Composed Image Retrieval (CIR) is the task of retrieving images matching a reference image augmented with a text, where the text describes changes to the reference image in natural language. Traditionally, models designed for CIR have relied on triplet data containing a reference image, reformulation text, and a target image. However, curating such triplet data often necessitates human intervention, leading to prohibitive costs. This challenge has hindered the scalability of CIR model training even with the availability of abundant unlabeled data. With the recent advances in foundational models, we advocate a shift in the CIR training paradigm where human annotations can be efficiently replaced by large language models (LLMs). Specifically, we demonstrate the capability of large captioning and language models in efficiently generating data for CIR only relying on unannotated image collections. Additionally, we introduce an embedding reformulation architecture that effectively combines image and text modalities. Our model, named InstructCIR, outperforms state-of-the-art methods in zero-shot composed image retrieval on CIRR and FashionIQ datasets. Furthermore, we demonstrate that by increasing the amount of generated data, our zero-shot model gets closer to the performance of supervised baselines.

Object Goal Navigation-requiring an agent to locate a specific object in an unseen environment-remains a core challenge in embodied AI. Although recent progress in Vision-Language Model (VLM)-based agents has demonstrated promising perception and decision-making abilities through prompting, none has yet established a fully modular world model design that reduces risky and costly interactions with the environment by predicting the future state of the world. We introduce WMNav, a novel World Model-based Navigation framework powered by Vision-Language Models (VLMs). It predicts possible outcomes of decisions and builds memories to provide feedback to the policy module. To retain the predicted state of the environment, WMNav proposes the online maintained Curiosity Value Map as part of the world model memory to provide dynamic configuration for navigation policy. By decomposing according to a human-like thinking process, WMNav effectively alleviates the impact of model hallucination by making decisions based on the feedback difference between the world model plan and observation. To further boost efficiency, we implement a two-stage action proposer strategy: broad exploration followed by precise localization. Extensive evaluation on HM3D and MP3D validates WMNav surpasses existing zero-shot benchmarks in both success rate and exploration efficiency (absolute improvement: +3.2% SR and +3.2% SPL on HM3D, +13.5% SR and +1.1% SPL on MP3D). Project page: https://b0b8k1ng.github.io/WMNav/.

This paper explores silent speech decoding in active brain-computer interface (BCI) systems, which offer more natural and flexible communication than traditional BCI applications. We collected a new silent speech dataset of over 120 hours of electroencephalogram (EEG) recordings from 12 subjects, capturing 24 commonly used English words for language model pretraining and decoding. Following the recent success of pretraining large models with self-supervised paradigms to enhance EEG classification performance, we propose Large Brain Language Model (LBLM) pretrained to decode silent speech for active BCI. To pretrain LBLM, we propose Future Spectro-Temporal Prediction (FSTP) pretraining paradigm to learn effective representations from unlabeled EEG data. Unlike existing EEG pretraining methods that mainly follow a masked-reconstruction paradigm, our proposed FSTP method employs autoregressive modeling in temporal and frequency domains to capture both temporal and spectral dependencies from EEG signals. After pretraining, we finetune our LBLM on downstream tasks, including word-level and semantic-level classification. Extensive experiments demonstrate significant performance gains of the LBLM over fully-supervised and pretrained baseline models. For instance, in the difficult cross-session setting, our model achieves 47.2% accuracy on semantic-level classification and 42.3% in word-level classification, outperforming baseline methods substantially. Our research advances silent speech decoding in active BCI systems, offering an innovative solution for EEG language model pretraining and a new dataset for fundamental research.

Accurate spatial reasoning in outdoor environments—covering geometry, object pose, and inter-object relationships—is fundamental to downstream tasks such as mapping, motion forecasting, and high-level planning in autonomous driving. We introduce SURDS, a large-scale benchmark designed to systematically evaluate the spatial reasoning capabilities of vision language models (VLMs). Built on the nuScenes dataset, SURDS comprises 41,080 vision–question–answer training instances and 9,250 evaluation samples, spanning six spatial categories: orientation, depth estimation, pixel-level localization, pairwise distance, lateral ordering, and front–behind relations. We benchmark leading general-purpose VLMs, including GPT, Gemini, and Qwen, revealing persistent limitations in fine-grained spatial understanding. To address these deficiencies, we go beyond static evaluation and explore whether alignment techniques can improve spatial reasoning performance. Specifically, we propose a reinforcement learning–based alignment scheme leveraging spatially grounded reward signals—capturing both perception-level accuracy (location) and reasoning consistency (logic). We further incorporate final-answer correctness and output-format rewards to guide fine-grained policy adaptation. Our GRPO-aligned variant achieves overall score of 40.80 in SURDS benchmark. Notably, it outperforms proprietary systems such as GPT-4o (13.30) and Gemini-2.0-flash (35.71). To our best knowledge, this is the first study to demonstrate that reinforcement learning–based alignment can significantly and consistently enhance the spatial reasoning capabilities of VLMs in real-world driving contexts. We release the SURDS benchmark, evaluation toolkit, and GRPO alignment code through: https://github.com/XiandaGuo/Drive-MLLM.

