CoPESD: A Multi-Level Surgical Motion Dataset for Training Large Vision-Language Models to Co-Pilot Endoscopic Submucosal Dissection

submucosal dissection (ESD) enables rapid resection of large lesions, minimizing recurrence rates and improving long-term overall survival. Despite these advantages, ESD is technically challenging and carries high risks of complications, necessitating skilled surgeons and precise instruments. Recent advancements in Large Visual-Language Models (LVLMs) offer promising decision support and predictive planning capabilities for robotic systems, which can augment the accuracy of ESD and reduce procedural risks. However, existing datasets for multi-level fine-grained ESD surgical motion understanding are scarce and lack detailed annotations. In this paper, we design a hierarchical decomposition of ESD motion granularity and introduce a multi-level surgical motion dataset (CoPESD) for training LVLMs as the robotic \textbf{Co}-\textbf{P}ilot of \textbf{E}ndoscopic \textbf{S}ubmucosal \textbf{D}issection. CoPESD includes 17,679 images with 32,699 bounding boxes and 88,395 multi-level motions, from over 35 hours of ESD videos for both robot-assisted and conventional surgeries. CoPESD enables granular analysis of ESD motions, focusing on the complex task of submucosal dissection. Extensive experiments on the LVLMs demonstrate the effectiveness of CoPESD in training LVLMs to predict following surgical robotic motions. As the first multimodal ESD motion dataset, CoPESD supports advanced research in ESD instruction-following and surgical automation. The dataset is available at \href{https://github.com/gkw0010/CoPESD}{https://github.com/gkw0010/CoPESD.}}

SAM2Point: Segment Any 3D as Videos in Zero-shot and Promptable Manners

We introduce SAM2Point, a preliminary exploration adapting Segment Anything Model 2 (SAM 2) for zero-shot and promptable 3D segmentation. SAM2Point interprets any 3D data as a series of multi-directional videos, and leverages SAM 2 for 3D-space segmentation, without further training or 2D-3D projection. Our framework supports various prompt types, including 3D points, boxes, and masks, and can generalize across diverse scenarios, such as 3D objects, indoor scenes, outdoor environments, and raw sparse LiDAR. Demonstrations on multiple 3D datasets, e.g., Objaverse, S3DIS, ScanNet, Semantic3D, and KITTI, highlight the robust generalization capabilities of SAM2Point. To our best knowledge, we present the most faithful implementation of SAM in 3D, which may serve as a starting point for future research in promptable 3D segmentation. Online Demo: https://huggingface.co/spaces/ZiyuG/SAM2Point . Code: https://github.com/ZiyuGuo99/SAM2Point .

LLaVA-OneVision: Easy Visual Task Transfer

We present LLaVA-OneVision, a family of open large multimodal models (LMMs) developed by consolidating our insights into data, models, and visual representations in the LLaVA-NeXT blog series. Our experimental results demonstrate that LLaVA-OneVision is the first single model that can simultaneously push the performance boundaries of open LMMs in three important computer vision scenarios: single-image, multi-image, and video scenarios. Importantly, the design of LLaVA-OneVision allows strong transfer learning across different modalities/scenarios, yielding new emerging capabilities. In particular, strong video understanding and cross-scenario capabilities are demonstrated through task transfer from images to videos.

MAVIS: Mathematical Visual Instruction Tuning

Multi-modal Large Language Models (MLLMs) have recently emerged as a significant focus in academia and industry. Despite their proficiency in general multi-modal scenarios, the mathematical problem-solving capabilities in visual contexts remain insufficiently explored. We identify three key areas within MLLMs that need to be improved: visual encoding of math diagrams, diagram-language alignment, and mathematical reasoning skills. This draws forth an urgent demand for large-scale, high-quality data and training pipelines in visual mathematics. In this paper, we propose MAVIS, the first MAthematical VISual instruction tuning paradigm for MLLMs, involving a series of mathematical visual datasets and specialized MLLMs. Targeting the three issues, MAVIS contains three progressive training stages from scratch. First, we curate MAVIS-Caption, consisting of 558K diagram-caption pairs, to fine-tune a math-specific vision encoder (CLIP-Math) through contrastive learning, tailored for improved diagram visual encoding. Second, we utilize MAVIS-Caption to align the CLIP-Math with a large language model (LLM) by a projection layer, enhancing vision-language alignment in mathematical domains. Third, we introduce MAVIS-Instruct, including 900K meticulously collected and annotated visual math problems, which is adopted to finally instruct-tune the MLLM for robust mathematical reasoning skills. In MAVIS-Instruct, we incorporate complete chain-of-thought (CoT) rationales for each problem, and minimize textual redundancy, thereby concentrating the model towards the visual elements. Data and Models are released at https://github.com/ZrrSkywalker/MAVIS

