MMMU provides 11.5K heterogeneous college-level multimodal questions that current models solve at 56-59% accuracy, establishing a new standard for expert multimodal evaluation.
hub Canonical reference
LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention
Canonical reference. 95% of citing Pith papers cite this work as background.
51
Pith papers citing it
Background
95% of classified citations
abstract
We present LLaMA-Adapter, a lightweight adaption method to efficiently fine-tune LLaMA into an instruction-following model. Using 52K self-instruct demonstrations, LLaMA-Adapter only introduces 1.2M learnable parameters upon the frozen LLaMA 7B model, and costs less than one hour for fine-tuning on 8 A100 GPUs. Specifically, we adopt a set of learnable adaption prompts, and prepend them to the word tokens at higher transformer layers. Then, a zero-initialized attention mechanism with zero gating is proposed, which adaptively injects the new instructional cues into LLaMA, while effectively preserves its pre-trained knowledge. With our efficient training, LLaMA-Adapter can generate high-quality responses, comparable to Alpaca with fully fine-tuned 7B parameters. Besides language commands, our approach can be simply extended to multi-modal instructions for learning image-conditioned LLaMA model, which achieves superior reasoning performance on ScienceQA and COCO Caption benchmarks. Furthermore, we also evaluate the zero-initialized attention mechanism for fine-tuning other pre-trained models (ViT, RoBERTa) on traditional vision and language tasks, demonstrating the superior generalization capacity of our approach. Code is released at https://github.com/OpenGVLab/LLaMA-Adapter.
hub tools
citation-role summary
background 19
baseline 1
citation-polarity summary
representative citing papers
MMMU-Pro is a stricter multimodal benchmark that removes text-only solvable questions, augments options, and requires reading text from images, yielding substantially lower model scores of 16.8-26.9%.
GPT-4-generated instruction data produces superior zero-shot performance in finetuned LLaMA models versus prior state-of-the-art data.
FTP-1 is the first foundation tactile policy pretrained on ~3000 hours of data from 26 sources across 21 sensors that improves performance on seen setups by 17.2% and transfers to unseen sensors with 31% success rate gain.
HARP is a train-based data selector for LLM finetuning that uses hierarchical active region pruning and empirical Bayes posteriors to achieve up to 8.9 point gains with roughly 7 times fewer training examples.
P²-DPO generates on-policy preference pairs targeting focus-and-enhance perception and visual robustness, combined with a calibration loss, to reduce hallucinations in LVLMs more effectively than human-feedback baselines.
EchoVQA is the first large-scale VQA dataset for echocardiography spanning high- and low-quality images across views, with acquisition guidance questions, paired with a low-parameter multimodal prompt model that reports SOTA on several benchmarks.
WorldSense provides the first benchmark requiring synergistic audio-video-text understanding on 1,662 real-world videos and 3,172 QA pairs, where the best current multimodal LLM reaches only 65.1% accuracy.
WE-MATH benchmark reveals most LMMs rely on rote memorization for visual math while GPT-4o has shifted toward knowledge generalization.
LLaMA-Adapter turns frozen LLaMA 7B into a capable instruction follower using only 1.2M new parameters and zero-init attention, matching Alpaca while extending to image-conditioned reasoning on ScienceQA and COCO.
AnchorSeg uses ordered query banks of latent reasoning tokens plus a spatial anchor token and a Token-Mask Cycle Consistency loss to achieve 67.7% gIoU and 68.1% cIoU on the ReasonSeg benchmark.
Set-of-Mark prompting marks segmented image regions with alphanumerics and masks to let GPT-4V achieve state-of-the-art zero-shot results on referring expression comprehension and segmentation benchmarks like RefCOCOg.
LLaVA is trained on GPT-4 generated visual instruction data to achieve 85.1% relative performance to GPT-4 on synthetic multimodal tasks and 92.53% accuracy on Science QA.
Qwen-RobotWorld is a language-conditioned video world model using Double-Stream MMDiT, an 8.6M-frame embodied corpus, and progressive curriculum training that ranks first on EWMBench and DreamGen Bench.
Lance presents a dual-stream mixture-of-experts model with modality-aware positional encoding and staged multi-task training that outperforms prior open-source unified models on image and video generation while keeping strong understanding performance.
