{"work":{"id":"94248955-4bc5-4517-98a0-66224a36d865","openalex_id":null,"doi":null,"arxiv_id":"2403.05135","raw_key":null,"title":"ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment","authors":null,"authors_text":"Xiwei Hu, Rui Wang, Yixiao Fang, Bin Fu, Pei Cheng, Gang Yu","year":2024,"venue":"cs.CV","abstract":"Diffusion models have demonstrated remarkable performance in the domain of text-to-image generation. However, most widely used models still employ CLIP as their text encoder, which constrains their ability to comprehend dense prompts, encompassing multiple objects, detailed attributes, complex relationships, long-text alignment, etc. In this paper, we introduce an Efficient Large Language Model Adapter, termed ELLA, which equips text-to-image diffusion models with powerful Large Language Models (LLM) to enhance text alignment without training of either U-Net or LLM. To seamlessly bridge two pre-trained models, we investigate a range of semantic alignment connector designs and propose a novel module, the Timestep-Aware Semantic Connector (TSC), which dynamically extracts timestep-dependent conditions from LLM. Our approach adapts semantic features at different stages of the denoising process, assisting diffusion models in interpreting lengthy and intricate prompts over sampling timesteps. Additionally, ELLA can be readily incorporated with community models and tools to improve their prompt-following capabilities. To assess text-to-image models in dense prompt following, we introduce Dense Prompt Graph Benchmark (DPG-Bench), a challenging benchmark consisting of 1K dense prompts. Extensive experiments demonstrate the superiority of ELLA in dense prompt following compared to state-of-the-art methods, particularly in multiple object compositions involving diverse attributes and relationships.","external_url":"https://arxiv.org/abs/2403.05135","cited_by_count":null,"metadata_source":"pith","metadata_fetched_at":"2026-05-25T04:20:19.410261+00:00","pith_arxiv_id":"2403.05135","created_at":"2026-05-09T06:25:47.861863+00:00","updated_at":"2026-06-05T21:23:00.469572+00:00","title_quality_ok":true,"display_title":"ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment","render_title":"ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment"},"hub":{"state":{"work_id":"94248955-4bc5-4517-98a0-66224a36d865","tier":"hub","tier_reason":"10+ Pith inbound or 1,000+ external citations","pith_inbound_count":62,"external_cited_by_count":null,"distinct_field_count":4,"first_pith_cited_at":"2024-09-27T16:06:11+00:00","last_pith_cited_at":"2026-05-22T17:59:42+00:00","author_build_status":"not_needed","summary_status":"needed","contexts_status":"needed","graph_status":"needed","ask_index_status":"not_needed","reader_status":"not_needed","recognition_status":"not_needed","updated_at":"2026-06-24T09:31:40.733442+00:00","tier_text":"hub"},"tier":"hub","role_counts":[{"context_role":"dataset","n":17},{"context_role":"background","n":7},{"context_role":"baseline","n":3}],"polarity_counts":[{"context_polarity":"use_dataset","n":17},{"context_polarity":"background","n":6},{"context_polarity":"baseline","n":3},{"context_polarity":"unclear","n":1}],"runs":{"context_extract":{"job_type":"context_extract","status":"succeeded","result":{"enqueued_papers":25},"error":null,"updated_at":"2026-05-14T17:59:37.569198+00:00"},"graph_features":{"job_type":"graph_features","status":"succeeded","result":{"co_cited":[{"title":"SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis","work_id":"8034c587-fba6-4941-87ba-c98f2ac962cb","shared_citers":19},{"title":"Qwen-Image Technical Report","work_id":"d06d7ecc-7579-4f89-a60b-4278a0f3c562","shared_citers":18},{"title":"Emerging Properties in Unified Multimodal Pretraining","work_id":"e0cfd82c-f5d4-44fd-b531-ec73ab0a805b","shared_citers":17},{"title":"Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling","work_id":"67d9e391-26d1-459e-ab56-07e60511c886","shared_citers":16},{"title":"BLIP3-o: A Family of Fully Open Unified Multimodal Models-Architecture, Training and Dataset","work_id":"86d896d2-592f-4d9b-938e-dfeb11f9388f","shared_citers":14},{"title":"Emu3: Next-Token Prediction is All You