COMBINER: Composed Image Retrieval Guided by Attribute-based Neighbor Relations

Composed Image Retrieval (CIR) represents a challenging retrieval task that targets locating specific images through multimodal inputs. Despite recent progress in CIR techniques, prior approaches often overlook cases where images appear visually alike yet differ in attributes, potentially undermining both multimodal feature fusion and similarity modeling. To mitigate this limitation, we design a unified representation of cross-modal features based on attribute prototypes. Nevertheless, the task is far from straightforward, owing to three core issues: (1) entanglement in attribute-level semantics, (2) inconsistency across modalities, and (3) supervised signal missing. To tackle the above obstacles, we introduce a COMposed image retrieval network guided By attrIbute-based NEighbor Relations (COMBINER). Specifically, we first design an Adaptive Semantic Disentanglement module, which is capable of disentangling attribute features based on multimodal primitive features. Secondly, we propose a Unified Prototype-based Composition module, which can construct cross-modal unified prototypes (CUP) and facilitate multimodal feature composition. Finally, we introduce a Dual Relations Modeling module, which can mine pairwise and neighbor relations based on attribute similarity. Compared to traditional neighbor relations modeling CIR methods, COMBINER represents the first study addressing the phenomenon of visually similar but attribute-unrelated samples. It achieves a more accurate understanding of the semantic relations among samples by employing an attribute prototype-based similarity metric. Comprehensive experiments conducted on three benchmark datasets confirm the effectiveness of our proposed COMBINER. The implementation of our method will be accessed at https://github.com/Lee-zixu/COMBINER

FashionLens: Toward Versatile Fashion Image Retrieval via Task-Adaptive Learning

Fashion image retrieval is a cornerstone of modern e-commerce systems. A unified framework that supports diverse query formats and search intentions is highly desired in practice. However, existing approaches focus on narrow retrieval tasks and do not fully capture such diversity. Therefore, in this work, we aim to develop a unified framework capable of handling diverse realistic fashion retrieval scenarios, achieving truly versatile fashion image retrieval. To establish a data foundation, we first introduce U-FIRE, a comprehensive benchmark that consolidates fragmented fashion datasets into a unified collection, supplemented by two manually curated datasets for testing generalization. Building upon this, we propose FashionLens, a unified framework based on Multimodal Large Language Models. To handle divergent matching objectives, we design a Proposal-Guided Spherical Query Calibrator that dynamically shifts query representations into task-aligned metric spaces via adaptive spherical linear interpolation. Additionally, to mitigate the optimization imbalance caused by varying task complexities and data scales, we develop a Gradient-Guided Adaptive Sampling strategy that automatically re-weights tasks based on realtime learning difficulty and the data scale prior. Experiments on U-FIRE show that FashionLens achieves state-of-the-art performance across diverse retrieval scenarios and generalizes robustly to unseen tasks. The data and code are publicly released at https://github.com/haokunwen/FashionLens.

OSGNet with MLLM Reranking @ Ego4D Episodic Memory Challenge 2026

In this report, we present our champion solutions for the Natural Language Queries and GoalStep tracks of the Ego4D Episodic Memory Challenge at CVPR 2026. Both tracks require accurately localizing temporal segments from long untrimmed egocentric videos. To address these tasks, we propose a reranking-based framework that effectively leverages the strong video-language reasoning capability of multimodal large language model (MLLM) while preserving the efficiency and candidate recall of conventional localization pipelines. Specifically, we first obtain a set of candidate segments from existing localization model OSGNet, and then employ MLLM to select the segment that best matches the given query, thereby refining the final prediction. Ultimately, our method achieved first place in both the Natural Language Queries and GoalStep tracks. Our code can be found at https://github.com/iLearn-Lab/CVPR25-OSGNet.

