Загрузка видео...

Не удалось загрузить видео

Возникла проблема при загрузке этого видео. Это может быть связано с временными проблемами сети или видео может быть недоступно.

На главную

Most don't know (1) how easy it is to invert embedding vectors back into sentences, (2) this is a perfect task text diffusion models. Here's a 78M parameter model and live demo that recovers 80% of tokens from Qwen3-Embedding and EmbeddingGemma vectors. Works even on multilingual input.

Jina AI

17,299 subscribers

12,977 просмотров • 5 месяцев назад •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

The next chapter about transformers is up on YouTube, digging into the attention mechanism: The model works with vectors representing tokens (think words), and this is the mechanism that allows those vectors to take in meaning from context.

The next chapter about transformers is up on YouTube, digging into the attention mechanism: The model works with vectors representing tokens (think words), and this is the mechanism that allows those vectors to take in meaning from context.

Grant Sanderson

810,238 просмотров • 2 лет назад

[CLIP] by Hand ✍️ The CLIP (Contrastive Language–Image Pre-training) model, a groundbreaking work by OpenAI, redefines the intersection of computer vision and natural language processing. It is the basis of all the multi-modal foundation models we see today. How does CLIP work? Goal: 🟨 Learn a shared embedding space for text and image [1] Given ↳ A mini batch of 3 text-image pairs ↳ OpenAI used 400 million text-image pairs to train its original CLIP model. Process 1st pair: "big table" [2] 🟪 Text → 2 Vectors (3D) ↳ Look up word embedding vectors using word2vec. [3] 🟩 Image → 2 Vectors (4D) ↳ Divide the image into two patches. ↳ Flatten each patch [4] Process other pairs ↳ Repeat [2]-[3] [5] 🟪 Text Encoder & 🟩 Image Encoder ↳ Encode input vectors into feature vectors ↳ Here, both encoders are simple one layer perceptron (linear + ReLU) ↳ In practice, the encoders are usually transformer models. [6] 🟪 🟩 Mean Pooling: 2 → 1 vector ↳ Average 2 feature vectors into a single vector by averaging across the columns ↳ The goal is to have one vector to represent each image or text [7] 🟪 🟩 -> 🟨 Projection ↳ Note that the text and image feature vectors from the encoders have different dimensions (3D vs. 4D). ↳ Use a linear layer to project image and text vectors to a 2D shared embedding space. 🏋️ Contrastive Pre-training 🏋️ [8] Prepare for MatMul ↳ Copy text vectors (T1,T2,T3) ↳ Copy the transpose of image vectors (I1,I2,I3) ↳ They are all in the 2D shared embedding space. [9] 🟦 MatMul ↳ Multiply T and I matrices. ↳ This is equivalent to taking dot product between every pair of image and text vectors. ↳ The purpose is to use dot product to estimate the similarity between a pair of image-text. [10] 🟦 Softmax: e^x ↳ Raise e to the power of the number in each cell ↳ To simplify hand calculation, we approximate e^□ with 3^□. [11] 🟦 Softmax: ∑ ↳ Sum each row for 🟩 image→🟪 text ↳ Sum each column for 🟪 text→ 🟩 image [12] 🟦 Softmax: 1 / sum ↳ Divide each element by the column sum to obtain a similarity matrix for 🟪 text→🟩 image ↳ Divide each element by the row sum to obtain a similarity matrix for 🟩 image→🟪 text [13] 🟥 Loss Gradients ↳ The "Targets" for the similarity matrices are Identity Matrices. ↳ Why? If I and T come from the same pair (i=j), we want the highest value, which is 1, and 0 otherwise. ↳ Apply the simple equation of [Similarity - Target] to compute gradients of for both directions. ↳ Why so simple? Because when Softmax and Cross-Entropy Loss are used together, the math magically works out that way. ↳ These gradients kick off the backpropagation process to update weights and biases of the encoders and projection layers (red borders).

