2023-12-28 更新
Black-Box Tuning of Vision-Language Models with Effective Gradient Approximation
Authors:Zixian Guo, Yuxiang Wei, Ming Liu, Zhilong Ji, Jinfeng Bai, Yiwen Guo, Wangmeng Zuo
Parameter-efficient fine-tuning (PEFT) methods have provided an effective way for adapting large vision-language models to specific tasks or scenarios. Typically, they learn a very small scale of parameters for pre-trained models in a white-box formulation, which assumes model architectures to be known and parameters to be accessible. However, large models are often not open-source due to considerations of preventing abuse or commercial factors, hence posing a barrier to the deployment of white-box PEFT methods. To alleviate the dependence on model accessibility, we introduce collaborative black-box tuning (CBBT) for both textual prompt optimization and output feature adaptation for black-box models. Specifically, considering that the backpropagation gradients are blocked, we approximate the gradients of textual prompts by analyzing the predictions with perturbed prompts. Secondly, a lightweight adapter is deployed over the output feature of the inaccessible model, further facilitating the model adaptation process. Empowered with these designs, our CBBT is extensively evaluated on eleven downstream benchmarks and achieves remarkable improvements compared to existing black-box VL adaptation methods. Code is released at https://github.com/guozix/cbbt.
PDF
点此查看论文截图
Align on the Fly: Adapting Chatbot Behavior to Established Norms
Authors:Chunpu Xu, Steffi Chern, Ethan Chern, Ge Zhang, Zekun Wang, Ruibo Liu, Jing Li, Jie Fu, Pengfei Liu
In this paper, we aim to align large language models with the ever-changing, complex, and diverse human values (e.g., social norms) across time and locations. This presents a challenge to existing alignment techniques, such as supervised fine-tuning, which internalize values within model parameters. To overcome this, we propose an On-the-fly Preference Optimization (OPO) method, which is a real-time alignment that works in a streaming way. It employs an external memory to store established rules for alignment, which can constrain LLMs’ behaviors without further training, allowing for convenient updates and customization of human values. We also introduce a scalable evaluation to assess the proposed method more effectively. Experimental results on both human-annotated and auto-generated questions from legal and moral domains indicate the effectiveness of the proposed OPO method. Our code and data are released at https://github.com/GAIR-NLP/OPO.
PDF
点此查看论文截图
ChartBench: A Benchmark for Complex Visual Reasoning in Charts
Authors:Zhengzhuo Xu, Sinan Du, Yiyan Qi, Chengjin Xu, Chun Yuan, Jian Guo
Multimodal Large Language Models (MLLMs) have demonstrated remarkable multimodal understanding and generation capabilities. However, their understanding of synthetic charts is limited, while existing benchmarks are simplistic and the charts deviate significantly from real-world examples, making it challenging to accurately assess MLLMs’ chart comprehension abilities. Hence, a challenging benchmark is essential for investigating progress and uncovering the limitations of current MLLMs on chart data. In this work, we propose to examine chart comprehension through more complex visual logic and introduce ChartBench, a comprehensive chart benchmark to accurately measure MLLMs’ fundamental chart comprehension and data reliability. Specifically, ChartBench consists of \textbf{41} categories, \textbf{2K} charts, and \textbf{16K} QA annotations. While significantly expanding chart types, ChartBench avoids direct labelling of data points, which requires MLLMs to infer values akin to humans by leveraging elements like color, legends, and coordinate systems. We also introduce an improved metric, \textit{Acc+}, which accurately reflects MLLMs’ chart comprehension abilities while avoiding labor-intensive manual evaluations or costly GPT-based evaluations. We conduct evaluations on \textbf{12} mainstream open-source models and \textbf{2} outstanding proprietary models. Through extensive experiments, we reveal the limitations of MLLMs on charts and provide insights to inspire the community to pay closer attention to MLLMs’ chart comprehension abilities. The benchmark and code will be publicly available for research.
PDF
点此查看论文截图
FedMS: Federated Learning with Mixture of Sparsely Activated Foundations Models
Authors:Panlong Wu, Kangshuo Li, Ting Wang, Fangxin Wang
Foundation models have shown great success in natural language processing, computer vision, and multimodal tasks. FMs have a large number of model parameters, thus requiring a substantial amount of data to help optimize the model during the training. Federated learning has revolutionized machine learning by enabling collaborative learning from decentralized data while still preserving the data privacy of clients. Despite the great benefits foundation models can have empowered by federated learning, they face severe computation, communication, and statistical challenges. In this paper, we propose a novel two-stage federated learning algorithm called FedMS. A global expert is trained in the first stage and a local expert is trained in the second stage to provide better personalization. We construct a Mixture of Foundation Models (MoFM) with these two experts and design a gate neural network with an inserted gate adapter that joins the aggregation every communication round in the second stage. To further adapt to edge computing scenarios with limited computational resources, we design a novel Sparsely Activated LoRA (SAL) algorithm that freezes the pre-trained foundation model parameters inserts low-rank adaptation matrices into transformer blocks and activates them progressively during the training. We employ extensive experiments to verify the effectiveness of FedMS, results show that FedMS outperforms other SOTA baselines by up to 55.25% in default settings.
