Mitigating Hallucination in Multimodal Large Language Model via Hallucination-targeted Direct Preference Optimization

Multimodal Large Language Models (MLLMs) are known to hallucinate, which limits their practical applications. Recent works have attempted to apply Direct Preference Optimization (DPO) to enhance the performance of MLLMs, but have shown inconsistent improvements in mitigating hallucinations. To address this issue more effectively, we introduce Hallucination-targeted Direct Preference Optimization (HDPO) to reduce hallucinations in MLLMs. Unlike previous approaches, our method tackles hallucinations from their diverse forms and causes. Specifically, we develop three types of preference pair data targeting the following causes of MLLM hallucinations: (1) insufficient visual capabilities, (2) long context generation, and (3) multimodal conflicts. Experimental results demonstrate that our method achieves superior performance across multiple hallucination evaluation datasets, surpassing most state-of-the-art (SOTA) methods and highlighting the potential of our approach. Ablation studies and in-depth analyses further confirm the effectiveness of our method and suggest the potential for further improvements through scaling up.

Beyond Coarse-Grained Matching in Video-Text Retrieval

Video-text retrieval has seen significant advancements, yet the ability of models to discern subtle differences in captions still requires verification. In this paper, we introduce a new approach for fine-grained evaluation. Our approach can be applied to existing datasets by automatically generating hard negative test captions with subtle single-word variations across nouns, verbs, adjectives, adverbs, and prepositions. We perform comprehensive experiments using four state-of-the-art models across two standard benchmarks (MSR-VTT and VATEX) and two specially curated datasets enriched with detailed descriptions (VLN-UVO and VLN-OOPS), resulting in a number of novel insights: 1) our analyses show that the current evaluation benchmarks fall short in detecting a model's ability to perceive subtle single-word differences, 2) our fine-grained evaluation highlights the difficulty models face in distinguishing such subtle variations. To enhance fine-grained understanding, we propose a new baseline that can be easily combined with current methods. Experiments on our fine-grained evaluations demonstrate that this approach enhances a model's ability to understand fine-grained differences.

Magnifier Prompt: Tackling Multimodal Hallucination via Extremely Simple Instructions

Hallucinations in multimodal large language models (MLLMs) hinder their practical applications. To address this, we propose a Magnifier Prompt (MagPrompt), a simple yet effective method to tackle hallucinations in MLLMs via extremely simple instructions. MagPrompt is based on the following two key principles, which guide the design of various effective prompts, demonstrating robustness: (1) MLLMs should focus more on the image. (2) When there are conflicts between the image and the model's inner knowledge, MLLMs should prioritize the image. MagPrompt is training-free and can be applied to open-source and closed-source models, such as GPT-4o and Gemini-pro. It performs well across many datasets and its effectiveness is comparable or even better than more complex methods like VCD. Furthermore, our prompt design principles and experimental analyses provide valuable insights into multimodal hallucination.

D&M: Enriching E-commerce Videos with Sound Effects by Key Moment Detection and SFX Matching

Videos showcasing specific products are increasingly important for E-commerce. Key moments naturally exist as the first appearance of a specific product, presentation of its distinctive features, the presence of a buying link, etc. Adding proper sound effects (SFX) to these key moments, or video decoration with SFX (VDSFX), is crucial for enhancing the user engaging experience. Previous studies about adding SFX to videos perform video to SFX matching at a holistic level, lacking the ability of adding SFX to a specific moment. Meanwhile, previous studies on video highlight detection or video moment retrieval consider only moment localization, leaving moment to SFX matching untouched. By contrast, we propose in this paper D&M, a unified method that accomplishes key moment detection and moment to SFX matching simultaneously. Moreover, for the new VDSFX task we build a large-scale dataset SFX-Moment from an E-commerce platform. For a fair comparison, we build competitive baselines by extending a number of current video moment detection methods to the new task. Extensive experiments on SFX-Moment show the superior performance of the proposed method over the baselines. Code and data will be released.

ASR-enhanced Multimodal Representation Learning for Cross-Domain Product Retrieval

E-commerce is increasingly multimedia-enriched, with products exhibited in a broad-domain manner as images, short videos, or live stream promotions. A unified and vectorized cross-domain production representation is essential. Due to large intra-product variance and high inter-product similarity in the broad-domain scenario, a visual-only representation is inadequate. While Automatic Speech Recognition (ASR) text derived from the short or live-stream videos is readily accessible, how to de-noise the excessively noisy text for multimodal representation learning is mostly untouched. We propose ASR-enhanced Multimodal Product Representation Learning (AMPere). In order to extract product-specific information from the raw ASR text, AMPere uses an easy-to-implement LLM-based ASR text summarizer. The LLM-summarized text, together with visual data, is then fed into a multi-branch network to generate compact multimodal embeddings. Extensive experiments on a large-scale tri-domain dataset verify the effectiveness of AMPere in obtaining a unified multimodal product representation that clearly improves cross-domain product retrieval.

