RAP: Retrieval-Augmented Planning with Contextual Memory for Multimodal LLM Agents

Owing to recent advancements, Large Language Models (LLMs) can now be deployed as agents for increasingly complex decision-making applications in areas including robotics, gaming, and API integration. However, reflecting past experiences in current decision-making processes, an innate human behavior, continues to pose significant challenges. Addressing this, we propose Retrieval-Augmented Planning (RAP) framework, designed to dynamically leverage past experiences corresponding to the current situation and context, thereby enhancing agents' planning capabilities. RAP distinguishes itself by being versatile: it excels in both text-only and multimodal environments, making it suitable for a wide range of tasks. Empirical evaluations demonstrate RAP's effectiveness, where it achieves SOTA performance in textual scenarios and notably enhances multimodal LLM agents' performance for embodied tasks. These results highlight RAP's potential in advancing the functionality and applicability of LLM agents in complex, real-world applications.

One Student Knows All Experts Know: From Sparse to Dense

Human education system trains one student by multiple experts. Mixture-of-experts (MoE) is a powerful sparse architecture including multiple experts. However, sparse MoE model is hard to implement, easy to overfit, and not hardware-friendly. In this work, inspired by human education model, we propose a novel task, knowledge integration, to obtain a dense student model (OneS) as knowledgeable as one sparse MoE. We investigate this task by proposing a general training framework including knowledge gathering and knowledge distillation. Specifically, we first propose Singular Value Decomposition Knowledge Gathering (SVD-KG) to gather key knowledge from different pretrained experts. We then refine the dense student model by knowledge distillation to offset the noise from gathering. On ImageNet, our OneS preserves $61.7\%$ benefits from MoE. OneS can achieve $78.4\%$ top-1 accuracy with only $15$M parameters. On four natural language processing datasets, OneS obtains $88.2\%$ MoE benefits and outperforms SoTA by $51.7\%$ using the same architecture and training data. In addition, compared with the MoE counterpart, OneS can achieve $3.7 \times$ inference speedup due to the hardware-friendly architecture.

Sparse-MLP: A Fully-MLP Architecture with Conditional Computation

Mixture-of-Experts (MoE) with sparse conditional computation has been proved an effective architecture for scaling attention-based models to more parameters with comparable computation cost. In this paper, we propose Sparse-MLP, scaling the recent MLP-Mixer model with sparse MoE layers, to achieve a more computation-efficient architecture. We replace a subset of dense MLP blocks in the MLP-Mixer model with Sparse blocks. In each Sparse block, we apply two stages of MoE layers: one with MLP experts mixing information within channels along image patch dimension, one with MLP experts mixing information within patches along the channel dimension. Besides, to reduce computational cost in routing and improve expert capacity, we design Re-represent layers in each Sparse block. These layers are to re-scale image representations by two simple but effective linear transformations. When pre-training on ImageNet-1k with MoCo v3 algorithm, our models can outperform dense MLP models by 2.5\% on ImageNet Top-1 accuracy with fewer parameters and computational cost. On small-scale downstream image classification tasks, i.e. Cifar10 and Cifar100, our Sparse-MLP can still achieve better performance than baselines.

Go Wider Instead of Deeper

The transformer has recently achieved impressive results on various tasks. To further improve the effectiveness and efficiency of the transformer, there are two trains of thought among existing works: (1) going wider by scaling to more trainable parameters; (2) going shallower by parameter sharing or model compressing along with the depth. However, larger models usually do not scale well when fewer tokens are available to train, and advanced parallelisms are required when the model is extremely large. Smaller models usually achieve inferior performance compared to the original transformer model due to the loss of representation power. In this paper, to achieve better performance with fewer trainable parameters, we propose a framework to deploy trainable parameters efficiently, by going wider instead of deeper. Specially, we scale along model width by replacing feed-forward network (FFN) with mixture-of-experts (MoE). We then share the MoE layers across transformer blocks using individual layer normalization. Such deployment plays the role to transform various semantic representations, which makes the model more parameter-efficient and effective. To evaluate our framework, we design WideNet and evaluate it on ImageNet-1K. Our best model outperforms Vision Transformer (ViT) by $1.46\%$ with $0.72 \times$ trainable parameters. Using $0.46 \times$ and $0.13 \times$ parameters, our WideNet can still surpass ViT and ViT-MoE by $0.83\%$ and $2.08\%$, respectively.