[CLIP] by Hand ✍️ The CLIP (Contrastive Language–Image Pre-training) model, a groundbreaking work by OpenAI, redefines the intersection of computer vision and natural language processing. It is the basis of all the multi-modal foundation models we see today. How does CLIP work? Goal: 🟨 Learn a shared embedding space for text and image [1] Given ↳ A mini batch of 3 text-image pairs ↳ OpenAI used 400 million text-image pairs to train its original CLIP model. Process 1st pair: "big table" [2] 🟪 Text → 2 Vectors (3D) ↳ Look up word embedding vectors using word2vec. [3] 🟩 Image → 2 Vectors (4D) ↳ Divide the image into two patches. ↳ Flatten each patch [4] Process other pairs ↳ Repeat [2]-[3] [5] 🟪 Text Encoder & 🟩 Image Encoder ↳ Encode input vectors into feature vectors ↳ Here, both encoders are simple one layer perceptron (linear + ReLU) ↳ In practice, the encoders are usually transformer models. [6] 🟪 🟩 Mean Pooling: 2 → 1 vector ↳ Average 2 feature vectors into a single vector by averaging across the columns ↳ The goal is to have one vector to represent each image or text [7] 🟪 🟩 -> 🟨 Projection ↳ Note that the text and image feature vectors from the encoders have different dimensions (3D vs. 4D). ↳ Use a linear layer to project image and text vectors to a 2D shared embedding space. 🏋️ Contrastive Pre-training 🏋️ [8] Prepare for MatMul ↳ Copy text vectors (T1,T2,T3) ↳ Copy the transpose of image vectors (I1,I2,I3) ↳ They are all in the 2D shared embedding space. [9] 🟦 MatMul ↳ Multiply T and I matrices. ↳ This is equivalent to taking dot product between every pair of image and text vectors. ↳ The purpose is to use dot product to estimate the similarity between a pair of image-text. [10] 🟦 Softmax: e^x ↳ Raise e to the power of the number in each cell ↳ To simplify hand calculation, we approximate e^□ with 3^□. [11] 🟦 Softmax: ∑ ↳ Sum each row for 🟩 image→🟪 text ↳ Sum each column for 🟪 text→ 🟩 image [12] 🟦 Softmax: 1 / sum ↳ Divide each element by the column sum to obtain a similarity matrix for 🟪 text→🟩 image ↳ Divide each element by the row sum to obtain a similarity matrix for 🟩 image→🟪 text [13] 🟥 Loss Gradients ↳ The "Targets" for the similarity matrices are Identity Matrices. ↳ Why? If I and T come from the same pair (i=j), we want the highest value, which is 1, and 0 otherwise. ↳ Apply the simple equation of [Similarity - Target] to compute gradients of for both directions. ↳ Why so simple? Because when Softmax and Cross-Entropy Loss are used together, the math magically works out that way. ↳ These gradients kick off the backpropagation process to update weights and biases of the encoders and projection layers (red borders).

Tom Yeh

67,834 просмотров • 2 лет назад

New short course: Building Multimodal Search and RAG", by Weaviate AI Database's Sebastia(N_) Witalec ✊🏽✊🏾✊🏿. Contrastive learning is used to train models to map vectors into an embedding space by pulling similar concepts closer together and pushing dissimilar concepts away from each other. This technique is also used to train multimodal embedding models that capture semantic similarity across different modalities like text, images, and audio. These multimodal embeddings can be used to build multimodal search and RAG systems. In this course, you'll learn how contrastive learning works, and how to add multimodality to RAG – so your models can draw on diverse, relevant context to answer questions. For example, a query about a financial report might synthesize information from text snippets, graphs, tables, and slides. You will also learn how visual instruction tuning lets you integrate image understanding into language models, and build a multi-vector recommender system using Weaviate’s open-source vector database. Please sign up here:

New short course: Building Multimodal Search and RAG", by Weaviate AI Database's Sebastia(N_) Witalec ✊🏽✊🏾✊🏿. Contrastive learning is used to train models to map vectors into an embedding space by pulling similar concepts closer together and pushing dissimilar concepts away from each other. This technique is also used to train multimodal embedding models that capture semantic similarity across different modalities like text, images, and audio. These multimodal embeddings can be used to build multimodal search and RAG systems. In this course, you'll learn how contrastive learning works, and how to add multimodality to RAG – so your models can draw on diverse, relevant context to answer questions. For example, a query about a financial report might synthesize information from text snippets, graphs, tables, and slides. You will also learn how visual instruction tuning lets you integrate image understanding into language models, and build a multi-vector recommender system using Weaviate’s open-source vector database. Please sign up here:

Andrew Ng

104,371 просмотров • 2 лет назад

1/N Most Vision-Language-Action models need tons of data for finetuning, and still fail for new objects and instructions. Introducing OTTER, a lightweight, easy-to-train model that uses text-aware visual features to nail unseen tasks out of the box! Here's how it works 👇

1/N Most Vision-Language-Action models need tons of data for finetuning, and still fail for new objects and instructions. Introducing OTTER, a lightweight, easy-to-train model that uses text-aware visual features to nail unseen tasks out of the box! Here's how it works 👇

Fangchen Liu

68,366 просмотров • 1 год назад

Google just proved that bigger isn't always better. Their 308M parameter model is outperforming models 2x its size. Google just released 𝗘𝗺𝗯𝗲𝗱𝗱𝗶𝗻𝗴𝗚𝗲𝗺𝗺𝗮, and it's proving that lightweight embedding models can punch way above their weight class. At just 308M parameters (578MB), it's the new state-of-the-art for models under 500M parameters across MTEB multilingual, English, and code benchmarks. But the really impressive part is that it ranks 8th overall on MTEB(Multilingual, v2) - that's 𝟭𝟳 𝗽𝗹𝗮𝗰𝗲𝘀 above the second-best sub-500M model, and it's delivering performance 𝗰𝗼𝗺𝗽𝗮𝗿𝗮𝗯𝗹𝗲 𝘁𝗼 𝗺𝗼𝗱𝗲𝗹𝘀 𝗻𝗲𝗮𝗿𝗹𝘆 𝗱𝗼𝘂𝗯𝗹𝗲 𝗶𝘁𝘀 𝘀𝗶𝘇𝗲. There are three key parts of their training recipe that sets it apart: 𝟭. 𝗘𝗻𝗰𝗼𝗱𝗲𝗿-𝗗𝗲𝗰𝗼𝗱𝗲𝗿 𝗜𝗻𝗶𝘁𝗶𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻 Instead of starting from a decoder-only Gemma 3 model, they first adapted it to encoder-decoder, then used just the encoder. By basing EmbeddingGemma off an LLM that already has world and language understanding, it gives it a stronger starting point. 𝟮. 𝗧𝗵𝗿𝗲𝗲-𝗟𝗼𝘀𝘀 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 They combine three different loss functions, instead of just having one: • Contrastive loss (NCE) with in-batch negatives and hardness weighting • Spread-out regularization to ensure embeddings utilize the full space (for quantization and ANN retrieval) • Embedding matching distillation from Gemini Embedding - not just learning from relevance scores, but directly aligning the embedding space with the teacher model 𝟯. 𝗠𝗼𝗱𝗲𝗹 𝗦𝗼𝘂𝗽𝗶𝗻𝗴 Rather than just averaging checkpoints from the same training run, they use optimization techniques to find multiple specialized training mixtures. Each mixture creates an "expert" model in different domains, and averaging all their parameters creates a final model that's actually better than individual models. Extras: • Matryoshka embeddings supporting 768, 512, 256, and 128 dimensions • Quantization-aware training - maintains quality even at int4 precision • 100+ languages from Gemma 3 pretraining • Exceptional performance on low-resource languages (check their XTREME-UP results) Is it the absolute best embedding model? No - Gemini Embedding still leads overall. But that's not really the point. EmbeddingGemma proves you can achieve state-of-the-art performance in a small package that's actually deployable on-device, in low-latency applications, and in resource-constrained environments. This makes good embeddings accessible for use cases that I'm seeing more and more: offline applications, privacy-sensitive deployments, and high-throughput scenarios where inference cost actually matters. Full paper: Shoutout to the EmbeddingGemma team at Google DeepMind for this awesome open source work 💙 and to Daniel Williams for helping me with this video! 🫶