PhD: A Prompted Visual Hallucination Evaluation Dataset

The rapid growth of Large Language Models (LLMs) has driven the development of Large Vision-Language Models (LVLMs). The challenge of hallucination, prevalent in LLMs, also emerges in LVLMs. However, most existing efforts mainly focus on object hallucination in LVLM, ignoring diverse types of LVLM hallucinations. In this study, we delve into the Intrinsic Vision-Language Hallucination (IVL-Hallu) issue, thoroughly analyzing different types of IVL-Hallu on their causes and reflections. Specifically, we propose several novel IVL-Hallu tasks and categorize them into four types: (a) object hallucination, which arises from the misidentification of objects, (b) attribute hallucination, which is caused by the misidentification of attributes, (c) multi-modal conflicting hallucination, which derives from the contradictions between textual and visual information, and (d) counter-common-sense hallucination, which owes to the contradictions between the LVLM knowledge and actual images. Based on these taxonomies, we propose a more challenging benchmark named PhD to evaluate and explore IVL-Hallu. An automated pipeline is proposed for generating different types of IVL-Hallu data. Extensive experiments on five SOTA LVLMs reveal their inability to effectively tackle our proposed IVL-Hallu tasks, with detailed analyses and insights on the origins and possible solutions of these new challenging IVL-Hallu tasks, facilitating future researches on IVL-Hallu and LVLM. The benchmark can be accessed at https://github.com/jiazhen-code/IntrinsicHallu

Adaptive Fusion of Radiomics and Deep Features for Lung Adenocarcinoma Subtype Recognition

The most common type of lung cancer, lung adenocarcinoma (LUAD), has been increasingly detected since the advent of low-dose computed tomography screening technology. In clinical practice, pre-invasive LUAD (Pre-IAs) should only require regular follow-up care, while invasive LUAD (IAs) should receive immediate treatment with appropriate lung cancer resection, based on the cancer subtype. However, prior research on diagnosing LUAD has mainly focused on classifying Pre-IAs/IAs, as techniques for distinguishing different subtypes of IAs have been lacking. In this study, we proposed a multi-head attentional feature fusion (MHA-FF) model for not only distinguishing IAs from Pre-IAs, but also for distinguishing the different subtypes of IAs. To predict the subtype of each nodule accurately, we leveraged both radiomics and deep features extracted from computed tomography images. Furthermore, those features were aggregated through an adaptive fusion module that can learn attention-based discriminative features. The utility of our proposed method is demonstrated here by means of real-world data collected from a multi-center cohort.

TeachCLIP: Multi-Grained Teaching for Efficient Text-to-Video Retrieval

For text-to-video retrieval (T2VR), which aims to retrieve unlabeled videos by ad-hoc textual queries, CLIP-based methods are dominating. Compared to CLIP4Clip which is efficient and compact, the state-of-the-art models tend to compute video-text similarity by fine-grained cross-modal feature interaction and matching, putting their scalability for large-scale T2VR into doubt. For efficient T2VR, we propose TeachCLIP with multi-grained teaching to let a CLIP4Clip based student network learn from more advanced yet computationally heavy models such as X-CLIP, TS2-Net and X-Pool . To improve the student's learning capability, we add an Attentional frame-Feature Aggregation (AFA) block, which by design adds no extra storage/computation overhead at the retrieval stage. While attentive weights produced by AFA are commonly used for combining frame-level features, we propose a novel use of the weights to let them imitate frame-text relevance estimated by the teacher network. As such, AFA provides a fine-grained learning (teaching) channel for the student (teacher). Extensive experiments on multiple public datasets justify the viability of the proposed method.

Cross-domain Collaborative Learning for Recognizing Multiple Retinal Diseases from Wide-Field Fundus Images

This paper addresses the emerging task of recognizing multiple retinal diseases from wide-field (WF) and ultra-wide-field (UWF) fundus images. For an effective reuse of existing labeled color fundus photo (CFP) data, we propose Cross-domain Collaborative Learning (CdCL). Inspired by the success of fixed-ratio based mixup in unsupervised domain adaptation, we re-purpose this strategy for the current task. Due to the intrinsic disparity between the field-of-view of CFP and WF/UWF images, a scale bias naturally exists in a mixup sample that the anatomic structure from a CFP image will be considerably larger than its WF/UWF counterpart. The CdCL method resolves the issue by Scale-bias Correction, which employs Transformers for producing scale-invariant features. As demonstrated by extensive experiments on multiple datasets covering both WF and UWF images, the proposed method compares favorably against a number of competitive baselines.

Renmin University of China at TRECVID 2022: Improving Video Search by Feature Fusion and Negation Understanding

We summarize our TRECVID 2022 Ad-hoc Video Search (AVS) experiments. Our solution is built with two new techniques, namely Lightweight Attentional Feature Fusion (LAFF) for combining diverse visual / textual features and Bidirectional Negation Learning (BNL) for addressing queries that contain negation cues. In particular, LAFF performs feature fusion at both early and late stages and at both text and video ends to exploit diverse (off-the-shelf) features. Compared to multi-head self attention, LAFF is much more compact yet more effective. Its attentional weights can also be used for selecting fewer features, with the retrieval performance mostly preserved. BNL trains a negation-aware video retrieval model by minimizing a bidirectionally constrained loss per triplet, where a triplet consists of a given training video, its original description and a partially negated description. For video feature extraction, we use pre-trained CLIP, BLIP, BEiT, ResNeXt-101 and irCSN. As for text features, we adopt bag-of-words, word2vec, CLIP and BLIP. Our training data consists of MSR-VTT, TGIF and VATEX that were used in our previous participation. In addition, we automatically caption the V3C1 collection for pre-training. The 2022 edition of the TRECVID benchmark has again been a fruitful participation for the RUCMM team. Our best run, with an infAP of 0.262, is ranked at the second place teamwise.