PDF
点此查看论文截图
MoTCoder: Elevating Large Language Models with Modular of Thought for Challenging Programming Tasks
Authors:Jingyao Li, Pengguang Chen, Jiaya Jia
Large Language Models (LLMs) have showcased impressive capabilities in handling straightforward programming tasks. However, their performance tends to falter when confronted with more challenging programming problems. We observe that conventional models often generate solutions as monolithic code blocks, restricting their effectiveness in tackling intricate questions. To overcome this limitation, we present Modular-of-Thought Coder (MoTCoder). We introduce a pioneering framework for MoT instruction tuning, designed to promote the decomposition of tasks into logical sub-tasks and sub-modules. Our investigations reveal that, through the cultivation and utilization of sub-modules, MoTCoder significantly improves both the modularity and correctness of the generated solutions, leading to substantial relative pass@1 improvements of 12.9% on APPS and 9.43% on CodeContests. Our codes are available at https://github.com/dvlab-research/MoTCoder.
PDF Model: https://huggingface.co/JingyaoLi/MoTCoder-15B-v1.0. Code: https://github.com/dvlab-research/MoTCoder
点此查看论文截图
DocMSU: A Comprehensive Benchmark for Document-level Multimodal Sarcasm Understanding
Authors:Hang Du, Guoshun Nan, Sicheng Zhang, Binzhu Xie, Junrui Xu, Hehe Fan, Qimei Cui, Xiaofeng Tao, Xudong Jiang
Multimodal Sarcasm Understanding (MSU) has a wide range of applications in the news field such as public opinion analysis and forgery detection. However, existing MSU benchmarks and approaches usually focus on sentence-level MSU. In document-level news, sarcasm clues are sparse or small and are often concealed in long text. Moreover, compared to sentence-level comments like tweets, which mainly focus on only a few trends or hot topics (e.g., sports events), content in the news is considerably diverse. Models created for sentence-level MSU may fail to capture sarcasm clues in document-level news. To fill this gap, we present a comprehensive benchmark for Document-level Multimodal Sarcasm Understanding (DocMSU). Our dataset contains 102,588 pieces of news with text-image pairs, covering 9 diverse topics such as health, business, etc. The proposed large-scale and diverse DocMSU significantly facilitates the research of document-level MSU in real-world scenarios. To take on the new challenges posed by DocMSU, we introduce a fine-grained sarcasm comprehension method to properly align the pixel-level image features with word-level textual features in documents. Experiments demonstrate the effectiveness of our method, showing that it can serve as a baseline approach to the challenging DocMSU. Our code and dataset are available at https://github.com/Dulpy/DocMSU.
PDF
点此查看论文截图
Large Language Models as Traffic Signal Control Agents: Capacity and Opportunity
Authors:Siqi Lai, Zhao Xu, Weijia Zhang, Hao Liu, Hui Xiong
Traffic signal control is crucial for optimizing the efficiency of road network by regulating traffic light phases. Existing research predominantly focuses on heuristic or reinforcement learning (RL)-based methods, which often lack transferability across diverse traffic scenarios and suffer from poor interpretability. This paper introduces a novel approach, LLMLight, utilizing large language models (LLMs) for traffic signal control tasks. By leveraging LLMs’ impressive generalization and zero-shot reasoning capabilities, LLMLight executes a human-like decision-making process for efficient traffic management. Specifically, the framework begins by composing task descriptions, current traffic conditions, and prior knowledge into a prompt. Subsequently, we utilize LLM’s chain-of-thought (CoT) reasoning ability to identify the next traffic signal phase, ensuring optimal efficiency in the road network. LLMLight achieves state-of-the-art (SOTA) or competitive results across five real-world traffic datasets. Notably, LLMLight showcases remarkable generalization, interpretability, and zero-shot reasoning abilities, even without any training for transportation management tasks. Our project is available at https://github.com/usail-hkust/LLMTSCS.
