Filter
Exclude
Time range
-
Near
You talk, I design. Because sometimes… words just aren’t pretty enough. 😎🎨 #GraphicDesigner #DesignHumor #VisualThinker
6
14
97
VisualThinker-R1-Zero R1-Zero's Aha Moment on just a 2B non-SFT Model VisualThinker-R1-Zero is a replication of DeepSeek-R1-Zero in visual reasoning. Successfully observe the emergent “aha moment” and increased response length in visual reasoning on just a 2B non-SFT models
4
9
7,264
Replying to @Yuchenj_UW
depends on how you view the "aha" moment! for the R1-V, our "aha" moment was the performance increase the visualThinker-R1's "aha" moment was the increase in response length
4
129
5 Mar 2025
Replying to @Yuchenj_UW
@AskPerplexity why did VisualThinker-R1-Zero not pick up reasoning with RL with a fine-tuned model but worked for the base model?
2
6
755
5 Mar 2025
“Aha moment” for multimodal reasoning on just a 2B model by RL! VisualThinker-R1-Zero authors find applying RL to fine-tuned vision-language models didn't replicate Deepseek-R1’s reasoning but directly applying RL training to the Qwen2-VL-2B base model worked! (similar to DeepSeek-R1-Zero) Without any SFT but only RL, the model achieves 59.47% accuracy on CVBench, beating the base model for ~30%. It's exciting to see RL just works for multimodal!
13
54
404
32,128
🚀 VisualThinker-R1-Zero: A Breakthrough in Visual Reasoning! 🎯 R1-Zero replicates DeepSeek-R1-Zero for visual tasks, showcasing: ✅ The emergent "aha moment" ✨ ✅ Longer, more detailed responses 📝 ✅ Achieved with just a 2B non-SFT model 🤖 Exciting progress in AI reasoning! 🔥 #AI #MachineLearning
5
13
969
4 Mar 2025
VisualThinker-R1-Zero R1-Zero's Aha Moment on just a 2B non-SFT Model VisualThinker-R1-Zero is a replication of DeepSeek-R1-Zero in visual reasoning. Successfully observe the emergent “aha moment” and increased response length in visual reasoning on just a 2B non-SFT models
11
58
326
32,255
🚨Exciting news from our @TurningPointAI team: the very first "aha moment" on multimodal reasoning during RL on a 2B base (non-instruct) model! 📎Blog (observations): turningpointai.notion.site/t… 📎Code for our VisualThinker-Zero-2B: github.com/turningpoint-ai/V… Key findings👇
🚀 We’re excited to share our latest work! Welcome to the first successful "aha moment" on multimodal reasoning. "Aha moment" is featured by improved response length & performance. It emerges during RL of an unaligned base model on multimodal tasks. Aha moment for language reasoning was originally observed on DeepSeek-R1-Zero. 🔍 Key Findings: 1. Directly applying GRPO on an unaligned 2B base model could elicit the multimodal “aha moment”: thinking capability marked by spontaneous reasoning strategy and increased reasoning length 2. Visual-centric task could benefit from long Chain-of-Thoughts 💻 Discover more on our notion blog and project page! Detailed Research Blog: Follow our complete journey and technical insights at our Notion Blog: 🔗turningpointai.notion.site/t… Reproduce Our Results: Access and build upon our implementation at GitHub: 🔗github.com/turningpoint-ai/V… Presented by: TurningPointAI Team 🔗turningpoint-ai.com/ #turningpointai #Smallmodel #MultimodalR1 #DeepseekR1 #R1 #Deepseek #AI #MultimodalReasoning #Qwen #QwenVL #DeepSeekR1zero
11
845
Alright, fellow visionaries and daydream doodlers, let's talk about how AI has been the ultimate wingman for us visual folks. I used to have these wild, vivid images swirling in my head like a cosmic art gallery, but bringing them to life? That was like trying to catch a rainbow with a butterfly net. Enter AI, stage left, with its magic wand of pixels and algorithms. Now, those pictures in my head aren't just fleeting visions; they're real, tangible artworks. With a few prompts, AI can take the abstract mess of colors and shapes from my mind and turn it into something you can actually see, touch, and share. It's like having a personal artist that never gets tired, never runs out of ideas, and can work faster than you can say 'Dali meets digital.' Whether it's for my next big project, a comic strip, or just to see if that dragon I imagined looks as fierce on screen as it does in my head, AI has got my back. No more struggling with brushes or cursing at photoshop layers; AI has made my inner world an outer reality. It's a game-changer for anyone whose brain works in pictures rather than paragraphs. So here's to AI, the silent partner in crime for all us visual thinkers. Now, if you'll excuse me, I've got some mind-boggling landscapes to materialize. #VisualThinker #AIFTW #MindToMatter
2
3
137
@HamzaYassin3: I gently asked R why he didn’t answer your 1st Q (I knew he knew): He said because there is no room above you when you 1st come in our front door (we have a wee porch) & he just didn’t want to sound cheeky. 😂 I totally did not visualise the porch! #visualthinker
2
3
179
What, this isn’t how you plan *your* semester?? #visualthinker
3
205
Today in the US it’s a big one, it’s #nationalferretday #illustration #scribe #visualthinker
1
22
61
2,451
16 Mar 2024
Caricature illustration of my client who became a very good friend. I also did some live caricatures at their wedding and they loved it. #cartoon #art #illustrationartists #cartoonist #caricature #visualthinker #visualcommunication
2
1
24
662
ayo share feed kamu & cerita kamu! - aku \VisualThinker, yg punya hobi dan bercita-cita menjadi fotografer. awal mula kenal fotografi dari ibu karena beliau juga mempunyai hobi yg sama saat muda. Saat ini sedang suka foto stage walaupun genre yg didalami lebih ke portrait😭
5
1
8
1,501
Y como siempre con #EQUIPAZOTOP que irá acompañando en los procesos creativos. Tenemos al #visualthinker @miguelpascual76; al #storytellerDelSentidoComún @llume38 y a la gran #educainfluencer @NovoaCris...Otro gran reto, y esta vez...¡estaré con vosotros!
1
2
452