REVIEW 31 cited by
LMFusion: Adapting Pretrained Language Models for Multimodal Generation
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
LMFusion: Adapting Pretrained Language Models for Multimodal Generation
read the original abstract
We present LMFusion, a framework for empowering pretrained text-only large language models (LLMs) with multimodal generative capabilities, enabling them to understand and generate both text and images in arbitrary sequences. LMFusion leverages existing Llama-3's weights for processing texts autoregressively while introducing additional and parallel transformer modules for processing images with diffusion. During training, the data from each modality is routed to its dedicated modules: modality-specific feedforward layers, query-key-value projections, and normalization layers process each modality independently, while the shared self-attention layers allow interactions across text and image features. By freezing the text-specific modules and only training the image-specific modules, LMFusion preserves the language capabilities of text-only LLMs while developing strong visual understanding and generation abilities. Compared to methods that pretrain multimodal generative models from scratch, our experiments demonstrate that, LMFusion improves image understanding by 20% and image generation by 3.6% using only 50% of the FLOPs while maintaining Llama-3's language capabilities. We also demonstrate that this framework can adapt existing vision-language models with multimodal generation ability. Overall, this framework not only leverages existing computational investments in text-only LLMs but also enables the parallel development of language and vision capabilities, presenting a promising direction for efficient multimodal model development.
Forward citations
Cited by 31 Pith papers
-
Where to Refine, When to Stop: Rethinking Redundancy via Latent Discrepancy for Efficient Visual Autoregressive Generation
LD-Pruning applies latent discrepancy to prune tokens and adaptively skip unconditional branches in VAR models for up to 2.35x faster inference with preserved quality.
-
Co-generation of Layout and Shape from Text via Autoregressive 3D Diffusion
3D-ARD+ unifies autoregressive token prediction with diffusion-based 3D latent generation to co-produce indoor scene layouts and object geometries that follow complex text-specified spatial and semantic constraints.
-
Demystifying Video Reasoning
Video diffusion models perform reasoning primarily along denoising steps via multi-path exploration and convergence (Chain-of-Steps), not sequential frames, with emergent memory, self-correction, and layer specialization.
-
Transfer between Modalities with MetaQueries
MetaQueries act as an efficient bridge allowing multimodal LLMs to augment diffusion-based image generation and editing without complex training or unfreezing the LLM backbone.
-
WISE: A World Knowledge-Informed Semantic Evaluation for Text-to-Image Generation
Text-to-image models show significant limitations in integrating world knowledge, as measured by the new WISE benchmark and WiScore metric across 20 models.
-
Streaming Multi-Agent Autoregressive Diffusion Model with World State Registers
Adding persistently updated, supervised world-state register tokens to streaming multi-agent diffusion improves cross-agent consistency and visual quality in two-agent Minecraft generation.
-
UniTac: A Unified Multimodal Model for Cross-Sensor Tactile Understanding and Generation
UniTac is the first unified multimodal model for cross-sensor tactile understanding and generation, using dual-level representations, two new understanding tasks, and a two-stage training paradigm with sensor-prior sa...
-
Polaris: Scaling Up Instruction-Guided Image Generation Towards Millions of Personalized Style Needs
Polaris retrieves and integrates relevant models from a large library of checkpoints and adapters to enable scalable instruction-guided image generation and editing without additional training.
-
Semantic Generative Tuning for Unified Multimodal Models
Semantic Generative Tuning uses image segmentation as a generative proxy to align misaligned representation spaces in unified multimodal models and improve both perception and generative layout fidelity.
-
STARFlow2: Bridging Language Models and Normalizing Flows for Unified Multimodal Generation
STARFlow2 presents an autoregressive flow-based architecture for unified multimodal text-image generation by interleaving a VLM stream with a TarFlow stream via residual skips and a unified latent space.
-
SpatialFusion: Endowing Unified Image Generation with Intrinsic 3D Geometric Awareness
SpatialFusion internalizes 3D geometric awareness into unified image generation models by pairing an MLLM with a spatial transformer that produces depth maps to constrain diffusion generation.
-
Meta-CoT: Enhancing Granularity and Generalization in Image Editing
Meta-CoT uses two-level decomposition of editing operations into meta-tasks and a CoT consistency reward to improve granularity and generalization, reporting 15.8% gains across 21 tasks.
-
LMGenDrive: Bridging Multimodal Understanding and Generative World Modeling for End-to-End Driving
LMGenDrive unifies LLM-based multimodal understanding with generative world models to output both future driving videos and control signals for end-to-end closed-loop autonomous driving.
