# MLLM-Guided VLM Fine-Tuning with Joint Inference for Zero-Shot Composed Image Retrieval

arXiv: https://arxiv.org/abs/2505.19707 · HTML: https://mlsanity.com/paper/2505.19707/

## Key facts

- Cited by 5 papers in the map (data through 14 Oct 2026).
- Builds on 38 of its 58 references that are in the map.
- First posted on arXiv 26 May 2025.

## Builds on (most cited first, 30 of 38)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [ImageNet Large Scale Visual Recognition Challenge](https://mlsanity.com/paper/1409.0575/) (2014)
- [Representation Learning with Contrastive Predictive Coding](https://mlsanity.com/paper/1807.03748/) (2018)
- [BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models](https://mlsanity.com/paper/2301.12597/) (2023)
- [InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks](https://mlsanity.com/paper/2312.14238/) (2023)
- [Align before Fuse: Vision and Language Representation Learning with Momentum Distillation](https://mlsanity.com/paper/2107.07651/) (2021)
- [MiniCPM-V: A GPT-4V Level MLLM on Your Phone](https://mlsanity.com/paper/2408.01800/) (2024)
- [MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategies](https://mlsanity.com/paper/2404.06395/) (2024)
- [A Corpus for Reasoning About Natural Language Grounded in Photographs](https://mlsanity.com/paper/1811.00491/) (2018)
- [InternLM-XComposer2: Mastering Free-form Text-Image Composition and Comprehension in Vision-Language Large Model](https://mlsanity.com/paper/2401.16420/) (2024)
- [Image Retrieval on Real-life Images with Pre-trained Vision-and-Language Models](https://mlsanity.com/paper/2108.04024/) (2021)
- [Enhancing the Reasoning Ability of Multimodal Large Language Models via Mixed Preference Optimization](https://mlsanity.com/paper/2411.10442/) (2024)
- [Composing Text and Image for Image Retrieval - An Empirical Odyssey](https://mlsanity.com/paper/1812.07119/) (2018)
- [Large Language Model Agent: A Survey on Methodology, Applications and Challenges](https://mlsanity.com/paper/2503.21460/) (2025)
- [Zero-Shot Composed Image Retrieval with Textual Inversion](https://mlsanity.com/paper/2303.15247/) (2023)
- [Pic2Word: Mapping Pictures to Words for Zero-shot Composed Image Retrieval](https://mlsanity.com/paper/2302.03084/) (2023)
- [Grounding Language Models to Images for Multimodal Inputs and Outputs](https://mlsanity.com/paper/2301.13823/) (2023)
- [Vision-by-Language for Training-Free Compositional Image Retrieval](https://mlsanity.com/paper/2310.09291/) (2023)
- [Sentence-level Prompts Benefit Composed Image Retrieval](https://mlsanity.com/paper/2310.05473/) (2023)
- [Composed Image Retrieval using Contrastive Learning and Task-oriented CLIP-based Features](https://mlsanity.com/paper/2308.11485/) (2023)
- [Context-I2W: Mapping Images to Context-dependent Words for Accurate Zero-Shot Composed Image Retrieval](https://mlsanity.com/paper/2309.16137/) (2023)
- [Target-Guided Composed Image Retrieval](https://mlsanity.com/paper/2309.01366/) (2023)
- [Knowledge-Enhanced Dual-stream Zero-shot Composed Image Retrieval](https://mlsanity.com/paper/2403.16005/) (2024)
- [Language-only Efficient Training of Zero-shot Composed Image Retrieval](https://mlsanity.com/paper/2312.01998/) (2023)
- [Simple but Effective Raw-Data Level Multimodal Fusion for Composed Image Retrieval](https://mlsanity.com/paper/2404.15875/) (2024)
- [iSEARLE: Improving Textual Inversion for Zero-Shot Composed Image Retrieval](https://mlsanity.com/paper/2405.02951/) (2024)
- [Self-Training Boosted Multi-Factor Matching Network for Composed Image Retrieval](https://mlsanity.com/paper/2305.09979/) (2023)
- [Where Does the Performance Improvement Come From? -- A Reproducibility Concern about Image-Text Retrieval](https://mlsanity.com/paper/2203.03853/) (2022)

## Built on by (most cited first, 5 of 5)

- [SPAZER: Spatial-Semantic Progressive Reasoning Agent for Zero-shot 3D Visual Grounding](https://mlsanity.com/paper/2506.21924/) (2025)
- [CoVR-R:Reason-Aware Composed Video Retrieval](https://mlsanity.com/paper/2603.20190/) (2026)
- [A Simple "Motivation" Can Enhance Reinforcement Finetuning of Large Reasoning Models](https://mlsanity.com/paper/2506.18485/) (2025)
- [Near-Oracle KV Selection via Pre-hoc Sparsity for Long-Context Inference](https://mlsanity.com/paper/2602.08329/) (2026)
- [Beyond Similarity: Foundation Models as an Efficient Backbone for Training-Free Composed Video Retrieval](https://mlsanity.com/paper/2609.10008/) (2026)

## Abstract (arXiv)

Existing Zero-Shot Composed Image Retrieval (ZS-CIR) methods typically train adapters that convert reference images into pseudo-text tokens, which are concatenated with the modifying text and processed by frozen text encoders in pretrained VLMs or LLMs. While this design leverages the strengths of large pretrained models, it only supervises the adapter to produce encoder-compatible tokens that loosely preserve visual semantics. Crucially, it does not directly optimize the composed query representation to capture the full intent of the composition or to align with the target semantics, thereby limiting retrieval performance, particularly in cases involving fine-grained or complex visual transformations. To address this problem, we propose MLLM-Guided VLM Fine-Tuning with Joint Inference (MVFT-JI), a novel approach that leverages a pretrained multimodal large language model (MLLM) to construct two complementary training tasks using only unlabeled images: target text retrieval taskand text-to-image retrieval task. By jointly optimizing these tasks, our method enables the VLM to inherently acquire robust compositional retrieval capabilities, supported by the provided theoretical justifications and empirical validation. Furthermore, during inference, we further prompt the MLLM to generate target texts from composed queries and compute retrieval scores by integrating similarities between (i) the composed query and candidate images, and (ii) the MLLM-generated target text and candidate images. This strategy effectively combines the VLM's semantic alignment strengths with the MLLM's reasoning capabilities.