UniCVR: From Alignment to Reranking for Unified Zero-Shot Composed Visual Retrieval

Composed image retrieval, multi-turn composed image retrieval, and composed video retrieval all share a common paradigm: composing the reference visual with modification text to retrieve the desired target. Despite this shared structure, the three tasks have been studied in isolation, with no prior work proposing a unified framework, let alone a zero-shot solution. In this paper, we propose UniCVR, the first unified zero-shot composed visual retrieval framework that jointly addresses all three tasks without any task-specific human-annotated data. UniCVR strategically combines two complementary strengths: Multimodal Large Language Models (MLLMs) for compositional query understanding and Vision-Language Pre-trained (VLP) models for structured visual retrieval. Concretely, UniCVR operates in two stages. In Stage I, we train the MLLM as a compositional query embedder via contrastive learning on a curated multi-source dataset of approximately 3.5M samples, bridging the heterogeneous embedding spaces between the MLLM and the frozen VLP gallery encoder. A cluster-based hard negative sampling strategy is proposed to strengthen contrastive supervision. In Stage II, we introduce an MLLM-guided dual-level reranking mechanism that applies adaptive budgeted subset scoring to a small number of top-ranked candidates, and then exploits the resulting relevance signals through a dual-level re-scoring scheme, producing more accurate final rankings with minimal computational overhead. Extensive experiments across five benchmarks covering all three tasks demonstrate that UniCVR achieves cutting-edge performance, validating its effectiveness and generalizability. Our data and code will be released upon acceptance.

MELT: Improve Composed Image Retrieval via the Modification Frequentation-Rarity Balance Network

Composed Image Retrieval (CIR) uses a reference image and a modification text as a query to retrieve a target image satisfying the requirement of ``modifying the reference image according to the text instructions''. However, existing CIR methods face two limitations: (1) frequency bias leading to ``Rare Sample Neglect'', and (2) susceptibility of similarity scores to interference from hard negative samples and noise. To address these limitations, we confront two key challenges: asymmetric rare semantic localization and robust similarity estimation under hard negative samples. To solve these challenges, we propose the Modification frEquentation-rarity baLance neTwork MELT. MELT assigns increased attention to rare modification semantics in multimodal contexts while applying diffusion-based denoising to hard negative samples with high similarity scores, enhancing multimodal fusion and matching. Extensive experiments on two CIR benchmarks validate the superior performance of MELT. Codes are available at https://github.com/luckylittlezhi/MELT.

VideoTemp-o3: Harmonizing Temporal Grounding and Video Understanding in Agentic Thinking-with-Videos

In long-video understanding, conventional uniform frame sampling often fails to capture key visual evidence, leading to degraded performance and increased hallucinations. To address this, recent agentic thinking-with-videos paradigms have emerged, adopting a localize-clip-answer pipeline in which the model actively identifies relevant video segments, performs dense sampling within those clips, and then produces answers. However, existing methods remain inefficient, suffer from weak localization, and adhere to rigid workflows. To solve these issues, we propose VideoTemp-o3, a unified agentic thinking-with-videos framework that jointly models video grounding and question answering. VideoTemp-o3 exhibits strong localization capability, supports on-demand clipping, and can refine inaccurate localizations. Specifically, in the supervised fine-tuning stage, we design a unified masking mechanism that encourages exploration while preventing noise. For reinforcement learning, we introduce dedicated rewards to mitigate reward hacking. Besides, from the data perspective, we develop an effective pipeline to construct high-quality long video grounded QA data, along with a corresponding benchmark for systematic evaluation across various video durations. Experimental results demonstrate that our method achieves remarkable performance on both long video understanding and grounding.

