Catastrophic forgetting poses a substantial challenge for managing intelligent agents controlled by a large model, causing performance degradation when these agents face new tasks. In our work, we propose a novel solution - the Progressive Prompt Decision Transformer (P2DT). This method enhances a transformer-based model by dynamically appending decision tokens during new task training, thus fostering task-specific policies. Our approach mitigates forgetting in continual and offline reinforcement learning scenarios. Moreover, P2DT leverages trajectories collected via traditional reinforcement learning from all tasks and generates new task-specific tokens during training, thereby retaining knowledge from previous studies. Preliminary results demonstrate that our model effectively alleviates catastrophic forgetting and scales well with increasing task environments.
相關內容
Federated Learning (FL) trains a black-box and high-dimensional model among different clients by exchanging parameters instead of direct data sharing, which mitigates the privacy leak incurred by machine learning. However, FL still suffers from membership inference attacks (MIA) or data reconstruction attacks (DRA). In particular, an attacker can extract the information from local datasets by constructing DRA, which cannot be effectively throttled by existing techniques, e.g., Differential Privacy (DP). In this paper, we aim to ensure a strong privacy guarantee for FL under DRA. We prove that reconstruction errors under DRA are constrained by the information acquired by an attacker, which means that constraining the transmitted information can effectively throttle DRA. To quantify the information leakage incurred by FL, we establish a channel model, which depends on the upper bound of joint mutual information between the local dataset and multiple transmitted parameters. Moreover, the channel model indicates that the transmitted information can be constrained through data space operation, which can improve training efficiency and the model accuracy under constrained information. According to the channel model, we propose algorithms to constrain the information transmitted in a single round of local training. With a limited number of training rounds, the algorithms ensure that the total amount of transmitted information is limited. Furthermore, our channel model can be applied to various privacy-enhancing techniques (such as DP) to enhance privacy guarantees against DRA. Extensive experiments with real-world datasets validate the effectiveness of our methods.
Production Machine Learning involves continuous training: hosting multiple versions of models over time, often with many model versions running at once. When model performance does not meet expectations, Machine Learning Engineers (MLEs) debug issues by exploring and analyzing numerous prior versions of code and training data to identify root causes and mitigate problems. Traditional debugging and logging tools often fall short in managing this experimental, multi-version context. FlorDB introduces Multiversion Hindsight Logging, which allows engineers to use the most recent version's logging statements to query past versions, even when older versions logged different data. Log statement propagation enables consistent injection of logging statements into past code versions, regardless of changes to the codebase. Once log statements are propagated across code versions, the remaining challenge in Multiversion Hindsight Logging is to efficiently replay the new log statements based on checkpoints from previous runs. Finally, a coherent user experience is required to help MLEs debug across all versions of code and data. To this end, FlorDB presents a unified relational model for efficient handling of historical queries, offering a comprehensive view of the log history to simplify the exploration of past code iterations. We present a performance evaluation on diverse benchmarks confirming its scalability and the ability to deliver real-time query responses, leveraging query-based filtering and checkpoint-based parallelism for efficient replay.
With excellent generalization ability, self-supervised speech models have shown impressive performance on various downstream speech tasks in the pre-training and fine-tuning paradigm. However, as the growing size of pre-trained models, fine-tuning becomes practically unfeasible due to heavy computation and storage overhead, as well as the risk of overfitting. Adapters are lightweight modules inserted into pre-trained models to facilitate parameter-efficient adaptation. In this paper, we propose an effective adapter framework designed for adapting self-supervised speech models to the speaker verification task. With a parallel adapter design, our proposed framework inserts two types of adapters into the pre-trained model, allowing the adaptation of latent features within intermediate Transformer layers and output embeddings from all Transformer layers. We conduct comprehensive experiments to validate the efficiency and effectiveness of the proposed framework. Experimental results on the VoxCeleb1 dataset demonstrate that the proposed adapters surpass fine-tuning and other parameter-efficient transfer learning methods, achieving superior performance while updating only 5% of the parameters.
Inverted landing is a routine behavior among a number of animal fliers. However, mastering this feat poses a considerable challenge for robotic fliers, especially to perform dynamic perching with rapid body rotations (or flips) and landing against gravity. Inverted landing in flies have suggested that optical flow senses are closely linked to the precise triggering and control of body flips that lead to a variety of successful landing behaviors. Building upon this knowledge, we aimed to replicate the flies' landing behaviors in small quadcopters by developing a control policy general to arbitrary ceiling-approach conditions. First, we employed reinforcement learning in simulation to optimize discrete sensory-motor pairs across a broad spectrum of ceiling-approach velocities and directions. Next, we converted the sensory-motor pairs to a two-stage control policy in a continuous augmented-optical flow space. The control policy consists of a first-stage Flip-Trigger Policy, which employs a one-class support vector machine, and a second-stage Flip-Action Policy, implemented as a feed-forward neural network. To transfer the inverted-landing policy to physical systems, we utilized domain randomization and system identification techniques for a zero-shot sim-to-real transfer. As a result, we successfully achieved a range of robust inverted-landing behaviors in small quadcopters, emulating those observed in flies.