LLaVA-NeXT-Interleave: Tackling Multi-image, Video, and 3D in Large Multimodal Models

Visual instruction tuning has made considerable strides in enhancing the capabilities of Large Multimodal Models (LMMs). However, existing open LMMs largely focus on single-image tasks, their applications to multi-image scenarios remains less explored. Additionally, prior LMM research separately tackles different scenarios, leaving it impossible to generalize cross scenarios with new emerging capabilities. To this end, we introduce LLaVA-NeXT-Interleave, which simultaneously tackles Multi-image, Multi-frame (video), Multi-view (3D), and Multi-patch (single-image) scenarios in LMMs. To enable these capabilities, we regard the interleaved data format as a general template and compile the M4-Instruct dataset with 1,177.6k samples, spanning 4 primary domains with 14 tasks and 41 datasets. We also curate the LLaVA-Interleave Bench to comprehensively evaluate the multi-image performance of LMMs. Through extensive experiments, LLaVA-NeXT-Interleave achieves leading results in multi-image, video, and 3D benchmarks, while maintaining the performance of single-image tasks. Besides, our model also exhibits several emerging capabilities, e.g., transferring tasks across different settings and modalities. Code is available at https://github.com/LLaVA-VL/LLaVA-NeXT

RoboMamba: Multimodal State Space Model for Efficient Robot Reasoning and Manipulation

A fundamental objective in robot manipulation is to enable models to comprehend visual scenes and execute actions. Although existing robot Multimodal Large Language Models (MLLMs) can handle a range of basic tasks, they still face challenges in two areas: 1) inadequate reasoning ability to tackle complex tasks, and 2) high computational costs for MLLM fine-tuning and inference. The recently proposed state space model (SSM) known as Mamba demonstrates promising capabilities in non-trivial sequence modeling with linear inference complexity. Inspired by this, we introduce RoboMamba, an end-to-end robotic MLLM that leverages the Mamba model to deliver both robotic reasoning and action capabilities, while maintaining efficient fine-tuning and inference. Specifically, we first integrate the vision encoder with Mamba, aligning visual data with language embedding through co-training, empowering our model with visual common sense and robot-related reasoning. To further equip RoboMamba with action pose prediction abilities, we explore an efficient fine-tuning strategy with a simple policy head. We find that once RoboMamba possesses sufficient reasoning capability, it can acquire manipulation skills with minimal fine-tuning parameters (0.1\% of the model) and time (20 minutes). In experiments, RoboMamba demonstrates outstanding reasoning capabilities on general and robotic evaluation benchmarks. Meanwhile, our model showcases impressive pose prediction results in both simulation and real-world experiments, achieving inference speeds 7 times faster than existing robot MLLMs. Our project web page: https://sites.google.com/view/robomamba-web

Video-MME: The First-Ever Comprehensive Evaluation Benchmark of Multi-modal LLMs in Video Analysis

In the quest for artificial general intelligence, Multi-modal Large Language Models (MLLMs) have emerged as a focal point in recent advancements. However, the predominant focus remains on developing their capabilities in static image understanding. The potential of MLLMs in processing sequential visual data is still insufficiently explored, highlighting the absence of a comprehensive, high-quality assessment of their performance. In this paper, we introduce Video-MME, the first-ever full-spectrum, Multi-Modal Evaluation benchmark of MLLMs in Video analysis. Our work distinguishes from existing benchmarks through four key features: 1) Diversity in video types, spanning 6 primary visual domains with 30 subfields to ensure broad scenario generalizability; 2) Duration in temporal dimension, encompassing both short-, medium-, and long-term videos, ranging from 11 seconds to 1 hour, for robust contextual dynamics; 3) Breadth in data modalities, integrating multi-modal inputs besides video frames, including subtitles and audios, to unveil the all-round capabilities of MLLMs; 4) Quality in annotations, utilizing rigorous manual labeling by expert annotators to facilitate precise and reliable model assessment. 900 videos with a total of 256 hours are manually selected and annotated by repeatedly viewing all the video content, resulting in 2,700 question-answer pairs. With Video-MME, we extensively evaluate various state-of-the-art MLLMs, including GPT-4 series and Gemini 1.5 Pro, as well as open-source image models like InternVL-Chat-V1.5 and video models like LLaVA-NeXT-Video. Our experiments reveal that Gemini 1.5 Pro is the best-performing commercial model, significantly outperforming the open-source models. Our dataset along with these findings underscores the need for further improvements in handling longer sequences and multi-modal data. Project Page: https://video-mme.github.io

