Sigma: Differential Rescaling of Query, Key and Value for Efficient Language Models

We introduce Sigma, an efficient large language model specialized for the system domain, empowered by a novel architecture including DiffQKV attention, and pre-trained on our meticulously collected system domain data. DiffQKV attention significantly enhances the inference efficiency of Sigma by optimizing the Query (Q), Key (K), and Value (V) components in the attention mechanism differentially, based on their varying impacts on the model performance and efficiency indicators. Specifically, we (1) conduct extensive experiments that demonstrate the model's varying sensitivity to the compression of K and V components, leading to the development of differentially compressed KV, and (2) propose augmented Q to expand the Q head dimension, which enhances the model's representation capacity with minimal impacts on the inference speed. Rigorous theoretical and empirical analyses reveal that DiffQKV attention significantly enhances efficiency, achieving up to a 33.36% improvement in inference speed over the conventional grouped-query attention (GQA) in long-context scenarios. We pre-train Sigma on 6T tokens from various sources, including 19.5B system domain data that we carefully collect and 1T tokens of synthesized and rewritten data. In general domains, Sigma achieves comparable performance to other state-of-arts models. In the system domain, we introduce the first comprehensive benchmark AIMicius, where Sigma demonstrates remarkable performance across all tasks, significantly outperforming GPT-4 with an absolute improvement up to 52.5%.

Chain-of-Reasoning: Towards Unified Mathematical Reasoning in Large Language Models via a Multi-Paradigm Perspective

Large Language Models (LLMs) have made notable progress in mathematical reasoning, yet they often rely on single-paradigm reasoning that limits their effectiveness across diverse tasks. In this paper, we introduce Chain-of-Reasoning (CoR), a novel unified framework that integrates multiple reasoning paradigms--Natural Language Reasoning (NLR), Algorithmic Reasoning (AR), and Symbolic Reasoning (SR)--to enable synergistic collaboration. CoR generates multiple potential answers using different reasoning paradigms and synthesizes them into a coherent final solution. We propose a Progressive Paradigm Training (PPT) strategy that allows models to progressively master these paradigms, culminating in the development of CoR-Math-7B. Experimental results demonstrate that CoR-Math-7B significantly outperforms current SOTA models, achieving up to a 41.0% absolute improvement over GPT-4 in theorem proving tasks and a 7.9% improvement over RL-based methods in arithmetic tasks. These results showcase the enhanced mathematical comprehensive ability of our model, achieving significant performance gains on specific tasks and enabling zero-shot generalization across tasks.

Scalable Vision Language Model Training via High Quality Data Curation

In this paper, we introduce SAIL-VL (ScAlable Vision Language Model TraIning via High QuaLity Data Curation), an open-source vision language model (VLM) of state-of-the-art (SOTA) performance with 2B parameters. We introduce three key improvements that contribute to SAIL-VL's leading performance: (1) Scalable high-quality visual understanding data construction: We implement a visual understanding data construction pipeline, which enables hundred-million-scale high-quality recaption data annotation. Equipped with this pipeline, we curate SAIL-Caption, a large-scale caption dataset with large quantity and the highest data quality compared with opensource caption datasets. (2) Scalable Pretraining with High-Quality Visual Understanding Data: We scale SAIL-VL's pretraining budget up to 131B tokens and show that even a 2B VLM benefits from scaled up training data sizes, exhibiting expected data size scaling laws in visual understanding and instruction following performance. (3) Scalable SFT via quantity and quality scaling: We introduce general guidance for instruction data curation to scale up instruction data continuously, allowing us to construct a large SFT dataset with the highest quality. To further improve SAIL-VL's performance, we propose quality scaling, a multi-stage training recipe with curriculum learning, to improve model performance scaling curves w.r.t. data sizes from logarithmic to be near-linear. SAIL-VL obtains the highest average score in 19 commonly used benchmarks in our evaluation and achieves top1 performance among VLMs of comparable sizes on OpenCompass (https://rank.opencompass.org.cn/leaderboard-multimodal). We release our SAIL-VL-2B model at HuggingFace (https://huggingface.co/BytedanceDouyinContent/SAIL-VL-2B).

Integrative Decoding: Improve Factuality via Implicit Self-consistency

Self-consistency-based approaches, which involve repeatedly sampling multiple outputs and selecting the most consistent one as the final response, prove to be remarkably effective in improving the factual accuracy of large language models. Nonetheless, existing methods usually have strict constraints on the task format, largely limiting their applicability. In this paper, we present Integrative Decoding (ID), to unlock the potential of self-consistency in open-ended generation tasks. ID operates by constructing a set of inputs, each prepended with a previously sampled response, and then processes them concurrently, with the next token being selected by aggregating of all their corresponding predictions at each decoding step. In essence, this simple approach implicitly incorporates self-consistency in the decoding objective. Extensive evaluation shows that ID consistently enhances factuality over a wide range of language models, with substantial improvements on the TruthfulQA (+11.2%), Biographies (+15.4%) and LongFact (+8.5%) benchmarks. The performance gains amplify progressively as the number of sampled responses increases, indicating the potential of ID to scale up with repeated sampling.

