john yue

Hello，I‘m a new guys and I came from China . If you like Science ,food ,or Video Games ,travel and so on .Let me know you ,We can be good friends.

[Wildminder @wildmindai]

Another cool stuff from NVIDIA.
LocateAnything - high-speed visual search engine. You provide a text prompt and it instantly pinpoints that object's exact location in an image.

- 10x speedup for dense object detection
- Qwen2.5-3B + Moon-ViT
- Fast/Slow/Hybrid modes
- trained on 138M samples for UI, docs, generic grounding.

https://t.co/bEvD6pRKaR

[Wildminder @wildmindai]

SUPIR upscaler is outdated. ASASR- turns blurry, low-quality photos into sharp, high-res images. Prevents the fake hallucinated details.

- improves OCR
- high segmentation accuracy
- based on FLUX.1 dev

this looks sweet

https://t.co/kx4A4N1wDc

[john yue @Luck2john]

@ljsabc 期待下周

[向阳乔木 @vista8]

阿里开源的OCR解析模型 Logics-Parsing V2好像很厉害。

连乐谱都能准确识别了，还有流程图、思维导图、代码和伪代码。

github: https://t.co/PmQ9cH8fvw

demo: https://t.co/TBwwRX5ZAO

模型地址：https://t.co/bcumWO1PyB

[Sayak Paul @RisingSayak]

Last year, I got to collaborate on a number of serious projects at the intersection of Diffusers x optimization ⚡️

First, NONE of them were bootstrapped with any AI agents but pure domain knowledge and expertise. So, besides just feeling good, it's also very reassuring to me to know how important those two traits are.

Now, coming to the projects that I think are worth mentioning:

* `flux-fast`: Showing a combination of `torch.compile` + unscaled FP8 FA3 + no CPU-GPU sync + dynamic FP8 is great for accelerating Flux.1-*. https://t.co/Fagw9bkFPA

* `torch.compile` x Diffusers: What does it take to get the most out of `torch.compile` in Diffusers across different user workloads? https://t.co/J8bPgBFK1y

* `lora-fast`: How to hotswap LoRAs into compiled models without incurring (slow) recompilation issues? How to set it up for success? https://t.co/FhY8ATz4c0

* `zerogpu-brrr`: How to optimize a ZeroGPU HF Space with AOT + FA3 and other goodies? This helps save 💰 and improve the user experience of your ZeroGPU applications. https://t.co/DdPsS6O5Ky

Hopefully, this will make you realize there's still a LOT that you can do (preferably pairing with AI) if you're curious and deeply invested in stuff you care about.

[Gorden Sun @Gorden_Sun]

daVinci-MagiHuman：开源音视频模型
由国内GAIR实验室发布，15B参数，能生成带有音频的视频（目前开源可用的同类模型只有LTX 2.3），支持中文、英文、日语等多种语言。效果还行，感觉生成的视频动作幅度不大。
在线体验：https://t.co/xsYdDP2QvA
模型：https://t.co/tYK1VRWxaV

[Victor M @victormustar]

Go try it: https://t.co/cbV5BGrqTu

[angrizan @chainsmoker89]

Track 1 : AI for Production

Letter Lora : Qwen Image Edit 2511

A custom trained LoRA that reverses engineers typography from real world photographs.

Lora Link : https://t.co/DvB46WCO4F

@Ali_TongyiLab
@Alibaba_Qwen
@ModelScope2022

#HappyQwensDay
#QwenImageLora

[Wildminder @wildmindai]

EffectMaker by Tencent.
Wan2.2 + Qwen3-VL does zero-shot effect transfer from reference videos to images without per-effect LoRAs.

> clones complex video effects to static images
> outperforms VFX-Creator and Omni-Effect .

professional-grade cinematic effects for film post-production, game design etc.
Just point at a reference and it clones the physics.

https://t.co/XeOAoA6ekP

[Tengfei Wang @DylanTFWang]

Autoregressive diffusion models drift for long videos? 📉
We fixed it.🚀 Speed + Stability = ✅

Meeting *Test-Time Correction (TTC)*. We stop error accumulation in its tracks without any retraining.
✅ Training-free
✅ 1 minute+ stable generation
✅ Negligible overhead

[Huaxiu Yao @HuaxiuYaoML]

Meet MetaClaw 🦞— Just talk to your agent, it learns and evolves.

💬 Conversations become training trajectories
⚡ Models update live with hot-swapped weights
🧠 Failures generate new reusable skills
💻 No GPU cluster required

Under the hood:
🔄 Online SkillRL training
Training runs asynchronously while the model continues serving.

🐚 Skill evolution
When the agent fails, an LLM analyzes the trajectory and generates new reusable skills.

