正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

low quant weights make the embedding model lose all discriminative power. I plotted the cosine correlation matrix of jina-v5, and one can see that low quant makes the model really blind. The off-diagonal similarities are pretty high on Q1/2/3, meaning everything looks similar in the semantic space. Q4 is... show more

Han Xiao

17,450 subscribers

62,912 次观看 • 2 个月前 •via X (Twitter)

教育科学技术

Anya Rossi• Live Now

Private livecam show

0 条评论

暂无评论

原始帖子的评论将显示在这里

相关视频

This is the 1st new Model Y test video from China. According to the tester, the new Y looks better than Model 3 in person. Suspension is superior. Sound quality is top-notch. 2nd row seat is comfortable and is power operated. Noticeably better than the legacy Model Y. Are you ready to order one?

This is the 1st new Model Y test video from China. According to the tester, the new Y looks better than Model 3 in person. Suspension is superior. Sound quality is top-notch. 2nd row seat is comfortable and is power operated. Noticeably better than the legacy Model Y. Are you ready to order one?

Ray

201,178 次观看 • 1 年前

Same model psd for both Same rigger Left is 2022 and the one on the Right is 2025 a few days ago It's crazy how much a riggers can really change how a model looks LIKE THAT DOESN'T EVEN LOOK LIKE THE SAME MODEL, BUT IT IS!!

Same model psd for both Same rigger Left is 2022 and the one on the Right is 2025 a few days ago It's crazy how much a riggers can really change how a model looks LIKE THAT DOESN'T EVEN LOOK LIKE THE SAME MODEL, BUT IT IS!!

💙💉 Khione || Vtuber

86,645 次观看 • 11 个月前

NEW: Google releases EmbeddingGemma, a state-of-the-art multilingual embedding model perfect for on-device use cases! At only 308M params, the model can run 100% locally in your browser! 🤯 Explore your documents in an interactive 3D universe with our demo: "The Semantic Galaxy"

NEW: Google releases EmbeddingGemma, a state-of-the-art multilingual embedding model perfect for on-device use cases! At only 308M params, the model can run 100% locally in your browser! 🤯 Explore your documents in an interactive 3D universe with our demo: "The Semantic Galaxy"

Xenova

23,338 次观看 • 9 个月前

Ever wondered what neural networks are and how they work? Systems like ChatGPT use neural networks to work as well as they do. Neural networks are composed of "layers" of neurons, layers with different functions; connections between layers called "weights"; and mathematical functions called "activation functions". If you’re interested in learning about these systems, check the comments. Ultimately, the neural network structure of the model serves to visually demonstrate that it is, in fact, a complex mathematical equation. When companies release the model's weights, they are releasing a key component needed to run the model's complete equation. Without the weights, the equation is incomplete. For the math-minded: the weights of a model are the learned numbers (they are variables during training) that are then used as constants in the mathematical functions that make up the model. Neural networks are ultimately just one big, hyper-complex mathematical function, and when a model is trained, it learns the constants associated with the high-dimensional input.

Ever wondered what neural networks are and how they work? Systems like ChatGPT use neural networks to work as well as they do. Neural networks are composed of "layers" of neurons, layers with different functions; connections between layers called "weights"; and mathematical functions called "activation functions". If you’re interested in learning about these systems, check the comments. Ultimately, the neural network structure of the model serves to visually demonstrate that it is, in fact, a complex mathematical equation. When companies release the model's weights, they are releasing a key component needed to run the model's complete equation. Without the weights, the equation is incomplete. For the math-minded: the weights of a model are the learned numbers (they are variables during training) that are then used as constants in the mathematical functions that make up the model. Neural networks are ultimately just one big, hyper-complex mathematical function, and when a model is trained, it learns the constants associated with the high-dimensional input.

Harper Carroll

26,930 次观看 • 7 个月前

i have been testing with my models a lot customization and sounds, and i figure out how to put low hp themes! jump sounds and put more than 3 clothes/skin color in one single model, here all the videos even with the themes of the 3 models that have the low hp theme for now

i have been testing with my models a lot customization and sounds, and i figure out how to put low hp themes! jump sounds and put more than 3 clothes/skin color in one single model, here all the videos even with the themes of the 3 models that have the low hp theme for now

samurott lover #1

15,685 次观看 • 3 个月前

Hong Kong has officially become "Tesla-land" 🇭🇰 Standing on a street corner in Central, you can witness the brand’s entire evolution in a single traffic light cycle. It is one of the few places on earth where you see total saturation across the board—from the high-end luxury of the Model S and the iconic falcon-wing doors of the Model X to the massive waves of Model 3s and Model Ys that have become the city’s new standard. It’s no longer just a trend; it’s the city's new automotive DNA. It’s the Octopus card of the road.

