Post-Semantic-Thinking: A Robust Strategy to Distill Reasoning Capacity from Large Language Models

Chain of thought finetuning aims to endow small student models with reasoning capacity to improve their performance towards a specific task by allowing them to imitate the reasoning procedure of large language models (LLMs) beyond simply predicting the answer to the question. However, the existing methods 1) generate rationale before the answer, making their answer correctness sensitive to the hallucination in the rationale;2) force the student model to repeat the exact LLMs rationale expression word-after-word, which could have the model biased towards learning the expression in rationale but count against the model from understanding the core logic behind it. Therefore, we propose a robust Post-Semantic-Thinking (PST) strategy to generate answers before rationale. Thanks to this answer-first setting, 1) the answering procedure can escape from the adverse effects caused by hallucinations in the rationale; 2) the complex reasoning procedure is tightly bound with the relatively concise answer, making the reasoning for questions easier with the prior information in the answer; 3) the efficiency of the method can also benefit from the setting since users can stop the generation right after answers are outputted when inference is conducted. Furthermore, the PST strategy loose the constraint against the generated rationale to be close to the LLMs gold standard in the hidden semantic space instead of the vocabulary space, thus making the small student model better comprehend the semantic reasoning logic in rationale. Extensive experiments conducted across 12 reasoning tasks demonstrate the effectiveness of PST.

Federated Data Model

In artificial intelligence (AI), especially deep learning, data diversity and volume play a pivotal role in model development. However, training a robust deep learning model often faces challenges due to data privacy, regulations, and the difficulty of sharing data between different locations, especially for medical applications. To address this, we developed a method called the Federated Data Model (FDM). This method uses diffusion models to learn the characteristics of data at one site and then creates synthetic data that can be used at another site without sharing the actual data. We tested this approach with a medical image segmentation task, focusing on cardiac magnetic resonance images from different hospitals. Our results show that models trained with this method perform well both on the data they were originally trained on and on data from other sites. This approach offers a promising way to train accurate and privacy-respecting AI models across different locations.

Spatiotemporal Diffusion Model with Paired Sampling for Accelerated Cardiac Cine MRI

Current deep learning reconstruction for accelerated cardiac cine MRI suffers from spatial and temporal blurring. We aim to improve image sharpness and motion delineation for cine MRI under high undersampling rates. A spatiotemporal diffusion enhancement model conditional on an existing deep learning reconstruction along with a novel paired sampling strategy was developed. The diffusion model provided sharper tissue boundaries and clearer motion than the original reconstruction in experts evaluation on clinical data. The innovative paired sampling strategy substantially reduced artificial noises in the generative results.

Clinically Feasible Diffusion Reconstruction for Highly-Accelerated Cardiac Cine MRI

The currently limited quality of accelerated cardiac cine reconstruction may potentially be improved by the emerging diffusion models, but the clinically unacceptable long processing time poses a challenge. We aim to develop a clinically feasible diffusion-model-based reconstruction pipeline to improve the image quality of cine MRI. A multi-in multi-out diffusion enhancement model together with fast inference strategies were developed to be used in conjunction with a reconstruction model. The diffusion reconstruction reduced spatial and temporal blurring in prospectively undersampled clinical data, as validated by experts inspection. The 1.5s per video processing time enabled the approach to be applied in clinical scenarios.

GenNBV: Generalizable Next-Best-View Policy for Active 3D Reconstruction

While recent advances in neural radiance field enable realistic digitization for large-scale scenes, the image-capturing process is still time-consuming and labor-intensive. Previous works attempt to automate this process using the Next-Best-View (NBV) policy for active 3D reconstruction. However, the existing NBV policies heavily rely on hand-crafted criteria, limited action space, or per-scene optimized representations. These constraints limit their cross-dataset generalizability. To overcome them, we propose GenNBV, an end-to-end generalizable NBV policy. Our policy adopts a reinforcement learning (RL)-based framework and extends typical limited action space to 5D free space. It empowers our agent drone to scan from any viewpoint, and even interact with unseen geometries during training. To boost the cross-dataset generalizability, we also propose a novel multi-source state embedding, including geometric, semantic, and action representations. We establish a benchmark using the Isaac Gym simulator with the Houses3K and OmniObject3D datasets to evaluate this NBV policy. Experiments demonstrate that our policy achieves a 98.26% and 97.12% coverage ratio on unseen building-scale objects from these datasets, respectively, outperforming prior solutions.