PDF
点此查看论文截图
LangSplat: 3D Language Gaussian Splatting
Authors:Minghan Qin, Wanhua Li, Jiawei Zhou, Haoqian Wang, Hanspeter Pfister
Human lives in a 3D world and commonly uses natural language to interact with a 3D scene. Modeling a 3D language field to support open-ended language queries in 3D has gained increasing attention recently. This paper introduces LangSplat, which constructs a 3D language field that enables precise and efficient open-vocabulary querying within 3D spaces. Unlike existing methods that ground CLIP language embeddings in a NeRF model, LangSplat advances the field by utilizing a collection of 3D Gaussians, each encoding language features distilled from CLIP, to represent the language field. By employing a tile-based splatting technique for rendering language features, we circumvent the costly rendering process inherent in NeRF. Instead of directly learning CLIP embeddings, LangSplat first trains a scene-wise language autoencoder and then learns language features on the scene-specific latent space, thereby alleviating substantial memory demands imposed by explicit modeling. Existing methods struggle with imprecise and vague 3D language fields, which fail to discern clear boundaries between objects. We delve into this issue and propose to learn hierarchical semantics using SAM, thereby eliminating the need for extensively querying the language field across various scales and the regularization of DINO features. Extensive experiments on open-vocabulary 3D object localization and semantic segmentation demonstrate that LangSplat significantly outperforms the previous state-of-the-art method LERF by a large margin. Notably, LangSplat is extremely efficient, achieving a {\speed} $\times$ speedup compared to LERF at the resolution of 1440 $\times$ 1080. We strongly recommend readers to check out our video results at https://langsplat.github.io
PDF Project Page: https://langsplat.github.io
点此查看论文截图
A bi-objective $ε$-constrained framework for quality-cost optimization in language model ensembles
Authors:Aditi Singla, Aditya Singh, Kanishk Kukreja
We propose an ensembling framework that uses diverse open-sourced Large Language Models (LLMs) to achieve high response quality while maintaining cost efficiency. We formulate a bi-objective optimization problem to represent the quality-cost tradeoff and then introduce an additional budget constraint that reduces the problem to a straightforward 0/1 knapsack problem. We empirically demonstrate that our framework outperforms the existing ensembling approaches in response quality while significantly reducing costs.
PDF
点此查看论文截图
RoleEval: A Bilingual Role Evaluation Benchmark for Large Language Models
Authors:Tianhao Shen, Sun Li, Deyi Xiong
The rapid evolution of large language models (LLMs) necessitates effective benchmarks for evaluating their role knowledge, which is essential for establishing connections with the real world and providing more immersive interactions. This paper introduces RoleEval, a bilingual benchmark designed to assess the memorization, utilization, and reasoning capabilities of role knowledge. RoleEval comprises RoleEval-Global (including internationally recognized characters) and RoleEval-Chinese (including characters popular in China), with 6,000 Chinese-English parallel multiple-choice questions focusing on 300 influential people and fictional characters drawn from a variety of domains including celebrities, anime, comics, movies, TV series, games, and fiction. These questions cover basic knowledge and multi-hop reasoning abilities, aiming to systematically probe various aspects such as personal information, relationships, abilities, and experiences of the characters. To maintain high standards, we perform a hybrid quality check process combining automatic and human verification, ensuring that the questions are diverse, challenging, and discriminative. Our extensive evaluations of RoleEval across various open-source and proprietary large language models, under both the zero- and few-shot settings, reveal insightful findings. Notably, while GPT-4 outperforms other models on RoleEval-Global, Chinese LLMs excel on RoleEval-Chinese, highlighting significant knowledge distribution differences. We expect that RoleEval will highlight the significance of assessing role knowledge for foundation models across various languages and cultural settings.
PDF Our dataset will be available at https://github.com/Magnetic2014/RoleEval
点此查看论文截图
JaColBERT and Hard Negatives, Towards Better Japanese-First Embeddings for Retrieval: Early Technical Report
Authors:Benjamin Clavié
Document retrieval in many languages has been largely relying on multi-lingual models, and leveraging the vast wealth of English training data. In Japanese, the best performing deep-learning based retrieval approaches rely on multilingual dense embeddings. In this work, we introduce (1) a hard-negative augmented version of the Japanese MMARCO dataset and (2) JaColBERT, a document retrieval model built on the ColBERT model architecture, specifically for Japanese. JaColBERT vastly outperform all previous monolingual retrieval approaches and competes with the best multilingual methods, despite unfavourable evaluation settings (out-of-domain vs. in-domain for the multilingual models). JaColBERT reaches an average Recall@10 of 0.813, noticeably ahead of the previous monolingual best-performing model (0.716) and only slightly behind multilingual-e5-base (0.820), though more noticeably behind multilingual-e5-large (0.856). These results are achieved using only a limited, entirely Japanese, training set, more than two orders of magnitudes smaller than multilingual embedding models. We believe these results show great promise to support retrieval-enhanced application pipelines in a wide variety of domains.
PDF
点此查看论文截图
Principled Instructions Are All You Need for Questioning LLaMA-1/2, GPT-3.5/4
Authors:Sondos Mahmoud Bsharat, Aidar Myrzakhan, Zhiqiang Shen
This paper introduces 26 guiding principles designed to streamline the process of querying and prompting large language models. Our goal is to simplify the underlying concepts of formulating questions for various scales of large language models, examining their abilities, and enhancing user comprehension on the behaviors of different scales of large language models when feeding into different prompts. Extensive experiments are conducted on LLaMA-1/2 (7B, 13B and 70B), GPT-3.5/4 to verify the effectiveness of the proposed principles on instructions and prompts design. We hope that this work provides a better guide for researchers working on the prompting of large language models. Project page is available at https://github.com/VILA-Lab/ATLAS.
PDF Github at: https://github.com/VILA-Lab/ATLAS
点此查看论文截图