LoRA of Change: Learning to Generate LoRA for the Editing Instruction from A Single Before-After Image Pair

In this paper, we propose the LoRA of Change (LoC) framework for image editing with visual instructions, i.e., before-after image pairs. Compared to the ambiguities, insufficient specificity, and diverse interpretations of natural language, visual instructions can accurately reflect users' intent. Building on the success of LoRA in text-based image editing and generation, we dynamically learn an instruction-specific LoRA to encode the "change" in a before-after image pair, enhancing the interpretability and reusability of our model. Furthermore, generalizable models for image editing with visual instructions typically require quad data, i.e., a before-after image pair, along with query and target images. Due to the scarcity of such quad data, existing models are limited to a narrow range of visual instructions. To overcome this limitation, we introduce the LoRA Reverse optimization technique, enabling large-scale training with paired data alone. Extensive qualitative and quantitative experiments demonstrate that our model produces high-quality images that align with user intent and support a broad spectrum of real-world visual instructions.

FoAR: Force-Aware Reactive Policy for Contact-Rich Robotic Manipulation

Contact-rich tasks present significant challenges for robotic manipulation policies due to the complex dynamics of contact and the need for precise control. Vision-based policies often struggle with the skill required for such tasks, as they typically lack critical contact feedback modalities like force/torque information. To address this issue, we propose FoAR, a force-aware reactive policy that combines high-frequency force/torque sensing with visual inputs to enhance the performance in contact-rich manipulation. Built upon the RISE policy, FoAR incorporates a multimodal feature fusion mechanism guided by a future contact predictor, enabling dynamic adjustment of force/torque data usage between non-contact and contact phases. Its reactive control strategy also allows FoAR to accomplish contact-rich tasks accurately through simple position control. Experimental results demonstrate that FoAR significantly outperforms all baselines across various challenging contact-rich tasks while maintaining robust performance under unexpected dynamic disturbances. Project website: https://tonyfang.net/FoAR/

TAPT: Test-Time Adversarial Prompt Tuning for Robust Inference in Vision-Language Models

Large pre-trained Vision-Language Models (VLMs) such as CLIP have demonstrated excellent zero-shot generalizability across various downstream tasks. However, recent studies have shown that the inference performance of CLIP can be greatly degraded by small adversarial perturbations, especially its visual modality, posing significant safety threats. To mitigate this vulnerability, in this paper, we propose a novel defense method called Test-Time Adversarial Prompt Tuning (TAPT) to enhance the inference robustness of CLIP against visual adversarial attacks. TAPT is a test-time defense method that learns defensive bimodal (textual and visual) prompts to robustify the inference process of CLIP. Specifically, it is an unsupervised method that optimizes the defensive prompts for each test sample by minimizing a multi-view entropy and aligning adversarial-clean distributions. We evaluate the effectiveness of TAPT on 11 benchmark datasets, including ImageNet and 10 other zero-shot datasets, demonstrating that it enhances the zero-shot adversarial robustness of the original CLIP by at least 48.9% against AutoAttack (AA), while largely maintaining performance on clean examples. Moreover, TAPT outperforms existing adversarial prompt tuning methods across various backbones, achieving an average robustness improvement of at least 36.6%.

Visual Cue Enhancement and Dual Low-Rank Adaptation for Efficient Visual Instruction Fine-Tuning

Fine-tuning multimodal large language models (MLLMs) presents significant challenges, including a reliance on high-level visual features that limits fine-grained detail comprehension, and data conflicts that arise from task complexity. To address these issues, we propose an efficient fine-tuning framework with two novel approaches: Vision Cue Enhancement (VCE) and Dual Low-Rank Adaptation (Dual-LoRA). VCE enhances the vision projector by integrating multi-level visual cues, improving the model's ability to capture fine-grained visual features. Dual-LoRA introduces a dual low-rank structure for instruction tuning, decoupling learning into skill and task spaces to enable precise control and efficient adaptation across diverse tasks. Our method simplifies implementation, enhances visual comprehension, and improves adaptability. Experiments on both downstream tasks and general benchmarks demonstrate the effectiveness of our proposed approach.

