple Preference Optimization with a Reference-Free Rewardcited by 963
- 2024ORPO: Monolithic Preference Optimization without Reference Modelcited by 562
- 2024A Systematic Survey of Prompt Engineering in Large Language Models: Techniques and Applicationscited by 431
- 2024Enhancing the Reasoning Ability of Multimodal Large Language Models via Mixed Preference Optimizationcited by 239
- 2024Personalized Steering of Large Language Models: Versatile Steering Vectors Through Bi-directional Preference Optimizationcited by 118
- 2022Do Androids Laugh at Electric Sheep? Humor "Understanding" Benchmarks from The New Yorker Caption Contestcited by 98
6 more
- 2018"Factual" or "Emotional": Stylized Image Captioning with Adaptive Learning and Attentioncited by 26
- 2024Altogether: Image Captioning via Re-aligning Alt-textcited by 17
- 2025Massive Supervised Fine-tuning Experiments Reveal How Data, Layer, and Training Factors Shape LLM Alignment Qualitycited by 14
- 2024Intuitive Fine-Tuning: Towards Simplifying Alignment into a Single Processcited by 8
- 2019Towards Generating Stylized Image Captions via Adversarial Trainingcited by 5
- 2024VickreyFeedback: Cost-efficient Data Construction for Reinforcement Learning from Human Feedbackcited by 2
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2025LOTUS: A Leaderboard for Detailed Image Captioning from Quality to Societal Bias and User Preferences
- 2024Removing Distributional Discrepancies in Captions Improves Image-Text Alignment
- 2025From Pixels to Feelings: Aligning MLLMs with Human Cognitive Perception of Images
- 2021What Vision-Language Models `See' when they See Scenes
- 2025Seeing Through Words, Speaking Through Pixels: Deep Representational Alignment Between Vision and Language Models
- 2024Synth-Align: Improving Trustworthiness in Vision-Language Model with Synthetic Preference Data Alignment
- 2025Evaluating Robustness of Vision-Language Models Under Noisy Conditions
- 2023Improving Compositional Text-to-image Generation with Large Vision-Language Models