Google just proved that bigger isn't always better. Their 308M parameter model is outperforming models 2x its size. Google just released 𝗘𝗺𝗯𝗲𝗱𝗱𝗶𝗻𝗴𝗚𝗲𝗺𝗺𝗮, and it's proving that lightweight embedding models can punch way above their weight class. At just 308M parameters (578MB), it's the new state-of-the-art for models under 500M parameters across MTEB multilingual, English, and code benchmarks. But the really impressive part is that it ranks 8th overall on MTEB(Multilingual, v2) - that's 𝟭𝟳 𝗽𝗹𝗮𝗰𝗲𝘀 above the second-best sub-500M model, and it's delivering performance 𝗰𝗼𝗺𝗽𝗮𝗿𝗮𝗯𝗹𝗲 𝘁𝗼 𝗺𝗼𝗱𝗲𝗹𝘀 𝗻𝗲𝗮𝗿𝗹𝘆 𝗱𝗼𝘂𝗯𝗹𝗲 𝗶𝘁𝘀 𝘀𝗶𝘇𝗲. There are three key parts of their training recipe that sets it apart: 𝟭. 𝗘𝗻𝗰𝗼𝗱𝗲𝗿-𝗗𝗲𝗰𝗼𝗱𝗲𝗿 𝗜𝗻𝗶𝘁𝗶𝗮𝗹𝗶𝘇𝗮𝘁𝗶𝗼𝗻 Instead of starting from a decoder-only Gemma 3 model, they first adapted it to encoder-decoder, then used just the encoder. By basing EmbeddingGemma off an LLM that already has world and language understanding, it gives it a stronger starting point. 𝟮. 𝗧𝗵𝗿𝗲𝗲-𝗟𝗼𝘀𝘀 𝗧𝗿𝗮𝗶𝗻𝗶𝗻𝗴 They combine three different loss functions, instead of just having one: • Contrastive loss (NCE) with in-batch negatives and hardness weighting • Spread-out regularization to ensure embeddings utilize the full space (for quantization and ANN retrieval) • Embedding matching distillation from Gemini Embedding - not just learning from relevance scores, but directly aligning the embedding space with the teacher model 𝟯. 𝗠𝗼𝗱𝗲𝗹 𝗦𝗼𝘂𝗽𝗶𝗻𝗴 Rather than just averaging checkpoints from the same training run, they use optimization techniques to find multiple specialized training mixtures. Each mixture creates an "expert" model in different domains, and averaging all their parameters creates a final model that's actually better than individual models. Extras: • Matryoshka embeddings supporting 768, 512, 256, and 128 dimensions • Quantization-aware training - maintains quality even at int4 precision • 100+ languages from Gemma 3 pretraining • Exceptional performance on low-resource languages (check their XTREME-UP results) Is it the absolute best embedding model? No - Gemini Embedding still leads overall. But that's not really the point. EmbeddingGemma proves you can achieve state-of-the-art performance in a small package that's actually deployable on-device, in low-latency applications, and in resource-constrained environments. This makes good embeddings accessible for use cases that I'm seeing more and more: offline applications, privacy-sensitive deployments, and high-throughput scenarios where inference cost actually matters. Full paper: Shoutout to the EmbeddingGemma team at Google DeepMind for this awesome open source work 💙 and to Daniel Williams for helping me with this video! 🫶

Victoria Slocum

21,592 просмотров • 8 месяцев назад

A new look on our Terrain Editor/Generator. Everything is computed on GPU so we were surprised that even when the whole terrain is recomputed from scratch, it is still realtime. Any ideas how to fix TAA ghosting since we don't know how to provide motion vectors?