Need","work_id":"720d288e-fac0-464c-9929-19efd9a52afc","shared_citers":12},{"title":"Step1X-Edit: A Practical Framework for General Image Editing","work_id":"3392f2c8-a1cb-4d6c-8c82-2cdccffa33f9","shared_citers":11},{"title":"Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer","work_id":"f1080a62-48e1-4255-b023-7556be57370d","shared_citers":11},{"title":"OmniGen2: Towards Instruction-Aligned Multimodal Generation","work_id":"d3153e5f-b6e2-4ab3-9f41-e24e24d64496","shared_citers":10},{"title":"Qwen3-VL Technical Report","work_id":"1fe243aa-e3c0-4da6-b391-4cbcfc88d5c0","shared_citers":10},{"title":"Seedream 3.0 Technical Report","work_id":"013e56d0-7f47-4d0e-bbca-e9540fc0e0cc","shared_citers":10},{"title":"Show-o: One Single Transformer to Unify Multimodal Understanding and Generation","work_id":"1393dc24-a6b2-44e1-b5d7-7009d1fa4811","shared_citers":10},{"title":"UniWorld-V1: High-Resolution Semantic Encoders for Unified Visual Understanding and Generation","work_id":"488a273e-95d8-46f1-87c7-2244068d00d0","shared_citers":10},{"title":"Chameleon: Mixed-Modal Early-Fusion Foundation Models","work_id":"2661b9a6-25cc-41a1-8100-612d2b801289","shared_citers":9},{"title":"Flow Matching for Generative Modeling","work_id":"6edb71c4-5d64-40af-a394-9757ea051a36","shared_citers":9},{"title":"ImgEdit: A Unified Image Editing Dataset and Benchmark","work_id":"059b5c3a-404c-4d30-a631-68c1d88a08a7","shared_citers":9},{"title":"Qwen2.5-VL Technical Report","work_id":"69dffacb-bfe8-442d-be86-48624c60426f","shared_citers":9},{"title":"Classifier-Free Diffusion Guidance","work_id":"acf2c588-c088-4a6c-938e-150ad7c666d7","shared_citers":8},{"title":"FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space","work_id":"5dfe19d5-3541-4803-8fe9-3c8b9e29b281","shared_citers":8},{"title":"Playground v2.5: Three insights towards enhancing aesthetic quality in text-to-image generation.ArXiv, abs/2402.17245","work_id":"53c64fda-8566-434d-bc10-2fdfbc6a55ad","shared_citers":8},{"title":"Show-o2: Improved Native Unified Multimodal Models","work_id":"77f00563-1ce6-4fba-9d4e-c8ce83f716ac","shared_citers":8},{"title":"Wan: Open and Advanced Large-Scale Video Generative Models","work_id":"ad3ebc3b-4224-46c9-b61d-bcf135da0a7c","shared_citers":8},{"title":"arXiv preprint arXiv:2505.22705 (2025)","work_id":"68d4c0f7-3dfd-438d-a823-6a93fd0a835d","shared_citers":7},{"title":"arXiv preprint arXiv:2507.22058 (2025)","work_id":"3ee0ee57-31b9-49d4-98fc-c12c499a14b9","shared_citers":7}],"time_series":[{"n":1,"year":2024},{"n":3,"year":2025},{"n":33,"year":2026}],"dependency_candidates":[]},"error":null,"updated_at":"2026-05-14T17:59:49.495747+00:00"},"identity_refresh":{"job_type":"identity_refresh","status":"succeeded","result":{"items":[{"title":"Qwen3 Technical Report","outcome":"unchanged","work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e","resolver":"local_arxiv","confidence":0.98,"old_work_id":"25a4e30c-1232-48e7-9925-02fa12ba7c9e"}],"counts":{"fixed":0,"merged":0,"unchanged":1,"quarantined":0,"needs_external_resolution":0},"errors":[],"attempted":1},"error":null,"updated_at":"2026-05-14T17:59:54.965882+00:00"},"summary_claims":{"job_type":"summary_claims","status":"succeeded","result":{"title":"ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment","claims":[{"claim_text":"Diffusion models have demonstrated remarkable performance in the domain of text-to-image generation. However, most widely used models still employ CLIP as their text encoder, which constrains their ability to comprehend dense prompts, encompassing multiple objects, detailed attributes, complex relationships, long-text alignment, etc. In this paper, we introduce an Efficient Large Language Model Adapter, termed ELLA, which equips text-to-image diffusion models with powerful Large Language Models (LLM) to enhance text alignment without training of either U-Net or LLM. To seamlessly bridge two pr","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment because it crossed a citation-hub threshold.","role_counts":[]},"error":null,"updated_at":"2026-05-14T17:59:50.311578+00:00"}},"summary":{"title":"ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment","claims":[{"claim_text":"Diffusion