The utilization of Large Language Models (LLMs) within the realm of reinforcement learning, particularly as planners, has garnered a significant degree of attention in recent scholarly literature. However, a substantial proportion of existing research predominantly focuses on planning models for robotics that transmute the outputs derived from perception models into linguistic forms, thus adopting a ‘pure-language’ strategy. In this research, we propose a hybrid End-to-End learning framework for autonomous driving by combining basic driving imitation learning with LLMs based on multi-modality prompt tokens. Instead of simply converting perception results from the separated train model into pure language input, our novelty lies in two aspects. 1) The end-to-end integration of visual and LiDAR sensory input into learnable multi-modality tokens, thereby intrinsically alleviating description bias by separated pre-trained perception models. 2) Instead of directly letting LLMs drive, this paper explores a hybrid setting of letting LLMs help the driving model correct mistakes and complicated scenarios. The results of our experiments suggest that the proposed methodology can attain driving scores of 49.21%, coupled with an impressive route completion rate of 91.34% in the offline evaluation conducted via CARLA. These performance metrics are comparable to the most advanced driving models.

In recent years, reinforcement learning and imitation learning have shown great potential for controlling humanoid robots’ motion. However, these methods typically create simulation environments and rewards for specific tasks, resulting in the requirements of multiple policies and limited capabilities for tackling complex and unknown tasks. To overcome these issues, we present a novel approach that combines adversarial imitation learning with large language models (LLMs). This innovative method enables the agent to learn reusable skills with a single policy and solve zero-shot tasks under the guidance of LLMs. In particular, we utilize the LLM as a strategic planner for applying previously learned skills to novel tasks through the comprehension of task-specific prompts. This empowers the robot to perform the specified actions in a sequence. To improve our model, we incorporate codebook-based vector quantization, allowing the agent to generate suitable actions in response to unseen textual commands from LLMs. Furthermore, we design general reward functions that consider the distinct motion features of humanoid robots, ensuring the agent imitates the motion data while maintaining goal orientation without additional guiding direction approaches or policies. To the best of our knowledge, this is the first framework that controls humanoid robots using a single learning policy network and LLM as a planner. Extensive experiments demonstrate that our method exhibits efficient and adaptive ability in complicated motion tasks.

Driving under drowsy conditions significantly escalates the risk of vehicular accidents. Recent endeavors to prevent driving accidents have focused on using electroencephalography (EEG) to identify drowsy mental states. However, a more versatile EEG-based assistant system is better equipped to achieve smooth human-machine interaction in driving scenarios. This system should be capable of understanding a driver’s intention while demonstrating resilience to artifacts induced by unexpected sudden movements. Building on this goal, this paper pioneers a novel research direction in the field of brain-computer interface for assisted driving, studying the neural patterns associated with driving intentions and presenting a novel method for driving intention prediction. In particular, our preliminary analysis of the EEG signal using independent component analysis suggests a close relation between the intention of driving maneuvers and the neural activities in central-frontal and parietal areas. Power spectral density analysis at a group level also reveals a notable distinction among various driving intentions in the frequency domain. To exploit these brain dynamics, we propose a novel Masked EEG Modeling (MEM) framework for predicting human driving intentions, including the intention for left turning, right turning, and straight proceeding. Extensive experiments, encompassing comprehensive quantitative and qualitative assessments on a publicly available driving dataset, demonstrate the proposed MEM method is proficient in predicting driving intentions across various vigilance states. Specifically, our model attains an accuracy of 85.19% when predicting driving intentions for drowsy subjects, which shows its promising potential for mitigating traffic accidents related to drowsy driving. Ablation also demonstrates that our method significantly enhances the flexibility in handling missing channels. Notably, our method maintains over 75% accuracy when more than half of the channels are missing or corrupted, underscoring its adaptability in real-life driving scenarios.

