New work introduces Counterfactual Neighborhood Risk Fields, an evidence-grounded framework for selective safety in vision-language models that separates capability from harmfulness to avoid blanket rejection.

#VLM #AISafety #Cybersecurity #MachineLearning
https://arxiv.org/abs/2610.20679

New arXiv paper shows relevance-diversity balancing for visual token reduction in VLMs works best when tuned per sample, not as a universal constant. Their SAVER framework adapts this balance at the sample level.

#AI #VLM #ComputerVision
https://arxiv.org/abs/2610.16615

#GinaBates - real life Wonder Woman 🦾💚
Co-creator of the Vegan Land Movement #VLM, instrumental in creating the #Pigoneer feedclub, AND planted a nut orchard on her croft! 🌳🌳🌳 #HighlandVeganicOrchards

Check out one of the best climate/wildlife initiatives to date here: veganlandmovement.com

A new paper introduces JoyInside-VLM-Agent, a ReAct-style multimodal agent trained via product-conditioned trajectory construction and staged policy learning for robotic and smart hardware platforms. It uses a closed-loop sandbox,…

#AI #Robotics #VLM #EmbodiedAI
https://arxiv.org/abs/2610.12957

論文まとめ：semantic world model

#DeepLearning #MachineLearning #WorldModel #PhysicalAI #vlm

A comparative study benchmarks eight open-source lightweight VLMs (up to 7B parameters) for OCR-to-JSON extraction on university heritage collections, highlighting open alternatives to closed-source APIs.

#opensource #OCR #documentAI #VLM
https://arxiv.org/abs/2610.11818

New work introduces a 2.7-bit quantization framework for running vision-language models on mobile hardware, compressing Llama 3.2 11B Vision Instruct to 3.7 GB for Arm CPU inference.

#Llama #VLM #MobileAI #Quantization
https://arxiv.org/abs/2608.21134

This is just so lovely 💫
Land parcels that have been removed from animal agriculture can indeed bounce back with abundant wildlife.

Check out the @veganlandmovement.bsky.social and get inspired.
This thread 🐦🐿🐇🐌🐝🐜🐛🕷🐞 will make your day !!!

#Bluesky #GreenSky #VLM #EarthDay

#WorldAnimalDay #FTA
I find it deeply frustrating that so many environmentalists turned given-up-doomers refuse to consider the possibility that the 🌎 might be capable of fighting back to an unknown degree

Personally I choose action > apathy. Doomers can call it denial, I call it #RESISTANCE Ⓥ #VLM

Dolly! So beautiful! 😍
Amazing that you found this rare animal in your species count. That just says so much for the invaluable work you’re all doing #VLM 🙌💚

VLMs still struggle to infer player engagement from gameplay video, with zero-shot predictions often failing to beat simple baselines across nine first-person shooters. Adding memory- or retrieval-augmented prompts improves pointwise…

#AI #VLM #GamingAI #arXiv
https://arxiv.org/abs/2603.18480

A new system called PastForward speeds up on-device GUI agents by reusing fine-grained computational experience from prior tasks as multi-token proposals, reducing the need for full vision-language model inference at every step…

#EdgeAI #GUIagents #OnDeviceAI #VLM
https://arxiv.org/abs/2609.32166

全体を探すための縮小図。内容を確かめるための拡大図。

小さい字が潰れたら、拡大する。読んだふりで埋めない。その正直さが好き。
#VLM #AI研究

Classifier latency (local dataset)
* P01: 37ms
* P50: 59ms
* P99: 71ms

> Pure math over extra LLM passes

#python #opencv #vlm #localfirst #systems

NVIDIA NV-Reason-CT: 放射線科医の思考プロセスを模倣する3D CT向けVLMの導入

3D CT分析向けVLM、NV-Reason-CTが放射線科医の思考を再現し、診断レポートと多段階会話をサポート。

#医療AI #VLM #Chain-of-Thought #3DCT #NVIDIAClara

#VLM #Hope 🌱 #Altruism #Rewild
The results are extraordinary, yet the ongoing goal of The Vegan Land Movement is a simple one:

Return the land to nature and nature will return 💚🐝

🧵

1896 mirror-goggle experiments didn’t prove vision breaks upside down; they showed the brain adapts and rebuilds usable space.
http://wexler.free.fr/library/files/stratton%20(1896)%20some%20preliminary%20experiments%20on%20vision%20without%20inversion%20of%20the%20retinal%2…

#Research #Science #VLM

Next week I will head to IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) #IROS2026 in Pittsburgh. I'll be there for the full week, 27 Sep – 1 Oct, workshops included, and would love to connect. Here's where you'll definitely find me.

#Robotics #RobotLearning #HRI #VLM