Ming

17,500 次观看 • 1 个月前

NEWS: The Tesla Model 3 has won Edmunds Top Rated Electric Car award for 2026. "The Model 3 is quite simply the best small affordable EV on the market. This is the best Model 3 yet. It is impressively well rounded thanks to improved build quality, ride comfort, and a compelling combination of efficiency, performance and value." "The Model 3 offers just about the perfect combination of everything — speed, range, comfort, space, tech, accessibility and convenience. It's a no-brainer if you want a sensible EV." - Jonathan Elfalan, director of vehicle testing.

NEWS: The Tesla Model 3 has won Edmunds Top Rated Electric Car award for 2026. "The Model 3 is quite simply the best small affordable EV on the market. This is the best Model 3 yet. It is impressively well rounded thanks to improved build quality, ride comfort, and a compelling combination of efficiency, performance and value." "The Model 3 offers just about the perfect combination of everything — speed, range, comfort, space, tech, accessibility and convenience. It's a no-brainer if you want a sensible EV." - Jonathan Elfalan, director of vehicle testing.

Sawyer Merritt

94,682 次观看 • 3 个月前

The Cybertruck is now the 3rd best-selling EV in the US for Q3. This makes the top 3 selling EVs in the United States all Teslas! #1 Model Y #2 Model 3 #3 Cybertruck

The Cybertruck is now the 3rd best-selling EV in the US for Q3. This makes the top 3 selling EVs in the United States all Teslas! #1 Model Y #2 Model 3 #3 Cybertruck

Teslaconomics

166,101 次观看 • 1 年前

80% of Most traders are asleep when the real setup forms. The 5AM window is where CRT quietly builds the day’s opportunity. In this 10-minute lecture, you’ll see: 5AM CRT Model: High Probability CRT Model (with ICT concepts) You’ll learn: • Why the 5AM candle is critical to the day’s range • How CRT defines the true high/low from this window • Where ICT concepts align with the setup • The confirmation that signals expansion • The mistake traders make before New York opens This isn’t a random time. It’s a repeatable window where structure is formed. If you trade CRT and ICT this model will change how you frame the morning. 🎥 10-minute in-depth breakdown attached. More CRT MODEL to be posted so be patient and wait for the remaining and chose the best that suits you.

80% of Most traders are asleep when the real setup forms. The 5AM window is where CRT quietly builds the day’s opportunity. In this 10-minute lecture, you’ll see: 5AM CRT Model: High Probability CRT Model (with ICT concepts) You’ll learn: • Why the 5AM candle is critical to the day’s range • How CRT defines the true high/low from this window • Where ICT concepts align with the setup • The confirmation that signals expansion • The mistake traders make before New York opens This isn’t a random time. It’s a repeatable window where structure is formed. If you trade CRT and ICT this model will change how you frame the morning. 🎥 10-minute in-depth breakdown attached. More CRT MODEL to be posted so be patient and wait for the remaining and chose the best that suits you.

R1X TRADES (GER40)

27,397 次观看 • 1 个月前

AI pricing is a function of 2 variables: agent autonomy & value attribution • Low autonomy, low attribution → seat-based Slack, early GitHub • Low autonomy, high attribution → hybrid (seats + credits) Cursor, Clay • High autonomy, low attribution → usage-based Twilio • High autonomy, high attribution → outcome-based FinAI, Chargeflow Pick the model that fits your product today, but build the pathway toward the top right

AI pricing is a function of 2 variables: agent autonomy & value attribution • Low autonomy, low attribution → seat-based Slack, early GitHub • Low autonomy, high attribution → hybrid (seats + credits) Cursor, Clay • High autonomy, low attribution → usage-based Twilio • High autonomy, high attribution → outcome-based FinAI, Chargeflow Pick the model that fits your product today, but build the pathway toward the top right

Dara

40,162 次观看 • 5 个月前

I tested the AI model Sora 2 on classic anime, the result is hardly believable... I can already see the hundreds of fanmades and parodies that are going to come out! Sora 2 is definitely a new step in AI anime..

I tested the AI model Sora 2 on classic anime, the result is hardly believable... I can already see the hundreds of fanmades and parodies that are going to come out! Sora 2 is definitely a new step in AI anime..

Naegiko

25,395,205 次观看 • 8 个月前

Shorekeeper Model feels like I am playing a 70$ game for FREE. Her model is so detailed that the amount of effort Kuro has put into the model design really shows off.

Shorekeeper Model feels like I am playing a 70$ game for FREE. Her model is so detailed that the amount of effort Kuro has put into the model design really shows off.