Recently, humanoid robots have made significant advances in their ability to perform challenging tasks due to the deployment of Reinforcement Learning (RL), however, the inherent complexity of humanoid robots, including the difficulty of designing complicated reward functions and training entire sophisticated systems, still poses a notable challenge. To conquer these challenges, after many iterations and in-depth investigations, we have meticulously developed a full-size humanoid robot, "Adam", whose innovative structural design greatly improves the efficiency and effectiveness of the imitation learning process. In addition, we have developed a novel imitation learning framework based on an adversarial motion prior, which applies not only to Adam but also to humanoid robots in general. Using the framework, Adam can exhibit unprecedented human-like characteristics in locomotion tasks. Our experimental results demonstrate that the proposed framework enables Adam to achieve human-comparable performance in complex locomotion tasks, marking the first time that human locomotion data has been used for imitation learning in a full-size humanoid robot. For more video demonstrations, please visit our YouTube channel: https://www.youtube.com/watch?v=7hK2ySYBa1I

In the realm of digital forensics and document authentication, writer identification plays a crucial role in determining the authors of documents based on handwriting styles. The primary challenge in writer-id is the “open-set scenario”, where the goal is accurately recognizing writers unseen during the model training. To overcome this challenge, representation learning is the key. This method can capture unique handwriting features, enabling it to recognize styles not previously encountered during training. Building on this concept, this paper introduces the Contrastive Masked Auto-Encoders (CMAE) for Character-level Open-Set Writer Identification. We merge Masked Auto-Encoders (MAE) with Contrastive Learning (CL) to simultaneously and respectively capture sequential information and distinguish diverse handwriting styles. Demonstrating its effectiveness, our model achieves state-of-the-art (SOTA) results on the CASIA online handwriting dataset, reaching an impressive precision rate of 89.7%. Our study advances universal writer-id with a sophisticated representation learning approach, contributing substantially to the ever-evolving landscape of digital handwriting analysis, and catering to the demands of an increasingly interconnected world.

The translation of brain dynamics into natural language is pivotal for brain-computer interfaces (BCIs), a field that has seen substantial growth in recent years. With the swift advancement of large language models, such as ChatGPT, the need to bridge the gap between the brain and languages becomes increasingly pressing. Current methods, however, require eye-tracking fixations or event markers to segment brain dynamics into word-level features, which can restrict the practical application of these systems. These event markers may not be readily available or could be challenging to acquire during real-time inference, and the sequence of eye fixations may not align with the order of spoken words. To tackle these issues, we introduce a novel framework, DeWave, that integrates discrete encoding sequences into open-vocabulary EEG-to-text translation tasks. DeWave uses a quantized variational encoder to derive discrete codex encoding and align it with pre-trained language models. This discrete codex representation brings forth two advantages: 1) it alleviates the order mismatch between eye fixations and spoken words by introducing text-EEG contrastive alignment training, and 2) it minimizes the interference caused by individual differences in EEG waves through an invariant discrete codex. Our model surpasses the previous baseline (40.1 and 31.7) by 3.06% and 6.34\%, respectively, achieving 41.35 BLEU-1 and 33.71 Rouge-F on the ZuCo Dataset. Furthermore, this work is the first to facilitate the translation of entire EEG signal periods without the need for word-level order markers (e.g., eye fixations), scoring 20.5 BLEU-1 and 29.5 Rouge-1 on the ZuCo Dataset, respectively.

Streaming label learning aims to model newly emerged labels for multilabel classification systems, which requires plenty of new label data for training. However, in changing environments, only a small amount of new label data can practically be collected. In this work, we formulate and study few-shot streaming label learning (FSLL), which models emerging new labels with only a few annotated examples by utilizing the knowledge learned from past labels. We propose a meta-learning framework, semantic inference network (SIN), which can learn and infer the semantic correlation between new labels and past labels to adapt FSLL tasks from a few examples effectively. SIN leverages label semantic representation to regularize the output space and acquires labelwise meta-knowledge based on gradient-based meta-learning. Moreover, SIN incorporates a novel label decision module with a meta-threshold loss to find the optimal confidence thresholds for each new label. Theoretically, we illustrate that the proposed semantic inference mechanism could constrain the complexity of hypotheses space to reduce the risk of overfitting and achieve better generalizability. Experimentally, extensive empirical results and ablation studies demonstrate the performance of SIN is superior to the prior state-of-the-art methods on FSLL.