UniMind unifies multi-task brain decoding from EEG by bridging signals to LLMs via a Neuro-Language Connector and dynamic task queries, outperforming prior models by 12% on average across ten datasets.
MathFlow decouples perception and inference stages in MLLMs for visual math, with a dedicated perception model delivering gains on the FlowVerse benchmark when paired with existing reasoners.
HybridVLA unifies diffusion and autoregression in a single VLA model via collaborative training and ensemble to raise robot manipulation success rates by 14% in simulation and 19% in real-world tasks.
Introduces the QEVD benchmark for asynchronous situated interaction in fitness coaching and proposes a streaming baseline to address limitations of existing vision-language models.
MVBench is a benchmark of 20 temporal video understanding tasks built by transforming static tasks into dynamic ones, with VideoChat2 outperforming prior MLLMs by over 15%.
AWQ quantizes LLM weights to low bits by scaling salient channels based on activation statistics, outperforming prior methods on language, coding, math, and multi-modal benchmarks.
Otter is a multi-modal model instruction-tuned on the MIMIC-IT dataset of over 3 million in-context instruction-response pairs to improve convergence and generalization on tasks with multiple images and videos.
LLM-X is a scalable architecture for direct negotiation and communication among personal LLM agents, featuring federated gateways, typed protocols, and policy enforcement, shown stable in experiments with up to 12 agents.
ALAM introduces algebraic consistency regularization on latent action transitions from videos, raising VLA success rates from 47.9% to 85.0% on MetaWorld MT50 and 94.1% to 98.1% on LIBERO.
citing papers explorer
-
P$^2$-DPO: Grounding Hallucination in Perceptual Processing via Calibration Direct Preference Optimization
P²-DPO generates on-policy preference pairs targeting focus-and-enhance perception and visual robustness, combined with a calibration loss, to reduce hallucinations in LVLMs more effectively than human-feedback baselines.
-
EchoVQA: Enabling Conversational Assistance for Point-of-Care Cardiac Ultrasound
EchoVQA is the first large-scale VQA dataset for echocardiography spanning high- and low-quality images across views, with acquisition guidance questions, paired with a low-parameter multimodal prompt model that reports SOTA on several benchmarks.
-
WorldSense: Evaluating Real-world Omnimodal Understanding for Multimodal LLMs
WorldSense provides the first benchmark requiring synergistic audio-video-text understanding on 1,662 real-world videos and 3,172 QA pairs, where the best current multimodal LLM reaches only 65.1% accuracy.
-
LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention
LLaMA-Adapter turns frozen LLaMA 7B into a capable instruction follower using only 1.2M new parameters and zero-init attention, matching Alpaca while extending to image-conditioned reasoning on ScienceQA and COCO.
-
AnchorSeg: Language Grounded Query Banks for Reasoning Segmentation
AnchorSeg uses ordered query banks of latent reasoning tokens plus a spatial anchor token and a Token-Mask Cycle Consistency loss to achieve 67.7% gIoU and 68.1% cIoU on the ReasonSeg benchmark.
-
Set-of-Mark Prompting Unleashes Extraordinary Visual Grounding in GPT-4V
Set-of-Mark prompting marks segmented image regions with alphanumerics and masks to let GPT-4V achieve state-of-the-art zero-shot results on referring expression comprehension and segmentation benchmarks like RefCOCOg.
-
Visual Instruction Tuning
LLaVA is trained on GPT-4 generated visual instruction data to achieve 85.1% relative performance to GPT-4 on synthetic multimodal tasks and 92.53% accuracy on Science QA.
-
Qwen-RobotWorld Technical Report: Unifying Embodied World Modeling through Language-Conditioned Video Generation
Qwen-RobotWorld is a language-conditioned video world model using Double-Stream MMDiT, an 8.6M-frame embodied corpus, and progressive curriculum training that ranks first on EWMBench and DreamGen Bench.
-
Lance: Unified Multimodal Modeling by Multi-Task Synergy
Lance presents a dual-stream mixture-of-experts model with modality-aware positional encoding and staged multi-task training that outperforms prior open-source unified models on image and video generation while keeping strong understanding performance.
-
MathFlow: Enhancing the Perceptual Flow of MLLMs for Visual Mathematical Problems
MathFlow decouples perception and inference stages in MLLMs for visual math, with a dedicated perception model delivering gains on the FlowVerse benchmark when paired with existing reasoners.