Hallucination, posed as a pervasive challenge of multi-modal large language models (MLLMs), has significantly impeded their real-world usage that demands precise judgment. Existing methods mitigate this issue with either training with specific designed data or inferencing with external knowledge from other sources, incurring inevitable additional costs. In this paper, we present OPERA, a novel MLLM decoding method grounded in an Over-trust Penalty and a Retrospection-Allocation strategy, serving as a nearly free lunch to alleviate the hallucination issue without additional data, knowledge, or training. Our approach begins with an interesting observation that, most hallucinations are closely tied to the knowledge aggregation patterns manifested in the self-attention matrix, i.e., MLLMs tend to generate new tokens by focusing on a few summary tokens, but not all the previous tokens. Such partial over-trust inclination results in the neglecting of image tokens and describes the image content with hallucination. Based on the observation, OPERA introduces a penalty term on the model logits during the beam-search decoding to mitigate the over-trust issue, along with a rollback strategy that retrospects the presence of summary tokens in the previously generated tokens, and re-allocate the token selection if necessary. With extensive experiments, OPERA shows significant hallucination-mitigating performance on different MLLMs and metrics, proving its effectiveness and generality. Our code is available at: //github.com/shikiw/OPERA.
In aerial combat, dogfighting poses intricate challenges that demand an understanding of both strategic maneuvers and the aerodynamics of agile fighter aircraft. In this paper, we introduce TempFuser, a novel long short-term temporal fusion transformer designed to learn tactical and agile flight maneuvers in aerial dogfights. Our approach employs two distinct LSTM-based input embeddings to encode long-term sparse and short-term dense state representations. By integrating these embeddings through a transformer encoder, our model captures the tactics and agility of fighter jets, enabling it to generate end-to-end flight commands that secure dominant positions and outmaneuver the opponent. After extensive training against various types of opponent aircraft in a high-fidelity flight simulator, our model successfully learns to perform complex fighter maneuvers, consistently outperforming several baseline models. Notably, our model exhibits human-like strategic maneuvers even when facing adversaries with superior specifications, all without relying on explicit prior knowledge. Moreover, it demonstrates robust pursuit performance in challenging supersonic and low-altitude environments. Demo videos are available at //sites.google.com/view/tempfuser.
With the urgent demand for generalized deep models, many pre-trained big models are proposed, such as BERT, ViT, GPT, etc. Inspired by the success of these models in single domains (like computer vision and natural language processing), the multi-modal pre-trained big models have also drawn more and more attention in recent years. In this work, we give a comprehensive survey of these models and hope this paper could provide new insights and helps fresh researchers to track the most cutting-edge works. Specifically, we firstly introduce the background of multi-modal pre-training by reviewing the conventional deep learning, pre-training works in natural language process, computer vision, and speech. Then, we introduce the task definition, key challenges, and advantages of multi-modal pre-training models (MM-PTMs), and discuss the MM-PTMs with a focus on data, objectives, network architectures, and knowledge enhanced pre-training. After that, we introduce the downstream tasks used for the validation of large-scale MM-PTMs, including generative, classification, and regression tasks. We also give visualization and analysis of the model parameters and results on representative downstream tasks. Finally, we point out possible research directions for this topic that may benefit future works. In addition, we maintain a continuously updated paper list for large-scale pre-trained multi-modal big models: //github.com/wangxiao5791509/MultiModal_BigModels_Survey
Multi-agent influence diagrams (MAIDs) are a popular form of graphical model that, for certain classes of games, have been shown to offer key complexity and explainability advantages over traditional extensive form game (EFG) representations. In this paper, we extend previous work on MAIDs by introducing the concept of a MAID subgame, as well as subgame perfect and trembling hand perfect equilibrium refinements. We then prove several equivalence results between MAIDs and EFGs. Finally, we describe an open source implementation for reasoning about MAIDs and computing their equilibria.
Recently pre-trained language representation models such as BERT have shown great success when fine-tuned on downstream tasks including information retrieval (IR). However, pre-training objectives tailored for ad-hoc retrieval have not been well explored. In this paper, we propose Pre-training with Representative wOrds Prediction (PROP) for ad-hoc retrieval. PROP is inspired by the classical statistical language model for IR, specifically the query likelihood model, which assumes that the query is generated as the piece of text representative of the "ideal" document. Based on this idea, we construct the representative words prediction (ROP) task for pre-training. Given an input document, we sample a pair of word sets according to the document language model, where the set with higher likelihood is deemed as more representative of the document. We then pre-train the Transformer model to predict the pairwise preference between the two word sets, jointly with the Masked Language Model (MLM) objective. By further fine-tuning on a variety of representative downstream ad-hoc retrieval tasks, PROP achieves significant improvements over baselines without pre-training or with other pre-training methods. We also show that PROP can achieve exciting performance under both the zero- and low-resource IR settings. The code and pre-trained models are available at //github.com/Albert-Ma/PROP.
Automatic KB completion for commonsense knowledge graphs (e.g., ATOMIC and ConceptNet) poses unique challenges compared to the much studied conventional knowledge bases (e.g., Freebase). Commonsense knowledge graphs use free-form text to represent nodes, resulting in orders of magnitude more nodes compared to conventional KBs (18x more nodes in ATOMIC compared to Freebase (FB15K-237)). Importantly, this implies significantly sparser graph structures - a major challenge for existing KB completion methods that assume densely connected graphs over a relatively smaller set of nodes. In this paper, we present novel KB completion models that can address these challenges by exploiting the structural and semantic context of nodes. Specifically, we investigate two key ideas: (1) learning from local graph structure, using graph convolutional networks and automatic graph densification and (2) transfer learning from pre-trained language models to knowledge graphs for enhanced contextual representation of knowledge. We describe our method to incorporate information from both these sources in a joint model and provide the first empirical results for KB completion on ATOMIC and evaluation with ranking metrics on ConceptNet. Our results demonstrate the effectiveness of language model representations in boosting link prediction performance and the advantages of learning from local graph structure (+1.5 points in MRR for ConceptNet) when training on subgraphs for computational efficiency. Further analysis on model predictions shines light on the types of commonsense knowledge that language models capture well.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191