All in one Al Robot platform showcased at #IROS2025 #Astribot #VLA #teleoperation #Humaniod_Robot #AI_Robot #Research

Mixing vision and robotics is incredibly hard, but when it finally works it feels like magic.

Who wants coconut water? — Interruptions welcomed. 🧃🤖 #Galaxea #Robotics #EmbodiedAI #AI #Humanoid #Robot

I’m excited to present our #NeurIPS2025 Spotlight work in San Diego! Humans rely on sound when vision misleads—but AI doesn’t. We introduce a neuroscience-inspired model that resolves cross-modal conflicts for sound localization.👂Paper&code: arxiv.org/pdf/2505.11217 #AI #AudioAI

🚀 Step-Audio-EditX is now open source!! ✨ Zero-Shot TTS with high timbre similarity ✨ Iterative editing of dozens of audio emotion and speaking style ✨ Fine-grained control over paralinguistic features Whether for audio editing, interactive design, or personalized scenarios,

🚨 New paper alert! We’re thrilled to share our new preprint “PrefixNLI: Detecting Factual Inconsistencies as Soon as They Arise” ✨ LLMs generate text one token at a time, but factuality checks still wait for a full sentence. We extend NLI to text prefixes, enabling the

🚀 Thrilled to share our new work, "RLoop: A Self-Improving Framework for Reinforcement Learning"! arxiv.org/pdf/2511.04285

Can LLMs accurately aggregate information over long, information-dense texts? Not yet… We introduce Oolong, a dataset of simple-to-verify information aggregation questions over long inputs. No model achieves >50% accuracy at 128K on Oolong!

Kimi this Kimi that. Just Kimi the bottom line. The Kimi.ai K2 Thinking model has taken 2nd place on Vals Index for open source. It beat out Z.ai’s GLM 4.5, though GLM 4.6 is still holding strong and doesn’t look to be budging anytime soon Here’s what we found in our

MLLMs are great at understanding videos, but struggle with spatial reasoning—like estimating distances or tracking objects across time. the bottleneck? getting precise 3D spatial annotations on real videos is expensive and error-prone. introducing SIMS-V 🤖 [1/n]

Last week, @GitHub released the 2025 Octoverse report. Over 180 million developers contributed to more than a billion projects, and for the first time, TypeScript surpassed Python and JavaScript. Activity on GitHub is at a record high, with more contributors, repos, and

Here's your weekend challenge: Implement speculative decoding. Step 1: Read the following paper and/or blog: arxiv.org/abs/2211.17192 galacodes.hashnode.dev/speculative-de… (cc Jay Gala | Building Indilingo) Step 2: Choose a family of models which come in various sizes. My choice would be the Gemma3 or Qwen

EdgeTAM now supported by DRONEFORGE what should we do with this model and an autonomous drone?

🚀 Meet AgiBot A2—your new interactive star for business & fun! 69kg, 169cm, and packed with tech that crushes service + performance. ✨ 96% noise-resistant recognition (hears you even in crowds!) ✨ 2h runtime + quick battery swap (non-stop action!) ✨ 40+ DOF for human-like

This paper creates a realistic benchmark to test if AI agents can truly perform end to end LLM research. Agents need about 6.5x more run time here than older tests, which shows the tasks are harder. InnovatorBench packs 20 tasks across 6 areas, covering data work, loss design,

The paper introduces HaluMem to pinpoint where AI memory systems hallucinate and shows errors grow during extraction and updates. It supports 1M+ token contexts. HaluMem tests memory step by step instead of only final answers. It measures 3 stages, extraction, updating, and

I was asked whether Chinese robots can play Chinese Kungfu, just like how Chinese people are always asked the same questions. Answers are: Yes, they can. But, I can’t 🤣

New AI at Meta paper explains when a smaller, curated dataset beats using everything. Standard training wastes effort because many examples are redundant or wrong. They formalize a label generator, a pruning oracle, and a learner. From this, they derive exact error laws and

The slide for my Free Transformer talk: fleuret.org/public/slides-…

🚨 Google just proposed training AI in space. Their new paper, “Towards a Future Space-Based, Highly Scalable AI Infrastructure System,” explores building orbital ML data centers powered directly by the Sun fleets of satellites running TPUs, networked by laser links. Why?