A new look on our Terrain Editor/Generator. Everything is computed on GPU so we were surprised that even when the whole terrain is recomputed from scratch, it is still realtime. Any ideas how to fix TAA ghosting since we don't know how to provide motion vectors?

bonzaj/plastic

14,698 просмотров • 1 месяц назад

This is a new diffusion-based language model from Inception. I don't know how well it works compared to an autoregressive LM of a similar size, but the inference process looks badass:

This is a new diffusion-based language model from Inception. I don't know how well it works compared to an autoregressive LM of a similar size, but the inference process looks badass:

BURKOV

58,434 просмотров • 1 год назад

How good a 2D rig can look? Darcy and Jarrod created this rough animation in Moho and then used vectors and brushes to transform it into an impressive live painting! “I’m impressed with just how much you can manipulate the vectors in Moho. The best part is it's fun!”

How good a 2D rig can look? Darcy and Jarrod created this rough animation in Moho and then used vectors and brushes to transform it into an impressive live painting! “I’m impressed with just how much you can manipulate the vectors in Moho. The best part is it's fun!”

Moho Animation Software

26,243 просмотров • 1 год назад

Introducing MirageLSD: The First Live-Stream Diffusion (LSD) AI Model Input any video stream, from a camera or video chat to a computer screen or game, and transform it into any world you desire, in real-time (<40ms latency). Here’s how it works (w/ demo you can use!):

Introducing MirageLSD: The First Live-Stream Diffusion (LSD) AI Model Input any video stream, from a camera or video chat to a computer screen or game, and transform it into any world you desire, in real-time (<40ms latency). Here’s how it works (w/ demo you can use!):

Decart

1,010,228 просмотров • 1 год назад

Matryoshka Representation Learning (MRL) is a super exciting approach to improving the quality and efficiency of embedding models and strategies ✨ MRL allows models to store more information in the earlier dimensions of their data vectors. This method not only boosts performance in tasks like classification and retrieval, but is also a super cool compression technique! Paper: For compression: It’s been so much fun learning and making this video with @DanielW966, thanks for all your help!

Matryoshka Representation Learning (MRL) is a super exciting approach to improving the quality and efficiency of embedding models and strategies ✨ MRL allows models to store more information in the earlier dimensions of their data vectors. This method not only boosts performance in tasks like classification and retrieval, but is also a super cool compression technique! Paper: For compression: It’s been so much fun learning and making this video with @DanielW966, thanks for all your help!

Victoria Slocum

120,436 просмотров • 1 год назад

🚀 The team at Google DeepMind just released Gemini Embedding 2, a frontier embeddings model with 3072 dimensions and state-of-the-art semantic quality. 👩‍💻 We built a demo showing how to integrate it across the LlamaIndex ecosystem, from LlamaParse to LlamaAgents: 𝗮𝘂𝗱𝗶𝗼-𝗸𝗯, a knowledge base for your audio notes. With audio-kb, you can: 🔹 Upload an MP3 or record directly from your terminal 🔹 LlamaParse extracts the transcript from the audio 🔹 Gemini Embedding 2 generates embeddings 🔹 Metadata + vectors are stored in SurrealDB and indexed with HNSW 🔍 Once ingested, you can search all your audio notes directly from the terminal. 🎙️ Perfect for turning voice memos, meetings, or lectures into a searchable knowledge base. 📖 Full blog: 💻 GitHub: ⚡ Try LlamaParse:

🚀 The team at Google DeepMind just released Gemini Embedding 2, a frontier embeddings model with 3072 dimensions and state-of-the-art semantic quality. 👩‍💻 We built a demo showing how to integrate it across the LlamaIndex ecosystem, from LlamaParse to LlamaAgents: 𝗮𝘂𝗱𝗶𝗼-𝗸𝗯, a knowledge base for your audio notes. With audio-kb, you can: 🔹 Upload an MP3 or record directly from your terminal 🔹 LlamaParse extracts the transcript from the audio 🔹 Gemini Embedding 2 generates embeddings 🔹 Metadata + vectors are stored in SurrealDB and indexed with HNSW 🔍 Once ingested, you can search all your audio notes directly from the terminal. 🎙️ Perfect for turning voice memos, meetings, or lectures into a searchable knowledge base. 📖 Full blog: 💻 GitHub: ⚡ Try LlamaParse:

LlamaIndex 🦙

34,881 просмотров • 4 месяцев назад

👀 Pixel perfect 💎✨ 🖼️ Edify Image from #NVIDIAResearch is a family of diffusion models that supports a wide range of applications, including text-to-image synthesis, 4K upsampling, ControlNets, 360° HDR panorama generation, and finetuning for image customization. 🧵 1/2

👀 Pixel perfect 💎✨ 🖼️ Edify Image from #NVIDIAResearch is a family of diffusion models that supports a wide range of applications, including text-to-image synthesis, 4K upsampling, ControlNets, 360° HDR panorama generation, and finetuning for image customization. 🧵 1/2

NVIDIA AI Developer

14,747 просмотров • 1 год назад

Pi0 from Physical Intelligence is one of the best generalist Vision-Language-Action (VLA) models for robotics (and predecessor of Pi0.5). 🧵 Here's how it works + how you can fine-tune it on your own robot for a simple task:

Pi0 from Physical Intelligence is one of the best generalist Vision-Language-Action (VLA) models for robotics (and predecessor of Pi0.5). 🧵 Here's how it works + how you can fine-tune it on your own robot for a simple task:

Ilia

32,956 просмотров • 10 месяцев назад

I got a diffusion model to run really fast, glued it to a LLM, and it blows my mind how much of the interactivity we expect from software (editing text, dragging things around, clicking on "links") just works™️ in pixel space. The future of software is bright

I got a diffusion model to run really fast, glued it to a LLM, and it blows my mind how much of the interactivity we expect from software (editing text, dragging things around, clicking on "links") just works™️ in pixel space. The future of software is bright

Zain Shah

22,392 просмотров • 17 дней назад

The future of AI is open-source. And ollama is the easiest way to build AI applications with open-source LLMs. Here's how to build a free, private RAG app using open-source tools. We'll use: - Ollama for LLMs and embedding models - PostgreSQL for data storage and retrieval - pgai Vectorizer for embedding creation and sync (I use Nomic for embeddings and tinnyllama as my LLM but you can substitute them for any models on Ollama)

The future of AI is open-source. And ollama is the easiest way to build AI applications with open-source LLMs. Here's how to build a free, private RAG app using open-source tools. We'll use: - Ollama for LLMs and embedding models - PostgreSQL for data storage and retrieval - pgai Vectorizer for embedding creation and sync (I use Nomic for embeddings and tinnyllama as my LLM but you can substitute them for any models on Ollama)

Avthar

34,261 просмотров • 1 год назад

I seriously cannot believe this is a 0.6B LLM! 🤯 Qwen just released Qwen3, a series of hybrid reasoning models that allow you to control how much "thinking" the model does for a given task. They can even run locally in your browser on WebGPU with 🤗 Transformers.js!

I seriously cannot believe this is a 0.6B LLM! 🤯 Qwen just released Qwen3, a series of hybrid reasoning models that allow you to control how much "thinking" the model does for a given task. They can even run locally in your browser on WebGPU with 🤗 Transformers.js!

Xenova

74,954 просмотров • 1 год назад

This is a big day. Meta is open-sourcing AudioCraft. You can now generate incredible music and sounds with a single prompt. It includes the most performant Generative AI Model (audio) on the market, the "Llama" of Audio. The research framework contains the weights and code of these models: ▸ MusicGen: controllable text-to-music model. ▸ AudioGen: text-to-sound model. ▸ EnCodec: high fidelity neural audio codec. ▸ Multi Band Diffusion: An EnCodec compatible decoder using diffusion. This is going to tremendously speed up audio research 👏

