AdaMMS: Model Merging for Heterogeneous Multimodal Large Language Models with Unsupervised Coefficient Optimization

Recently, model merging methods have demonstrated powerful strengths in combining abilities on various tasks from multiple Large Language Models (LLMs). While previous model merging methods mainly focus on merging homogeneous models with identical architecture, they meet challenges when dealing with Multimodal Large Language Models (MLLMs) with inherent heterogeneous property, including differences in model architecture and the asymmetry in the parameter space. In this work, we propose AdaMMS, a novel model merging method tailored for heterogeneous MLLMs. Our method tackles the challenges in three steps: mapping, merging and searching. Specifically, we first design mapping function between models to apply model merging on MLLMs with different architecture. Then we apply linear interpolation on model weights to actively adapt the asymmetry in the heterogeneous MLLMs. Finally in the hyper-parameter searching step, we propose an unsupervised hyper-parameter selection method for model merging. As the first model merging method capable of merging heterogeneous MLLMs without labeled data, extensive experiments on various model combinations demonstrated that AdaMMS outperforms previous model merging methods on various vision-language benchmarks.

Beyond Single Frames: Can LMMs Comprehend Temporal and Contextual Narratives in Image Sequences?

Large Multimodal Models (LMMs) have achieved remarkable success across various visual-language tasks. However, existing benchmarks predominantly focus on single-image understanding, leaving the analysis of image sequences largely unexplored. To address this limitation, we introduce StripCipher, a comprehensive benchmark designed to evaluate capabilities of LMMs to comprehend and reason over sequential images. StripCipher comprises a human-annotated dataset and three challenging subtasks: visual narrative comprehension, contextual frame prediction, and temporal narrative reordering. Our evaluation of $16$ state-of-the-art LMMs, including GPT-4o and Qwen2.5VL, reveals a significant performance gap compared to human capabilities, particularly in tasks that require reordering shuffled sequential images. For instance, GPT-4o achieves only 23.93% accuracy in the reordering subtask, which is 56.07% lower than human performance. Further quantitative analysis discuss several factors, such as input format of images, affecting the performance of LLMs in sequential understanding, underscoring the fundamental challenges that remain in the development of LMMs.

SG-FSM: A Self-Guiding Zero-Shot Prompting Paradigm for Multi-Hop Question Answering Based on Finite State Machine

Large Language Models with chain-of-thought prompting, such as OpenAI-o1, have shown impressive capabilities in natural language inference tasks. However, Multi-hop Question Answering (MHQA) remains challenging for many existing models due to issues like hallucination, error propagation, and limited context length. To address these challenges and enhance LLMs' performance on MHQA, we propose the Self-Guiding prompting Finite State Machine (SG-FSM), designed to strengthen multi-hop reasoning abilities. Unlike traditional chain-of-thought methods, SG-FSM tackles MHQA by iteratively breaking down complex questions into sub-questions, correcting itself to improve accuracy. It processes one sub-question at a time, dynamically deciding the next step based on the current context and results, functioning much like an automaton. Experiments across various benchmarks demonstrate the effectiveness of our approach, outperforming strong baselines on challenging datasets such as Musique. SG-FSM reduces hallucination, enabling recovery of the correct final answer despite intermediate errors. It also improves adherence to specified output formats, simplifying evaluation significantly.

FAST-GSC: Fast and Adaptive Semantic Transmission for Generative Semantic Communication

