mudler/Qwen3.6-35B-A3B-APEX-GGUF
35B • Updated • 83.7k • 69
Henry Kroll III
hellork
AI & ML interests
Small, open-source LLMs for mobile with <= 4Gb graphics, text generation, image manipulation recognition, 3D model generation, language preservation and translation, and business use cases.
Recent Activity
liked a model 2 days ago
mudler/Qwen3.6-35B-A3B-APEX-GGUF updated a collection 10 months ago
Under 10b iq4_nl gguf updated a collection 10 months ago
Under 10b iq4_nl ggufOrganizations
liked a
model 2 days ago
updated a
collection 10 months ago
reacted to andito's post with 🔥 10 months ago
Post
4096
🧠👁️ Can AI visualize solutions?
Humans often solve visual problems by sketching ideas in our minds. What if Vision-Language Models (VLMs) could do something similar, not by generating full images, but by using internal “mental sketches”?
That’s the idea behind Mirage, a new framework that empowers VLMs to reason using latent visual tokens. Instead of just thinking in words, Mirage mixes in abstract visual representations that help the model solve complex tasks.
These aren't photorealistic images. They're compact, internal representations optimized purely to support reasoning.
🔧 Mirage is trained in two phases:
1) Grounding: It learns to produce latent tokens anchored in real images.
2) Refinement: The model drops the images and learns to generate visual tokens on its own.
📈 And yes, it works!
On challenging benchmarks like Visual Spatial Planning, Jigsaw puzzles, and Spatial Attention Tasks, Mirage clearly outperforms GPT-4o and other strong baselines.
Smart sketches > empty words.
By mimicking the way humans visualize solutions, Mirage gives AI a new kind of imagination, one that’s faster, more efficient, and more human-like.
Kudos to the teams at UMass Amherst and MIT behind this exciting work.
Check the paper: Machine Mental Imagery: Empower Multimodal Reasoning with Latent Visual Tokens (2506.17218)
Humans often solve visual problems by sketching ideas in our minds. What if Vision-Language Models (VLMs) could do something similar, not by generating full images, but by using internal “mental sketches”?
That’s the idea behind Mirage, a new framework that empowers VLMs to reason using latent visual tokens. Instead of just thinking in words, Mirage mixes in abstract visual representations that help the model solve complex tasks.
These aren't photorealistic images. They're compact, internal representations optimized purely to support reasoning.
🔧 Mirage is trained in two phases:
1) Grounding: It learns to produce latent tokens anchored in real images.
2) Refinement: The model drops the images and learns to generate visual tokens on its own.
📈 And yes, it works!
On challenging benchmarks like Visual Spatial Planning, Jigsaw puzzles, and Spatial Attention Tasks, Mirage clearly outperforms GPT-4o and other strong baselines.
Smart sketches > empty words.
By mimicking the way humans visualize solutions, Mirage gives AI a new kind of imagination, one that’s faster, more efficient, and more human-like.
Kudos to the teams at UMass Amherst and MIT behind this exciting work.
Check the paper: Machine Mental Imagery: Empower Multimodal Reasoning with Latent Visual Tokens (2506.17218)
- 4 replies
liked 2
models 10 months ago
Just wanted to say thanks Calcuis
😎👍 2
39
#1 opened 11 months ago
by 
Winnougan
updated a
collection 10 months ago
liked a
model 10 months ago
liked 2
Spaces 11 months ago
upvoted an article 11 months ago
Article
Making LLMs even more accessible with bitsandbytes, 4-bit quantization and QLoRA
- +3
•
177
updated a
Space 12 months ago
published a
Space 12 months ago
updated a
model 12 months ago
published a
model 12 months ago
updated a
model 12 months ago
published a
model 12 months ago
updated a
collection 12 months ago
updated a
model 12 months ago