Triplet-constraint Transformer with Multi-scale Refinement for Dose Prediction in Radiotherapy

Radiotherapy is a primary treatment for cancers with the aim of applying sufficient radiation dose to the planning target volume (PTV) while minimizing dose hazards to the organs at risk (OARs). Convolutional neural networks (CNNs) have automated the radiotherapy plan-making by predicting the dose maps. However, current CNN-based methods ignore the remarkable dose difference in the dose map, i.e., high dose value in the interior PTV while low value in the exterior PTV, leading to a suboptimal prediction. In this paper, we propose a triplet-constraint transformer (TCtrans) with multi-scale refinement to predict the high-quality dose distribution. Concretely, a novel PTV-guided triplet constraint is designed to refine dose feature representations in the interior and exterior PTV by utilizing the explicit geometry of PTV. Furthermore, we introduce a multi-scale refinement (MSR) module to effectively fulfill the triplet constraint in different decoding layers with multiple scales. Besides, a transformer encoder is devised to learn the important global dosimetric knowledge. Experiments on a clinical cervical cancer dataset demonstrate the superiority of our method.

EmbodiedScan: A Holistic Multi-Modal 3D Perception Suite Towards Embodied AI

In the realm of computer vision and robotics, embodied agents are expected to explore their environment and carry out human instructions. This necessitates the ability to fully understand 3D scenes given their first-person observations and contextualize them into language for interaction. However, traditional research focuses more on scene-level input and output setups from a global view. To address the gap, we introduce EmbodiedScan, a multi-modal, ego-centric 3D perception dataset and benchmark for holistic 3D scene understanding. It encompasses over 5k scans encapsulating 1M ego-centric RGB-D views, 1M language prompts, 160k 3D-oriented boxes spanning over 760 categories, some of which partially align with LVIS, and dense semantic occupancy with 80 common categories. Building upon this database, we introduce a baseline framework named Embodied Perceptron. It is capable of processing an arbitrary number of multi-modal inputs and demonstrates remarkable 3D perception capabilities, both within the two series of benchmarks we set up, i.e., fundamental 3D perception tasks and language-grounded tasks, and in the wild. Codes, datasets, and benchmarks will be available at https://github.com/OpenRobotLab/EmbodiedScan.

Spectroscopy-Guided Discovery of Three-Dimensional Structures of Disordered Materials with Diffusion Models

The ability to rapidly develop materials with desired properties has a transformative impact on a broad range of emerging technologies. In this work, we introduce a new framework based on the diffusion model, a recent generative machine learning method to predict 3D structures of disordered materials from a target property. For demonstration, we apply the model to identify the atomic structures of amorphous carbons ($a$-C) as a representative material system from the target X-ray absorption near edge structure (XANES) spectra--a common experimental technique to probe atomic structures of materials. We show that conditional generation guided by XANES spectra reproduces key features of the target structures. Furthermore, we show that our model can steer the generative process to tailor atomic arrangements for a specific XANES spectrum. Finally, our generative model exhibits a remarkable scale-agnostic property, thereby enabling generation of realistic, large-scale structures through learning from a small-scale dataset (i.e., with small unit cells). Our work represents a significant stride in bridging the gap between materials characterization and atomic structure determination; in addition, it can be leveraged for materials discovery in exploring various material properties as targeted.

Improving End-to-End Speech Processing by Efficient Text Data Utilization with Latent Synthesis

Training a high performance end-to-end speech (E2E) processing model requires an enormous amount of labeled speech data, especially in the era of data-centric artificial intelligence. However, labeled speech data are usually scarcer and more expensive for collection, compared to textual data. We propose Latent Synthesis (LaSyn), an efficient textual data utilization framework for E2E speech processing models. We train a latent synthesizer to convert textual data into an intermediate latent representation of a pre-trained speech model. These pseudo acoustic representations of textual data augment acoustic data for model training. We evaluate LaSyn on low-resource automatic speech recognition (ASR) and spoken language understanding (SLU) tasks. For ASR, LaSyn improves an E2E baseline trained on LibriSpeech train-clean-100, with relative word error rate reductions over 22.3% on different test sets. For SLU, LaSyn improves our E2E baseline by absolute 4.1% for intent classification accuracy and 3.8% for slot filling SLU-F1 on SLURP, and absolute 4.49% and 2.25% for exact match (EM) and EM-Tree accuracies on STOP respectively. With fewer parameters, the results of LaSyn are competitive to published state-of-the-art works. The results demonstrate the quality of the augmented training data.