The rapidly evolving field of generative artificial intelligence technology has introduced innovative approaches for developing semantic communication (SemCom) frameworks, leading to the emergence of a new paradigm-generative SemCom (GSC). However, the complex processes involved in semantic extraction and generative inference may result in considerable latency in resource-constrained scenarios. To tackle these issues, we introduce a new GSC framework that involves fast and adaptive semantic transmission (FAST-GSC). This framework incorporates one innovative communication mechanism and two enhancement strategies at the transmitter and receiver, respectively. Aiming to reduce task latency, our communication mechanism enables fast semantic transmission by parallelizing the processes of semantic extraction at the transmitter and inference at the receiver. Preliminary evaluations indicate that while this mechanism effectively reduces task latency, it could potentially compromise task performance. To address this issue, we propose two additional methods for enhancement. First, at the transmitter, we employ reinforcement learning to discern the intrinsic temporal dependencies among the semantic units and design their extraction and transmission sequence accordingly. Second, at the receiver, we design a semantic difference calculation module and propose a sequential conditional denoising approach to alleviate the stringent immediacy requirement for the reception of semantic features. Extensive experiments demonstrate that our proposed architecture achieves a performance score comparable to the conventional GSC architecture while realizing a 52% reduction in residual task latency that extends beyond the fixed inference duration.

FSM: A Finite State Machine Based Zero-Shot Prompting Paradigm for Multi-Hop Question Answering

Large Language Models (LLMs) with chain-of-thought (COT) prompting have demonstrated impressive abilities on simple nature language inference tasks. However, they tend to perform poorly on Multi-hop Question Answering (MHQA) tasks due to several challenges, including hallucination, error propagation and limited context length. We propose a prompting method, Finite State Machine (FSM) to enhance the reasoning capabilities of LLM for complex tasks in addition to improved effectiveness and trustworthiness. Different from COT methods, FSM addresses MHQA by iteratively decomposing a question into multi-turn sub-questions, and self-correcting in time, improving the accuracy of answers in each step. Specifically, FSM addresses one sub-question at a time and decides on the next step based on its current result and state, in an automaton-like format. Experiments on benchmarks show the effectiveness of our method. Although our method performs on par with the baseline on relatively simpler datasets, it excels on challenging datasets like Musique. Moreover, this approach mitigates the hallucination phenomenon, wherein the correct final answer can be recovered despite errors in intermediate reasoning. Furthermore, our method improves LLMs' ability to follow specified output format requirements, significantly reducing the difficulty of answer interpretation and the need for reformatting.

Harnessing the Power of AI-Generated Content for Semantic Communication

Semantic Communication (SemCom) is envisaged as the next-generation paradigm to address challenges stemming from the conflicts between the increasing volume of transmission data and the scarcity of spectrum resources. However, existing SemCom systems face drawbacks, such as low explainability, modality rigidity, and inadequate reconstruction functionality. Recognizing the transformative capabilities of AI-generated content (AIGC) technologies in content generation, this paper explores a pioneering approach by integrating them into SemCom to address the aforementioned challenges. We employ a three-layer model to illustrate the proposed AIGC-assisted SemCom (AIGC-SCM) architecture, emphasizing its clear deviation from existing SemCom. Grounded in this model, we investigate various AIGC technologies with the potential to augment SemCom's performance. In alignment with SemCom's goal of conveying semantic meanings, we also introduce the new evaluation methods for our AIGC-SCM system. Subsequently, we explore communication scenarios where our proposed AIGC-SCM can realize its potential. For practical implementation, we construct a detailed integration workflow and conduct a case study in a virtual reality image transmission scenario. The results demonstrate our ability to maintain a high degree of alignment between the reconstructed content and the original source information, while substantially minimizing the data volume required for transmission. These findings pave the way for further enhancements in communication efficiency and the improvement of Quality of Service. At last, we present future directions for AIGC-SCM studies.