Retrieval Augmented Recipe Generation

Given the potential applications of generating recipes from food images, this area has garnered significant attention from researchers in recent years. Existing works for recipe generation primarily utilize a two-stage training method, first generating ingredients and then obtaining instructions from both the image and ingredients. Large Multi-modal Models (LMMs), which have achieved notable success across a variety of vision and language tasks, shed light to generating both ingredients and instructions directly from images. Nevertheless, LMMs still face the common issue of hallucinations during recipe generation, leading to suboptimal performance. To tackle this, we propose a retrieval augmented large multimodal model for recipe generation. We first introduce Stochastic Diversified Retrieval Augmentation (SDRA) to retrieve recipes semantically related to the image from an existing datastore as a supplement, integrating them into the prompt to add diverse and rich context to the input image. Additionally, Self-Consistency Ensemble Voting mechanism is proposed to determine the most confident prediction recipes as the final output. It calculates the consistency among generated recipe candidates, which use different retrieval recipes as context for generation. Extensive experiments validate the effectiveness of our proposed method, which demonstrates state-of-the-art (SOTA) performance in recipe generation tasks on the Recipe1M dataset.

Domain Expansion and Boundary Growth for Open-Set Single-Source Domain Generalization

Open-set single-source domain generalization aims to use a single-source domain to learn a robust model that can be generalized to unknown target domains with both domain shifts and label shifts. The scarcity of the source domain and the unknown data distribution of the target domain pose a great challenge for domain-invariant feature learning and unknown class recognition. In this paper, we propose a novel learning approach based on domain expansion and boundary growth to expand the scarce source samples and enlarge the boundaries across the known classes that indirectly broaden the boundary between the known and unknown classes. Specifically, we achieve domain expansion by employing both background suppression and style augmentation on the source data to synthesize new samples. Then we force the model to distill consistent knowledge from the synthesized samples so that the model can learn domain-invariant information. Furthermore, we realize boundary growth across classes by using edge maps as an additional modality of samples when training multi-binary classifiers. In this way, it enlarges the boundary between the inliers and outliers, and consequently improves the unknown class recognition during open-set generalization. Extensive experiments show that our approach can achieve significant improvements and reach state-of-the-art performance on several cross-domain image classification datasets.

A Prompt Refinement-based Large Language Model for Metro Passenger Flow Forecasting under Delay Conditions

Accurate short-term forecasts of passenger flow in metro systems under delay conditions are crucial for emergency response and service recovery, which pose significant challenges and are currently under-researched. Due to the rare occurrence of delay events, the limited sample size under delay condictions make it difficult for conventional models to effectively capture the complex impacts of delays on passenger flow, resulting in low forecasting accuracy. Recognizing the strengths of large language models (LLMs) in few-shot learning due to their powerful pre-training, contextual understanding, ability to perform zero-shot and few-shot reasoning, to address the issues that effectively generalize and adapt with minimal data, we propose a passenger flow forecasting framework under delay conditions that synthesizes an LLM with carefully designed prompt engineering. By Refining prompt design, we enable the LLM to understand delay event information and the pattern from historical passenger flow data, thus overcoming the challenges of passenger flow forecasting under delay conditions. The propmpt engineering in the framework consists of two main stages: systematic prompt generation and prompt refinement. In the prompt generation stage, multi-source data is transformed into descriptive texts understandable by the LLM and stored. In the prompt refinement stage, we employ the multidimensional Chain of Thought (CoT) method to refine the prompts. We verify the proposed framework by conducting experiments using real-world datasets specifically targeting passenger flow forecasting under delay conditions of Shenzhen metro in China. The experimental results demonstrate that the proposed model performs particularly well in forecasting passenger flow under delay conditions.