Dual Knowledge-Enhanced Two-Stage Reasoner for Multimodal Dialog Systems

Textual response generation is pivotal for multimodal \mbox{task-oriented} dialog systems, which aims to generate proper textual responses based on the multimodal context. While existing efforts have demonstrated remarkable progress, there still exist the following limitations: 1) \textit{neglect of unstructured review knowledge} and 2) \textit{underutilization of large language models (LLMs)}. Inspired by this, we aim to fully utilize dual knowledge (\textit{i.e., } structured attribute and unstructured review knowledge) with LLMs to promote textual response generation in multimodal task-oriented dialog systems. However, this task is non-trivial due to two key challenges: 1) \textit{dynamic knowledge type selection} and 2) \textit{intention-response decoupling}. To address these challenges, we propose a novel dual knowledge-enhanced two-stage reasoner by adapting LLMs for multimodal dialog systems (named DK2R). To be specific, DK2R first extracts both structured attribute and unstructured review knowledge from external knowledge base given the dialog context. Thereafter, DK2R uses an LLM to evaluate each knowledge type's utility by analyzing LLM-generated provisional probe responses. Moreover, DK2R separately summarizes the intention-oriented key clues via dedicated reasoning, which are further used as auxiliary signals to enhance LLM-based textual response generation. Extensive experiments conducted on a public dataset verify the superiority of DK2R. We have released the codes and parameters.

Mitigating Hallucination Through Theory-Consistent Symmetric Multimodal Preference Optimization

Direct Preference Optimization (DPO) has emerged as an effective approach for mitigating hallucination in Multimodal Large Language Models (MLLMs). Although existing methods have achieved significant progress by utilizing vision-oriented contrastive objectives for enhancing MLLMs' attention to visual inputs and hence reducing hallucination, they suffer from non-rigorous optimization objective function and indirect preference supervision. To address these limitations, we propose a Symmetric Multimodal Preference Optimization (SymMPO), which conducts symmetric preference learning with direct preference supervision (i.e., response pairs) for visual understanding enhancement, while maintaining rigorous theoretical alignment with standard DPO. In addition to conventional ordinal preference learning, SymMPO introduces a preference margin consistency loss to quantitatively regulate the preference gap between symmetric preference pairs. Comprehensive evaluation across five benchmarks demonstrate SymMPO's superior performance, validating its effectiveness in hallucination mitigation of MLLMs.

Modality Reliability Guided Multimodal Recommendation

Multimodal recommendation faces an issue of the performance degradation that the uni-modal recommendation sometimes achieves the better performance. A possible reason is that the unreliable item modality data hurts the fusion result. Several existing studies have introduced weights for different modalities to reduce the contribution of the unreliable modality data in predicting the final user rating. However, they fail to provide appropriate supervisions for learning the modality weights, making the learned weights imprecise. Therefore, we propose a modality reliability guided multimodal recommendation framework that uniquely learns the modality weights supervised by the modality reliability. Considering that there is no explicit label provided for modality reliability, we resort to automatically identify it through the BPR recommendation objective. In particular, we define a modality reliability vector as the supervision label by the difference between modality-specific user ratings to positive and negative items, where a larger difference indicates a higher reliability of the modality as the BPR objective is better satisfied. Furthermore, to enhance the effectiveness of the supervision, we calculate the confidence level for the modality reliability vector, which dynamically adjusts the supervision strength and eliminates the harmful supervision. Extensive experiments on three real-world datasets show the effectiveness of the proposed method.

Fine-grained Textual Inversion Network for Zero-Shot Composed Image Retrieval

Composed Image Retrieval (CIR) allows users to search target images with a multimodal query, comprising a reference image and a modification text that describes the user's modification demand over the reference image. Nevertheless, due to the expensive labor cost of training data annotation, recent researchers have shifted to the challenging task of zero-shot CIR (ZS-CIR), which targets fulfilling CIR without annotated triplets. The pioneer ZS-CIR studies focus on converting the CIR task into a standard text-to-image retrieval task by pre-training a textual inversion network that can map a given image into a single pseudo-word token. Despite their significant progress, their coarse-grained textual inversion may be insufficient to capture the full content of the image accurately. To overcome this issue, in this work, we propose a novel Fine-grained Textual Inversion Network for ZS-CIR, named FTI4CIR. In particular, FTI4CIR comprises two main components: fine-grained pseudo-word token mapping and tri-wise caption-based semantic regularization. The former maps the image into a subject-oriented pseudo-word token and several attribute-oriented pseudo-word tokens to comprehensively express the image in the textual form, while the latter works on jointly aligning the fine-grained pseudo-word tokens to the real-word token embedding space based on a BLIP-generated image caption template. Extensive experiments conducted on three benchmark datasets demonstrate the superiority of our proposed method.