MEDiC: Autonomous Surgical Robotic Assistance to Maximizing Exposure for Dissection and Cautery

Surgical automation has the capability to improve the consistency of patient outcomes and broaden access to advanced surgical care in underprivileged communities. Shared autonomy, where the robot automates routine subtasks while the surgeon retains partial teleoperative control, offers great potential to make an impact. In this paper we focus on one important skill within surgical shared autonomy: Automating robotic assistance to maximize visual exposure and apply tissue tension for dissection and cautery. Ensuring consistent exposure to visualize the surgical site is crucial for both efficiency and patient safety. However, achieving this is highly challenging due to the complexities of manipulating deformable volumetric tissues that are prevalent in surgery.To address these challenges we propose \methodname, a framework for autonomous surgical robotic assistance to \methodfullname. We integrate a differentiable physics model with perceptual feedback to achieve our two key objectives: 1) Maximizing tissue exposure and applying tension for a specified dissection site through visual-servoing conrol and 2) Selecting optimal control positions for a dissection target based on deformable Jacobian analysis. We quantitatively assess our method through repeated real robot experiments on a tissue phantom, and showcase its capabilities through dissection experiments using shared autonomy on real animal tissue.

AutoPeel: Adhesion-aware Safe Peeling Trajectory Optimization for Robotic Wound Care

Chronic wounds, including diabetic ulcers, pressure ulcers, and ulcers secondary to venous hypertension, affects more than 6.5 million patients and a yearly cost of more than $25 billion in the United States alone. Chronic wound treatment is currently a manual process, and we envision a future where robotics and automation will aid in this treatment to reduce cost and improve patient care. In this work, we present the development of the first robotic system for wound dressing removal which is reported to be the worst aspect of living with chronic wounds. Our method leverages differentiable physics-based simulation to perform gradient-based Model Predictive Control (MPC) for optimized trajectory planning. By integrating fracture mechanics of adhesion, we are able to model the peeling effect inherent to dressing adhesion. The system is further guided by carefully designed objective functions that promote both efficient and safe control, reducing the risk of tissue damage. We validated the efficacy of our approach through a series of experiments conducted on both synthetic skin phantoms and real human subjects. Our results demonstrate the system's ability to achieve precise and safe dressing removal trajectories, offering a promising solution for automating this essential healthcare procedure.

SIMPNet: Spatial-Informed Motion Planning Network

Current robotic manipulators require fast and efficient motion-planning algorithms to operate in cluttered environments. State-of-the-art sampling-based motion planners struggle to scale to high-dimensional configuration spaces and are inefficient in complex environments. This inefficiency arises because these planners utilize either uniform or hand-crafted sampling heuristics within the configuration space. To address these challenges, we present the Spatial-informed Motion Planning Network (SIMPNet). SIMPNet consists of a stochastic graph neural network (GNN)-based sampling heuristic for informed sampling within the configuration space. The sampling heuristic of SIMPNet encodes the workspace embedding into the configuration space through a cross-attention mechanism. It encodes the manipulator's kinematic structure into a graph, which is used to generate informed samples within the framework of sampling-based motion planning algorithms. We have evaluated the performance of SIMPNet using a UR5e robotic manipulator operating within simple and complex workspaces, comparing it against baseline state-of-the-art motion planners. The evaluation results show the effectiveness and advantages of the proposed planner compared to the baseline planners.

Grasping Force Control and Adaptation for a Cable-Driven Robotic Hand

This paper introduces a unique force control and adaptation algorithm for a lightweight and low-complexity five-fingered robotic hand, namely an Integrated-Finger Robotic Hand (IFRH). The force control and adaptation algorithm is intuitive to design, easy to implement, and improves the grasping functionality through feedforward adaptation automatically. Specifically, we have extended Youla-parameterization which is traditionally used in feedback controller design into a feedforward iterative learning control algorithm (ILC). The uniqueness of such an extension is that both the feedback and feedforward controllers are parameterized over one unified design parameter which can be easily customized based on the desired closed-loop performance. While Youla-parameterization and ILC have been explored in the past on various applications, our unique parameterization and computational methods make the design intuitive and easy to implement. This provides both robust and adaptive learning capabilities, and our application rivals the complexity of many robotic hand control systems. Extensive experimental tests have been conducted to validate the effectiveness of our method.

PianoBART: Symbolic Piano Music Generation and Understanding with Large-Scale Pre-Training

Learning musical structures and composition patterns is necessary for both music generation and understanding, but current methods do not make uniform use of learned features to generate and comprehend music simultaneously. In this paper, we propose PianoBART, a pre-trained model that uses BART for both symbolic piano music generation and understanding. We devise a multi-level object selection strategy for different pre-training tasks of PianoBART, which can prevent information leakage or loss and enhance learning ability. The musical semantics captured in pre-training are fine-tuned for music generation and understanding tasks. Experiments demonstrate that PianoBART efficiently learns musical patterns and achieves outstanding performance in generating high-quality coherent pieces and comprehending music. Our code and supplementary material are available at https://github.com/RS2002/PianoBart.

Task Oriented In-Domain Data Augmentation

Large Language Models (LLMs) have shown superior performance in various applications and fields. To achieve better performance on specialized domains such as law and advertisement, LLMs are often continue pre-trained on in-domain data. However, existing approaches suffer from two major issues. First, in-domain data are scarce compared with general domain-agnostic data. Second, data used for continual pre-training are not task-aware, such that they may not be helpful to downstream applications. We propose TRAIT, a task-oriented in-domain data augmentation framework. Our framework is divided into two parts: in-domain data selection and task-oriented synthetic passage generation. The data selection strategy identifies and selects a large amount of in-domain data from general corpora, and thus significantly enriches domain knowledge in the continual pre-training data. The synthetic passages contain guidance on how to use domain knowledge to answer questions about downstream tasks. By training on such passages, the model aligns with the need of downstream applications. We adapt LLMs to two domains: advertisement and math. On average, TRAIT improves LLM performance by 8% in the advertisement domain and 7.5% in the math domain.