GRA: Detecting Oriented Objects through Group-wise Rotating and Attention

Oriented object detection, an emerging task in recent years, aims to identify and locate objects across varied orientations. This requires the detector to accurately capture the orientation information, which varies significantly within and across images. Despite the existing substantial efforts, simultaneously ensuring model effectiveness and parameter efficiency remains challenging in this scenario. In this paper, we propose a lightweight yet effective Group-wise Rotating and Attention (GRA) module to replace the convolution operations in backbone networks for oriented object detection. GRA can adaptively capture fine-grained features of objects with diverse orientations, comprising two key components: Group-wise Rotating and Group-wise Attention. Group-wise Rotating first divides the convolution kernel into groups, where each group extracts different object features by rotating at a specific angle according to the object orientation. Subsequently, Group-wise Attention is employed to adaptively enhance the object-related regions in the feature. The collaborative effort of these components enables GRA to effectively capture the various orientation information while maintaining parameter efficiency. Extensive experimental results demonstrate the superiority of our method. For example, GRA achieves a new state-of-the-art (SOTA) on the DOTA-v2.0 benchmark, while saving the parameters by nearly 50% compared to the previous SOTA method. Code will be released.

2023 Low-Power Computer Vision Challenge (LPCVC) Summary

This article describes the 2023 IEEE Low-Power Computer Vision Challenge (LPCVC). Since 2015, LPCVC has been an international competition devoted to tackling the challenge of computer vision (CV) on edge devices. Most CV researchers focus on improving accuracy, at the expense of ever-growing sizes of machine models. LPCVC balances accuracy with resource requirements. Winners must achieve high accuracy with short execution time when their CV solutions run on an embedded device, such as Raspberry PI or Nvidia Jetson Nano. The vision problem for 2023 LPCVC is segmentation of images acquired by Unmanned Aerial Vehicles (UAVs, also called drones) after disasters. The 2023 LPCVC attracted 60 international teams that submitted 676 solutions during the submission window of one month. This article explains the setup of the competition and highlights the winners' methods that improve accuracy and shorten execution time.

Weaver: Foundation Models for Creative Writing

This work introduces Weaver, our first family of large language models (LLMs) dedicated to content creation. Weaver is pre-trained on a carefully selected corpus that focuses on improving the writing capabilities of large language models. We then fine-tune Weaver for creative and professional writing purposes and align it to the preference of professional writers using a suit of novel methods for instruction data synthesis and LLM alignment, making it able to produce more human-like texts and follow more diverse instructions for content creation. The Weaver family consists of models of Weaver Mini (1.8B), Weaver Base (6B), Weaver Pro (14B), and Weaver Ultra (34B) sizes, suitable for different applications and can be dynamically dispatched by a routing agent according to query complexity to balance response quality and computation cost. Evaluation on a carefully curated benchmark for assessing the writing capabilities of LLMs shows Weaver models of all sizes outperform generalist LLMs several times larger than them. Notably, our most-capable Weaver Ultra model surpasses GPT-4, a state-of-the-art generalist LLM, on various writing scenarios, demonstrating the advantage of training specialized LLMs for writing purposes. Moreover, Weaver natively supports retrieval-augmented generation (RAG) and function calling (tool usage). We present various use cases of these abilities for improving AI-assisted writing systems, including integration of external knowledge bases, tools, or APIs, and providing personalized writing assistance. Furthermore, we discuss and summarize a guideline and best practices for pre-training and fine-tuning domain-specific LLMs.

STAR-RIS-Assisted Privacy Protection in Semantic Communication System

Semantic communication (SemCom) has emerged as a promising architecture in the realm of intelligent communication paradigms. SemCom involves extracting and compressing the core information at the transmitter while enabling the receiver to interpret it based on established knowledge bases (KBs). This approach enhances communication efficiency greatly. However, the open nature of wireless transmission and the presence of homogeneous KBs among subscribers of identical data type pose a risk of privacy leakage in SemCom. To address this challenge, we propose to leverage the simultaneous transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) to achieve privacy protection in a SemCom system. In this system, the STAR-RIS is utilized to enhance the signal transmission of the SemCom between a base station and a destination user, as well as to covert the signal to interference specifically for the eavesdropper (Eve). Simulation results demonstrate that our generated task-level disturbance outperforms other benchmarks in protecting SemCom privacy, as evidenced by the significantly lower task success rate achieved by Eve.