Towards Effective Utilization of Mixed-Quality Demonstrations in Robotic Manipulation via Segment-Level Selection and Optimization

Data is crucial for robotic manipulation, as it underpins the development of robotic systems for complex tasks. While high-quality, diverse datasets enhance the performance and adaptability of robotic manipulation policies, collecting extensive expert-level data is resource-intensive. Consequently, many current datasets suffer from quality inconsistencies due to operator variability, highlighting the need for methods to utilize mixed-quality data effectively. To mitigate these issues, we propose "Select Segments to Imitate" (S2I), a framework that selects and optimizes mixed-quality demonstration data at the segment level, while ensuring plug-and-play compatibility with existing robotic manipulation policies. The framework has three components: demonstration segmentation dividing origin data into meaningful segments, segment selection using contrastive learning to find high-quality segments, and trajectory optimization to refine suboptimal segments for better policy learning. We evaluate S2I through comprehensive experiments in simulation and real-world environments across six tasks, demonstrating that with only 3 expert demonstrations for reference, S2I can improve the performance of various downstream policies when trained with mixed-quality demonstrations. Project website: https://tonyfang.net/s2i/.

EAGLE: Towards Efficient Arbitrary Referring Visual Prompts Comprehension for Multimodal Large Language Models

Recently, Multimodal Large Language Models (MLLMs) have sparked great research interests owing to their exceptional content-reasoning and instruction-following capabilities. To effectively instruct an MLLM, in addition to conventional language expressions, the practice of referring to objects by painting with brushes on images has emerged as a prevalent tool (referred to as "referring visual prompts") due to its efficacy in aligning the user's intention with specific image regions. To accommodate the most common referring visual prompts, namely points, boxes, and masks, existing approaches initially utilize specialized feature encoding modules to capture the semantics of the highlighted areas indicated by these prompts. Subsequently, these encoded region features are adapted to MLLMs through fine-tuning on a meticulously curated multimodal instruction dataset. However, such designs suffer from redundancy in architecture. Moreover, they face challenges in effectively generalizing when encountering a diverse range of arbitrary referring visual prompts in real-life scenarios. To address the above issues, we propose EAGLE, a novel MLLM that empowers comprehension of arbitrary referring visual prompts with less training efforts than existing approaches. Specifically, our EAGLE maintains the innate format of the referring visual prompts as colored patches rendered on the given image for conducting the instruction tuning. Our approach embeds referring visual prompts as spatial concepts conveying specific spatial areas comprehensible to the MLLM, with the semantic comprehension of these regions originating from the MLLM itself. Besides, we also propose a Geometry-Agnostic Learning paradigm (GAL) to further disentangle the MLLM's region-level comprehension with the specific formats of referring visual prompts. Extensive experiments are conducted to prove the effectiveness of our proposed method.

EventHallusion: Diagnosing Event Hallucinations in Video LLMs

Recently, Multimodal Large Language Models (MLLMs) have made significant progress in the video comprehension field. Despite remarkable content reasoning and instruction following capabilities they demonstrated, the hallucination problem of these VideoLLMs is less explored compared with its counterpart in the image domain. To mitigate this gap, we first propose EventHallusion, a novel benchmark that focuses on assessing the VideoLMMs' hallucination phenomenon on video event comprehension. Based on the observation that existing VideoLLMs are entangled with the priors stemming from their foundation models, our EventHallusion is curated by meticulously collecting videos and annotating questions to intentionally mislead the VideoLLMs into interpreting events based on these priors rather than accurately understanding the video content. On the other hand, we also propose a simple yet effective method, called Temporal Contrastive Decoding (TCD), to tackle the hallucination problems of VideoLLMs. The proposed TCD suppresses the model's preference toward their priors by comparing the original video with a constructed counterpart, whose temporal cues are disrupted, during the autoregressive decoding stage. Through comprehensive evaluation of eight open-source and two closed-source VideoLLMs on the proposed EventHallusion benchmark, we find that the open-source models suffer significantly from hallucination problems, whereas the closed-source models perform markedly better. By further equipping open-sourced VideoLLMs with the proposed TCD approach, evident performance improvements are achieved across most metrics in the EventHallusion benchmark. Our codes and benchmark data are available at https://github.com/Stevetich/EventHallusion.