TripletMix: Triplet Data Augmentation for 3D Understanding

Data augmentation has proven to be a vital tool for enhancing the generalization capabilities of deep learning models, especially in the context of 3D vision where traditional datasets are often limited. Despite previous advancements, existing methods primarily cater to unimodal data scenarios, leaving a gap in the augmentation of multimodal triplet data, which integrates text, images, and point clouds. Simultaneously augmenting all three modalities enhances diversity and improves alignment across modalities, resulting in more comprehensive and robust 3D representations. To address this gap, we propose TripletMix, a novel approach to address the previously unexplored issue of multimodal data augmentation in 3D understanding. TripletMix innovatively applies the principles of mixed-based augmentation to multimodal triplet data, allowing for the preservation and optimization of cross-modal connections. Our proposed TripletMix combines feature-level and input-level augmentations to achieve dual enhancement between raw data and latent features, significantly improving the model's cross-modal understanding and generalization capabilities by ensuring feature consistency and providing diverse and realistic training samples. We demonstrate that TripletMix not only improves the baseline performance of models in various learning scenarios including zero-shot and linear probing classification but also significantly enhances model generalizability. Notably, we improved the zero-shot classification accuracy on ScanObjectNN from 51.3 percent to 61.9 percent, and on Objaverse-LVIS from 46.8 percent to 51.4 percent. Our findings highlight the potential of multimodal data augmentation to significantly advance 3D object recognition and understanding.

Self-Corrected Multimodal Large Language Model for End-to-End Robot Manipulation

Robot manipulation policies have shown unsatisfactory action performance when confronted with novel task or object instances. Hence, the capability to automatically detect and self-correct failure action is essential for a practical robotic system. Recently, Multimodal Large Language Models (MLLMs) have shown promise in visual instruction following and demonstrated strong reasoning abilities in various tasks. To unleash general MLLMs as an end-to-end robotic agent, we introduce a Self-Corrected (SC)-MLLM, equipping our model not only to predict end-effector poses but also to autonomously recognize and correct failure actions. Specifically, we first conduct parameter-efficient fine-tuning to empower MLLM with pose prediction ability, which is reframed as a language modeling problem. When facing execution failures, our model learns to identify low-level action error causes (i.e., position and rotation errors) and adaptively seeks prompt feedback from experts. Based on the feedback, SC-MLLM rethinks the current failure scene and generates the corrected actions. Furthermore, we design a continuous policy learning method for successfully corrected samples, enhancing the model's adaptability to the current scene configuration and reducing the frequency of expert intervention. To evaluate our SC-MLLM, we conduct extensive experiments in both simulation and real-world settings. SC-MLLM agent significantly improve manipulation accuracy compared to previous state-of-the-art robotic MLLM (ManipLLM), increasing from 57\% to 79\% on seen object categories and from 47\% to 69\% on unseen novel categories.

SPP: Sparsity-Preserved Parameter-Efficient Fine-Tuning for Large Language Models

Large Language Models (LLMs) have become pivotal in advancing the field of artificial intelligence, yet their immense sizes pose significant challenges for both fine-tuning and deployment. Current post-training pruning methods, while reducing the sizes of LLMs, often fail to maintain their original performance. To address these challenges, this paper introduces SPP, a Sparsity-Preserved Parameter-efficient fine-tuning method. Different from existing post-training pruning approaches that struggle with performance retention, SPP proposes to employ lightweight learnable column and row matrices to optimize sparse LLM weights, keeping the structure and sparsity of pruned pre-trained models intact. By element-wise multiplication and residual addition, SPP ensures the consistency of model sparsity pattern and ratio during both training and weight-merging processes. We demonstrate the effectiveness of SPP by applying it to the LLaMA and LLaMA-2 model families with recent post-training pruning methods. Our results show that SPP significantly enhances the performance of models with different sparsity patterns (i.e. unstructured and N:M sparsity), especially for those with high sparsity ratios (e.g. 75%), making it a promising solution for the efficient fine-tuning of sparse LLMs. Code will be made available at https://github.com/Lucky-Lance/SPP.