This is a big day. Meta is open-sourcing AudioCraft. You can now generate incredible music and sounds with a single prompt. It includes the most performant Generative AI Model (audio) on the market, the "Llama" of Audio. The research framework contains the weights and code of these models: ▸ MusicGen: controllable text-to-music model. ▸ AudioGen: text-to-sound model. ▸ EnCodec: high fidelity neural audio codec. ▸ Multi Band Diffusion: An EnCodec compatible decoder using diffusion. This is going to tremendously speed up audio research 👏

Lior Alexander

231,622 просмотров • 3 лет назад

We're incredibly excited to showcase our partnership with IBM and the true power of 🤝 We've made embedding a persistent, on-chain 3D avatar into any platform as easy as embedding a YouTube video. 🌐 Frictionless deployment to onboard the next millions of developers and users 💻 Open-source AI infrastructure 🔗 Web3 meets enterprise-grade scale Every AI needs a body, brain (LLM memory), a wallet (jobs with x402) and every digital identity needs a network. The foundation is built, the partnerships are live, and the $THREE ecosystem is powering it all. Don't just watch the internet evolve. Build it.

three.ws

143,506 просмотров • 1 месяц назад

BREAKING: a new Chinese model is out. And no. This is NOT DeepSeek. Meet Qwen-2.5, from the giant Alibaba: 1. It can code, write text, search the web. 2. It can generate images, like Dall-E. 3. It can even generate videos. Here's everything you need to know:

BREAKING: a new Chinese model is out. And no. This is NOT DeepSeek. Meet Qwen-2.5, from the giant Alibaba: 1. It can code, write text, search the web. 2. It can generate images, like Dall-E. 3. It can even generate videos. Here's everything you need to know:

Ruben Hassid

363,505 просмотров • 1 год назад

LLM-grounded Diffusion: Enhancing Prompt Understanding of Text-to-Image Diffusion Models with Large Language Models paper page: github: Recent advancements in text-to-image generation with diffusion models have yielded remarkable results synthesizing highly realistic and diverse images. However, these models still encounter difficulties when generating images from prompts that demand spatial or common sense reasoning. We propose to equip diffusion models with enhanced reasoning capabilities by using off-the-shelf pretrained large language models (LLMs) in a novel two-stage generation process. First, we adapt an LLM to be a text-guided layout generator through in-context learning. When provided with an image prompt, an LLM outputs a scene layout in the form of bounding boxes along with corresponding individual descriptions. Second, we steer a diffusion model with a novel controller to generate images conditioned on the layout. Both stages utilize frozen pretrained models without any LLM or diffusion model parameter optimization. We validate the superiority of our design by demonstrating its ability to outperform the base diffusion model in accurately generating images according to prompts that necessitate both language and spatial reasoning. Additionally, our method naturally allows dialog-based scene specification and is able to handle prompts in a language that is not well-supported by the underlying diffusion model.

LLM-grounded Diffusion: Enhancing Prompt Understanding of Text-to-Image Diffusion Models with Large Language Models paper page: github: Recent advancements in text-to-image generation with diffusion models have yielded remarkable results synthesizing highly realistic and diverse images. However, these models still encounter difficulties when generating images from prompts that demand spatial or common sense reasoning. We propose to equip diffusion models with enhanced reasoning capabilities by using off-the-shelf pretrained large language models (LLMs) in a novel two-stage generation process. First, we adapt an LLM to be a text-guided layout generator through in-context learning. When provided with an image prompt, an LLM outputs a scene layout in the form of bounding boxes along with corresponding individual descriptions. Second, we steer a diffusion model with a novel controller to generate images conditioned on the layout. Both stages utilize frozen pretrained models without any LLM or diffusion model parameter optimization. We validate the superiority of our design by demonstrating its ability to outperform the base diffusion model in accurately generating images according to prompts that necessitate both language and spatial reasoning. Additionally, our method naturally allows dialog-based scene specification and is able to handle prompts in a language that is not well-supported by the underlying diffusion model.

AK

83,657 просмотров • 3 лет назад