-
Demystifying Video Reasoning
Video diffusion models reason along the denoising trajectory (Chain-of-Steps), not primarily across frames, and this mechanism can be nudged by ensembling latent trajectories.
-
LLaMo: Scaling Pretrained Language Models for Unified Motion Understanding and Generation with Continuous Autoregressive Tokens
LLaMo scales pretrained LLMs for unified motion-language tasks by encoding motion into continuous causal latents and adding a flow-matching head for real-time autoregressive generation and captioning.
-
ChatUMM: Robust Context Tracking for Conversational Interleaved Generation
A unified multimodal model trained on synthetic multi-turn dialogues that can interleave image generation with text across a conversation.
-
UniVideo: Unified Understanding, Generation, and Editing for Videos
UniVideo combines a frozen MLLM and a video DiT to unify video understanding, generation, in-context editing, visual prompting, and zero-shot free-form video edits under one instruction interface.
-
Reconstruction Alignment Improves Unified Multimodal Models
RECA, a self-supervised post-training objective that conditions unified multimodal models on their own visual understanding embeddings to reconstruct input images, improves text-to-image and editing benchmarks across ...
-
Interleaving Reasoning for Better Text-to-Image Generation
A text-image-text-image multi-turn pipeline improves text-to-image generation, gaining up to 8 points over the base model on several benchmarks while adding an image-conditioned reflection step.
-
Galaxea Open-World Dataset and G0 Dual-System VLA Model
A new open-world mobile manipulation dataset and a dual-system VLA model show that single-embodiment pre-training, not cross-embodiment pre-training, drives strong downstream task performance.
-
Semantic Generative Tuning for Unified Multimodal Models
Semantic Generative Tuning applies segmentation-based generative proxies during post-training to align and improve both understanding and generation in unified multimodal models.
-
Reversing the Flow: Generation-to-Understanding Synergy in Large Multimodal Models
Generation-to-Understanding synergy lets multimodal models create self-generated visual edits as intermediate steps, improving performance on twelve benchmarks while revealing limits in task-aligned self-reflection.
-
Free Lunch for Unified Multimodal Models: Enhancing Generation via Reflective Rectification with Inherent Understanding
UniRect-CoT is a training-free rectification chain-of-thought framework that treats diffusion denoising as visual reasoning and uses the model's inherent understanding to align and correct intermediate generation results.
-
CG-MLLM: Captioning and Generating 3D content via Multi-modal Large Language Models
CG-MLLM is a multimodal LLM using a Mixture-of-Transformer architecture with separate TokenAR and BlockAR components integrated with a pre-trained vision-language backbone and 3D VAE to enable 3D captioning and high-f...
-
FasterVAR: Plug-and-Play Acceleration for Visual Autoregressive Models
Stage-aware pruning of late generation steps, using random projection and cached-feature restoration, speeds up VAR text-to-image models by up to 3.4x with minimal quality loss.
-
SRUM: Fine-Grained Self-Rewarding for Unified Multimodal Models
A unified multimodal model can improve its own text-to-image generation by using its understanding module as a rewarder in a global-plus-local reward-weighted training loop.
-
OmniGen2: Towards Instruction-Aligned Multimodal Generation
OmniGen2 introduces a unified generative model with two distinct decoding pathways and a decoupled image tokenizer that achieves competitive results on text-to-image and editing benchmarks plus state-of-the-art consis...
-
Emerging Properties in Unified Multimodal Pretraining
BAGEL is a unified decoder-only model that develops emerging complex multimodal reasoning abilities after pretraining on large-scale interleaved data and outperforms prior open-source unified models.
-
BLIP3-o: A Family of Fully Open Unified Multimodal Models-Architecture, Training and Dataset
BLIP3-o uses a diffusion transformer to generate CLIP image features and a sequential pretraining strategy to build open models that perform strongly on both image understanding and generation benchmarks.
-
MMCORE: MultiModal COnnection with Representation Aligned Latent Embeddings
MMCORE transfers VLM reasoning into diffusion-based image generation and editing via aligned latent embeddings from learnable queries, outperforming baselines on text-to-image and editing tasks.
-
Show-o2: Improved Native Unified Multimodal Models
Show-o2 unifies text, image, and video understanding and generation in a single autoregressive-plus-flow-matching model built on 3D causal VAE representations.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.