📌 Skill injection
Relevant skills are retrieved and injected into the system prompt at each step to guide behavior in real time.

Built on Kimi-2.5 via Tinker cloud LoRA
→ fine-tuning costs about $10 and requires no GPU cluster

💻 Fully open-sourced
https://t.co/RR24ZJvZ7T

Built with @openclaw and @thinkymachines

Kudos to the team @richardxp888, Jianwen Chen, @Xinyu2ML, @lillianwei423, @StephenQS0710, Zeyu Zheng, @cihangxie!

[AK @_akhaliq]

MatAnyone 2 is out on Hugging Face

Scaling Video Matting via a Learned Quality Evaluator

paper: https://t.co/KPMaG8teJ2
app: https://t.co/wkMpaOdoCh

[PaddlePaddle @PaddlePaddle]

🔥 PaddleOCR-VL is now available in the llama.cpp ecosystem

This brings document parsing VLMs closer to local and lightweight deployment workflows — making it easier for developers to explore portable, community-friendly multimodal document AI.

Why it matters
🔹 Easier access to PaddleOCR-VL in GGUF-based workflows
🔹 More flexible paths for local inference and lightweight deployment
🔹 A simpler way to experiment with structured document understanding beyond traditional OCR stacks

⚠️ Important note for developers
If llama-cli works but llama-server throws errors, try explicitly passing --chat-template-file when launching llama-server.

Chat template / GGUF resources
👉 PaddleOCR-VL-1.5-GGUF: https://t.co/F8qxYL0b3X
👉 PaddleOCR-VL-GGUF: https://t.co/BLorWbajVl

A meaningful step toward making document intelligence more open, more portable, and more developer-ready.

#PaddleOCR #llamacpp #GGUF #DocumentAI #MultimodalAI #OCR #OpenSourceAI

[Machine Delusions @Machinedelusion]

⏰Released the PixelGen Repo for @ComfyUI

This is a pixel diffusion model. No Vae! All diffused in pixel space!

LoRa training nodes included as well. Link Below👇

[Michael Yuan @juntao]

Rust implementation for Speech-to-Text based on open-source Qwen3 models

* Self-contained binary build — no external dependencies
* Uses libtorch on Linux with optional Nvidia GPU support
* Uses MLX on MacOS with Apple GPU/NPU support

🔨 CLI for AI agents and humans: https://t.co/knsZlastgQ

🖥️ OpenAI compatible API server: https://t.co/qjDqCf9hor

🤖 OpenClaw skill: https://t.co/tE6lzTjYpy

Why and how
https://t.co/VxRt9oSZ8a

[Ling Yang @LingYang_PU]

What if your AI agent got better just by talking to you?
Introducing OpenClaw-RL — a fully async RL framework that turns your everyday conversations into training signals. Your agent learns your habits, your workflows, your preferences. Privately. Continuously. #Clawdbot #openclaw

🔑 Two learning modes:
• Binary RL — likes/dislikes become rewards
• On-Policy Distillation — your textual feedback becomes token-level guidance

Self-hosted. Zero API keys. Your data never leaves your machine.
👉 https://t.co/ry18qekutm

[Niels Rogge @NielsRogge]

Opus 4.6 made a @Gradio demo for it too!

It uses a "chunked window" approach, allowing it to run up to 160 frames per second (FPS)

I'm really impressed by coding agents - porting a model + demo in less than 2 days

[Sayak Paul @RisingSayak]

Editing images is a series of state transitions between the source image and the edited image that we want. Yet, the existing paradigm doesn't explicitly include any transitioning priors in the editing process.

This becomes particularly prevalent for edits, involving causal dynamics (e.g., refraction, deformation).

To model this kind of physics-informed information, we leverage the rich priors present in videos and introduce PhysicEdit 🔥

TL;DR: We fine-tune QwenImage Edit on a curated dataset of videos with reasoning traces and fixed-length transition queries to do solid physics-aware image editing!

In the process, we introduce a cool dataset "PhysicTran38K", consisting of 38K transition trajectories across five physical domains and devise a method to provide supervision from it QwenImage Edit.

Hop in to learn more ⬇️

[liz tan @liztansz]

Say you have trained your deep learning model. It works. But do you know what it has actually learned?

🚀 We’ve built SymTorch: a library that translates deep learning models into human-readable equations.

I've attached here a quick video demonstrating how SymTorch works.

[Sayak Paul @RisingSayak]

If you stuck around till here, thank you!

Please check out the work here: https://t.co/tRpiaC9Xed. We're also open-sourcing everything -- checkpoints, code, and the dataset ❤️

Best collaborators ever: @ben_nebulous, @zhuole1025, and others!