-
HybridVLA: Collaborative Diffusion and Autoregression in a Unified Vision-Language-Action Model
HybridVLA unifies diffusion and autoregression in a single VLA model via collaborative training and ensemble to raise robot manipulation success rates by 14% in simulation and 19% in real-world tasks.
-
What to Say and When to Say it: Live Fitness Coaching as a Testbed for Situated Interaction
Introduces the QEVD benchmark for asynchronous situated interaction in fitness coaching and proposes a streaming baseline to address limitations of existing vision-language models.
-
MVBench: A Comprehensive Multi-modal Video Understanding Benchmark
MVBench is a benchmark of 20 temporal video understanding tasks built by transforming static tasks into dynamic ones, with VideoChat2 outperforming prior MLLMs by over 15%.
-
Otter: A Multi-Modal Model with In-Context Instruction Tuning
Otter is a multi-modal model instruction-tuned on the MIMIC-IT dataset of over 3 million in-context instruction-response pairs to improve convergence and generalization on tasks with multiple images and videos.
-
ReasonEdit: Towards Interpretable Image Editing Evaluation via Reinforcement Learning
ReasonEdit uses a new CoT dataset and reinforcement learning to produce interpretable, human-aligned evaluations of text-guided image edits.
-
ShareGPT4V: Improving Large Multi-Modal Models with Better Captions
A new 1.2M-caption dataset generated via GPT-4V improves LMMs on MME and MMBench by 222.8/22.0/22.3 and 2.7/1.3/1.5 points respectively when used for supervised fine-tuning.
-
Video-LLaVA: Learning United Visual Representation by Alignment Before Projection
Video-LLaVA creates a unified visual representation for images and videos via pre-projection alignment, enabling mutual enhancement from joint training and strong results on image and video benchmarks.
-
IP-Adapter: Text Compatible Image Prompt Adapter for Text-to-Image Diffusion Models
IP-Adapter adds effective image prompting to text-to-image diffusion models using a lightweight decoupled cross-attention adapter that works alongside text prompts and other controls.
-
CLIP-SVD: Efficient and Interpretable Vision-Language Adaptation via Singular Values
CLIP-SVD performs parameter-efficient adaptation of CLIP by fine-tuning singular values from SVD of weight matrices, reporting SOTA few-shot accuracy on 21 datasets plus a language-based interpretability analysis.
-
LLaVA-Octopus: Unlocking Instruction-Driven Adaptive Projector Fusion for Video Understanding
LLaVA-Octopus introduces instruction-driven adaptive fusion of multiple visual projectors in a multimodal LLM to improve video understanding performance.
-
PLLaVA : Parameter-free LLaVA Extension from Images to Videos for Video Dense Captioning
A temporal pooling layer added to LLaVA smooths video feature distributions and lifts performance on dense video captioning and QA to new SOTA levels without extra parameters.
-
AppAgent: Multimodal Agents as Smartphone Users
AppAgent lets large language models operate diverse smartphone apps via visual interactions and learns app usage from exploration or demonstrations.
-
LLaMA-Adapter V2: Parameter-Efficient Visual Instruction Model
LLaMA-Adapter V2 achieves open-ended visual instruction following in LLMs by unlocking more parameters, early fusion of visual tokens, and joint training on disjoint parameter groups with only 14M added parameters.
-
LLaVA-OneVision: Easy Visual Task Transfer
LLaVA-OneVision is the first single open LMM to simultaneously achieve strong performance in single-image, multi-image, and video scenarios with cross-scenario transfer capabilities.
-
InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks
InternVL scales a vision model to 6B parameters and aligns it with LLMs using web data to achieve state-of-the-art results on 32 visual-linguistic benchmarks.
-
UnAC: Adaptive Visual Prompting with Abstraction and Stepwise Checking for Complex Multimodal Reasoning
UnAC improves LMM performance on visual reasoning benchmarks by combining adaptive visual prompting, image abstraction, and gradual self-checking.
-
OpenFlamingo: An Open-Source Framework for Training Large Autoregressive Vision-Language Models
OpenFlamingo provides open-source autoregressive vision-language models that achieve 80-89% of Flamingo performance on seven vision-language datasets.
-
A Survey on Multimodal Large Language Models
This survey organizes the architectures, training strategies, data, evaluation methods, extensions, and challenges of Multimodal Large Language Models.