models have demonstrated remarkable performance in the domain of text-to-image generation. However, most widely used models still employ CLIP as their text encoder, which constrains their ability to comprehend dense prompts, encompassing multiple objects, detailed attributes, complex relationships, long-text alignment, etc. In this paper, we introduce an Efficient Large Language Model Adapter, termed ELLA, which equips text-to-image diffusion models with powerful Large Language Models (LLM) to enhance text alignment without training of either U-Net or LLM. To seamlessly bridge two pr","claim_type":"abstract","evidence_strength":"source_metadata"}],"why_cited":"Pith tracks ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment because it crossed a citation-hub threshold.","role_counts":[]},"graph":{"co_cited":[{"title":"SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis","work_id":"8034c587-fba6-4941-87ba-c98f2ac962cb","shared_citers":19},{"title":"Qwen-Image Technical Report","work_id":"d06d7ecc-7579-4f89-a60b-4278a0f3c562","shared_citers":18},{"title":"Emerging Properties in Unified Multimodal Pretraining","work_id":"e0cfd82c-f5d4-44fd-b531-ec73ab0a805b","shared_citers":17},{"title":"Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling","work_id":"67d9e391-26d1-459e-ab56-07e60511c886","shared_citers":16},{"title":"BLIP3-o: A Family of Fully Open Unified Multimodal Models-Architecture, Training and Dataset","work_id":"86d896d2-592f-4d9b-938e-dfeb11f9388f","shared_citers":14},{"title":"Emu3: Next-Token Prediction is All You Need","work_id":"720d288e-fac0-464c-9929-19efd9a52afc","shared_citers":12},{"title":"Step1X-Edit: A Practical Framework for General Image Editing","work_id":"3392f2c8-a1cb-4d6c-8c82-2cdccffa33f9","shared_citers":11},{"title":"Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformer","work_id":"f1080a62-48e1-4255-b023-7556be57370d","shared_citers":11},{"title":"OmniGen2: Towards Instruction-Aligned Multimodal Generation","work_id":"d3153e5f-b6e2-4ab3-9f41-e24e24d64496","shared_citers":10},{"title":"Qwen3-VL Technical Report","work_id":"1fe243aa-e3c0-4da6-b391-4cbcfc88d5c0","shared_citers":10},{"title":"Seedream 3.0 Technical Report","work_id":"013e56d0-7f47-4d0e-bbca-e9540fc0e0cc","shared_citers":10},{"title":"Show-o: One Single Transformer to Unify Multimodal Understanding and Generation","work_id":"1393dc24-a6b2-44e1-b5d7-7009d1fa4811","shared_citers":10},{"title":"UniWorld-V1: High-Resolution Semantic Encoders for Unified Visual Understanding and Generation","work_id":"488a273e-95d8-46f1-87c7-2244068d00d0","shared_citers":10},{"title":"Chameleon: Mixed-Modal Early-Fusion Foundation Models","work_id":"2661b9a6-25cc-41a1-8100-612d2b801289","shared_citers":9},{"title":"Flow Matching for Generative Modeling","work_id":"6edb71c4-5d64-40af-a394-9757ea051a36","shared_citers":9},{"title":"ImgEdit: A Unified Image Editing Dataset and Benchmark","work_id":"059b5c3a-404c-4d30-a631-68c1d88a08a7","shared_citers":9},{"title":"Qwen2.5-VL Technical Report","work_id":"69dffacb-bfe8-442d-be86-48624c60426f","shared_citers":9},{"title":"Classifier-Free Diffusion Guidance","work_id":"acf2c588-c088-4a6c-938e-150ad7c666d7","shared_citers":8},{"title":"FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space","work_id":"5dfe19d5-3541-4803-8fe9-3c8b9e29b281","shared_citers":8},{"title":"Playground v2.5: Three insights towards enhancing aesthetic quality in text-to-image generation.ArXiv, abs/2402.17245","work_id":"53c64fda-8566-434d-bc10-2fdfbc6a55ad","shared_citers":8},{"title":"Show-o2: Improved Native Unified Multimodal Models","work_id":"77f00563-1ce6-4fba-9d4e-c8ce83f716ac","shared_citers":8},{"title":"Wan: Open and Advanced Large-Scale Video Generative Models","work_id":"ad3ebc3b-4224-46c9-b61d-bcf135da0a7c","shared_citers":8},{"title":"arXiv preprint arXiv:2505.22705 (2025)","work_id":"68d4c0f7-3dfd-438d-a823-6a93fd0a835d","shared_citers":7},{"title":"arXiv preprint arXiv:2507.22058 (2025)","work_id":"3ee0ee57-31b9-49d4-98fc-c12c499a14b9","shared_citers":7}],"time_series":[{"n":1,"year":2024},{"n":3,"year":2025},{"n":33,"year":2026}],"dependency_candidates":[]},"authors":[]}}