Continual learning is an intellectual ability of artificial agents to learn new streaming labels from sequential data. The main impediment to continual learning is catastrophic forgetting, a severe performance degradation on previously learned tasks. Although simply replaying all previous data or continuously adding the model parameters could alleviate the issue, it is impractical in real-world applications due to the limited available resources. Inspired by the mechanism of the human brain to deepen its past impression, we propose a novel framework, Deep Retrieval and Imagination (DRI), which consists of two components: 1) an embedding network that constructs a unified embedding space without adding model parameters on the arrival of new tasks; and 2) a generative model to produce additional (imaginary) data based on the limited memory. By retrieving the past experiences and corresponding imaginary data, DRI distills knowledge and rebalances the embedding space to further mitigate forgetting. Theoretical analysis demonstrates that DRI can reduce the loss approximation error and improve the robustness through retrieval and imagination, bringing better generalizability to the network. Extensive experiments show that DRI performs significantly better than the existing state-of-the-art continual learning methods and effectively alleviates catastrophic forgetting.

Weakly-Supervised Semantic Segmentation (WSSS) segments objects without heavy burden of dense annotation. While as a price, generated pseudo-masks exist obvious noisy pixels, which result in sub-optimal segmentation models trained over these pseudo-masks. But rare studies notice or work on this problem, even these noisy pixels are inevitable after their improvements on pseudo-mask. So we try to improve WSSS in the aspect of noise mitigation. And we observe that many noisy pixels are of high confidences, especially when the response range is too wide or narrow, presenting an uncertain status. Thus, in this paper, we simulate noisy variations of response by scaling the prediction map in multiple times for uncertainty estimation. The uncertainty is then used to weight the segmentation loss to mitigate noisy supervision signals. We call this method URN, abbreviated from Uncertainty estimation via Response scaling for Noise mitigation. Experiments validate the benefits of URN, and our method achieves state-of-the-art results at 71.2% and 41.5% on PASCAL VOC 2012 and MS COCO 2014 respectively, without extra models like saliency detection. Code is available at https://github.com/XMed-Lab/URN.

Continual learning methods aim at training a neural network from sequential data with streaming labels, relieving catastrophic forgetting. However, existing methods are based on and designed for convolutional neural networks (CNNs), which have not utilized the full potential of newly emerged powerful vision transformers. In this paper, we propose a novel attention-based framework Lifelong Vision Transformer (LVT), to achieve a better stability-plasticity trade-off for continual learning. Specifically, an inter-task attention mechanism is presented in LVT, which implicitly absorbs the previous tasks' information and slows down the drift of important attention between previous tasks and the current task. LVT designs a dual-classifier structure that independently injects new representation to avoid catas-trophic interference and accumulates the new and previous knowledge in a balanced manner to improve the overall performance. Moreover, we develop a confidence-aware memory update strategy to deepen the impression of the previous tasks. The extensive experimental results show that our approach achieves state-of-the-art performance with even fewer parameters on continual learning benchmarks.

Few-shot learning can adapt the classification model to new labels with only a few labeled examples. Previous studies mainly focus on the scenario of a single category label per example but have not solved the more challenging multi-label scenario with exponential-sized output space and low-data effectively. In this paper, we propose a semantic-aware meta-learning model for multi-label few-shot learning. Our approach can learn and infer the semantic correlation between unseen labels and historical labels to quickly adapt multi-label tasks from only a few examples. Specifically, features can be mapped into the semantic embedding space via label word vectors to explore and exploit the label correlation, and thus cope with the challenge on the overwhelming size of the output space. Then a novel semantic inference mechanism is designed for leveraging prior knowledge learned from historical labels, which will produce good generalization performance on new labels to alleviate the low-data problem. Finally, extensive empirical results show that the proposed method significantly outperforms the existing state-of-the-art methods on the multi-label few-shot learning tasks.

It is desirable to include more controllable attributes to enhance the diversity of generated responses in open-domain dialogue systems. However, existing methods can generate responses with only one controllable attribute or lack a flexible way to generate them with multiple controllable attributes. In this paper, we propose a Progressively trained Hierarchical Encoder-Decoder (PHED) to tackle this task. More specifically, PHED deploys Conditional Variational AutoEncoder (CVAE) on Transformer to include one aspect of attributes at one stage. A vital characteristic of the CVAE is to separate the latent variables at each stage into two types: a global variable capturing the common semantic features and a specific variable absorbing the attribute information at that stage. PHED then couples the CVAE latent variables with the Transformer encoder and is trained by minimizing a newly derived ELBO and controlled losses to produce the next stage's input and produce responses as required. Finally, we conduct extensive evaluations to show that PHED significantly outperforms the state-of-the-art neural generation models and produces more diverse responses as expected.