Naru

902,435 次观看 • 1 年前

🚨You can now use the new upcoming OpenAI model GPT 5.2 inside Cursor. Here is the full walkthrough. - Open the editor, go to settings and then the model tab. Add a custom model and enter the text "gpt-5.2-high" and "gpt-5.2". - After that you can select the model and ask questions. To verify, I started my test on the usage page which had zero gpt-5.2-high requests and consumption. After the test I could see the details in usage and the cost incurred while using it. Enjoy

🚨You can now use the new upcoming OpenAI model GPT 5.2 inside Cursor. Here is the full walkthrough. - Open the editor, go to settings and then the model tab. Add a custom model and enter the text "gpt-5.2-high" and "gpt-5.2". - After that you can select the model and ask questions. To verify, I started my test on the usage page which had zero gpt-5.2-high requests and consumption. After the test I could see the details in usage and the cost incurred while using it. Enjoy

AshutoshShrivastava

424,035 次观看 • 6 个月前

The new open-source Text to Speech model: Fish Speech 1.4 is brilliant! Trained on a massive 700K hours of multilingual speech data in 8 languages - Instant voice cloning 🗣️ - Ultra-low latency ⚡ - Compact model (~1GB weights) 🏋️‍♂️

The new open-source Text to Speech model: Fish Speech 1.4 is brilliant! Trained on a massive 700K hours of multilingual speech data in 8 languages - Instant voice cloning 🗣️ - Ultra-low latency ⚡ - Compact model (~1GB weights) 🏋️‍♂️

Rohan Paul

228,836 次观看 • 1 年前

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

21,211 次观看 • 7 个月前

NEW: Perplexity just released pplx-embed, a series of powerful embedding models. Inspired by their logo, I built a small semantic search demo: enter a query and the model retrieves the most relevant page. Everything runs locally in your browser with Transformers.js and Three.js

Xenova

11,092 次观看 • 3 个月前

✨ Made a new mini feature on Photo AI: [ Grab from 3d model ] So the problem is we're at that stage in time (typical for AI) where image-to-3d models are not good enough but are fun to play with, but we know they'll be good enough in 1-2 years With [ Make 3d model ] you already can turn any Photo AI pic into a 3d model but it still looks hyper clunky and deformed, but it works! One cool idea I had to make that more useful and made now: Let people make a 3d model then change the view of the it with the 3d viewer, then press [ o ] and it grabs a frame of the 3d That image you can then [ Remix ] (img2img), and it becomes a real photo again and that in turn you can then turn into a video again with [ Make video ] So that essentially gives you a fully freeform camera position control to take photos with One thing I need to fix is the background/skybox, I kinda need to take the original photo and remove the person and just get the background for the 3d model viewer, in this case it should be white, but it's a start!

✨ Made a new mini feature on Photo AI: [ Grab from 3d model ] So the problem is we're at that stage in time (typical for AI) where image-to-3d models are not good enough but are fun to play with, but we know they'll be good enough in 1-2 years With [ Make 3d model ] you already can turn any Photo AI pic into a 3d model but it still looks hyper clunky and deformed, but it works! One cool idea I had to make that more useful and made now: Let people make a 3d model then change the view of the it with the 3d viewer, then press [ o ] and it grabs a frame of the 3d That image you can then [ Remix ] (img2img), and it becomes a real photo again and that in turn you can then turn into a video again with [ Make video ] So that essentially gives you a fully freeform camera position control to take photos with One thing I need to fix is the background/skybox, I kinda need to take the original photo and remove the person and just get the background for the 3d model viewer, in this case it should be white, but it's a start!

@levelsio

119,210 次观看 • 11 个月前

In business, balance is everything. 0il at $78 a barrel isn't just a number, it's the sweet spot where profits flow and people can still breathe. Too low, the industry collapses. Too high, the market rebels. The real game isn't chasing extremes, it's mastering the middle.

In business, balance is everything. 0il at $78 a barrel isn't just a number, it's the sweet spot where profits flow and people can still breathe. Too low, the industry collapses. Too high, the market rebels. The real game isn't chasing extremes, it's mastering the middle.

Bigjhay!

208,842 次观看 • 1 个月前

Dump all the images, videos, text, and audio data that you have into Supabase (storage + database), and make them searchable using the new Gemini Embedding 2 model! I know we all miss him a lot!

Dump all the images, videos, text, and audio data that you have into Supabase (storage + database), and make them searchable using the new Gemini Embedding 2 model! I know we all miss him a lot!

Tyler Shukert

11,446 次观看 • 3 个月前

I've seen so many people criticize the stock sonic model and say the model needs changing. But no, the model is fine. the problem is with how it's always posed. here for example with the unleashed model you can simply deform the brows and the muzzle to curve. This is pretty much the sonic unleashed game model but with a good rig that makes anything possible with the face. (Rig by DANCADA³ᴰ)

I've seen so many people criticize the stock sonic model and say the model needs changing. But no, the model is fine. the problem is with how it's always posed. here for example with the unleashed model you can simply deform the brows and the muzzle to curve. This is pretty much the sonic unleashed game model but with a good rig that makes anything possible with the face. (Rig by DANCADA³ᴰ)

TBSF

192,128 次观看 • 1 年前