Papers11

#embodied AI

From Perception to Action: An Interactive Benchmark for Vision Reasoning

The paper introduces CHAIN, a hands-on 3D playground that tests if AI can not only see objects but also plan and act under real physics.

#interactive benchmark#vision-language models#physical reasoning

Learning Situated Awareness in the Real World

Intermediate

Chuhan Li, Ruilin Han et al.Feb 18arXiv

SAW-Bench is a new test that checks if AI can understand the world from a first-person view, like wearing smart glasses.

#situated awareness#egocentric video#observer-centric reasoning

Theory of Space: Can Foundation Models Construct Spatial Beliefs through Active Exploration?

Intermediate

Pingyue Zhang, Zihan Huang et al.Feb 4arXiv

This paper asks a simple question with big consequences: can today’s AI models actively explore a new space and build a trustworthy internal map of it?

#active exploration#cognitive map#spatial belief

Research on World Models Is Not Merely Injecting World Knowledge into Specific Tasks

Intermediate

Bohan Zeng, Kaixin Zhu et al.Feb 2arXiv

This paper argues that true world models are not just sprinkling facts into single tasks, but building a unified system that can see, think, remember, act, and generate across many situations.

#world models#unified framework#multimodal reasoning

Rethinking Video Generation Model for the Embodied World

Beginner

Yufan Deng, Zilin Pan et al.Jan 21arXiv

Robots need videos that not only look pretty but also follow real-world physics and finish the task asked of them.

#robot video generation#embodied AI#benchmark

TwinBrainVLA: Unleashing the Potential of Generalist VLMs for Embodied Tasks via Asymmetric Mixture-of-Transformers

Intermediate

Bin Yu, Shijie Lian et al.Jan 20arXiv

TwinBrainVLA is a robot brain with two halves: a frozen generalist that keeps world knowledge safe and a trainable specialist that learns to move precisely.

#Vision-Language-Action#catastrophic forgetting#Asymmetric Mixture-of-Transformers

Aligning Agentic World Models via Knowledgeable Experience Learning

Intermediate

Baochang Ren, Yunzhi Yao et al.Jan 19arXiv

WorldMind teaches AI agents to learn the rules of the real world while they act, instead of cramming everything into fixed model weights.

#agentic world models#predictive coding#physical hallucinations

Fast-ThinkAct: Efficient Vision-Language-Action Reasoning via Verbalizable Latent Planning

Intermediate

Chi-Pin Huang, Yunze Man et al.Jan 14arXiv

Fast-ThinkAct teaches a robot to plan with a few tiny hidden "thought tokens" instead of long paragraphs, making it much faster while staying smart.

#Vision-Language-Action#latent reasoning#verbalizable planning

VL-LN Bench: Towards Long-horizon Goal-oriented Navigation with Active Dialogs

Intermediate

Wensi Huang, Shaohao Zhu et al.Dec 26arXiv

Real life directions are often vague, so the paper creates a task where a robot can ask questions while it searches for a very specific object in a big house.

#embodied AI#interactive navigation#instance goal navigation

PhysBrain: Human Egocentric Data as a Bridge from Vision Language Models to Physical Intelligence

Intermediate

Xiaopeng Lin, Shijie Lian et al.Dec 18arXiv

Robots learn best from what they would actually see, which is a first-person (egocentric) view, but most AI models are trained on third-person videos and get confused.

#egocentric vision#first-person video#vision-language model

Openpi Comet: Competition Solution For 2025 BEHAVIOR Challenge

Intermediate

Junjie Bai, Yu-Wei Chao et al.Dec 10arXiv

This paper shows how to make home-helper robots better at long, multi-step chores by smart training on diverse tasks and by polishing the model after training using its own best attempts.

#Vision-Language-Action#long-horizon manipulation#rejection sampling fine-tuning