What if you could fine-tune a diffusion model by simply following its natural "frequency rhythm"? FeRA is here. It's a new framework that tunes AI image generators by aligning updates with the model's intrinsic frequency energy—how it naturally builds details from blurry to
@jiqizhixin
-
ByteDance Robot Copilot: VR-Guided Autonomous Manipulation
By
–
What if a robot could be your copilot, letting you teach it complex skills with minimal effort?
— 机器之心 JIQIZHIXIN (@jiqizhixin) 21 décembre 2025
ByteDance presents a new Shared Autonomy framework!
A human uses VR to guide the robot's arm, while an autonomous AI policy (DexGrasp-VLA) takes over the fine, tactile work of the… pic.twitter.com/WXnzZOo0UiWhat if a robot could be your copilot, letting you teach it complex skills with minimal effort? ByteDance presents a new Shared Autonomy framework! A human uses VR to guide the robot's arm, while an autonomous AI policy (DexGrasp-VLA) takes over the fine, tactile work of the
-
PsiBot Introduces First Embodied Native Human Data Collection Solution
By
–
What if we could collect human data in a revolutionary way for real-world applications? PsiBot presents the world's first embodied native human data collection solution, Psi-SynEngine! It includes a portable exoskeleton tactile glove data collection set, a large – scale in
-
Modern LLM Architecture: DeepSeek V3 vs Mistral 3
By
–
This blogpost is a masterpiece! The Big LLM Architecture Comparison From DeepSeek V3 to Mistral 3 Large: A Look At Modern LLM Architecture Design Link:
-
ODB-dLLM: Faster Parallelizable Diffusion-Based Language Models
By
–
Diffusion-based LLMs are fast and parallelizable, but bidirectional attention makes inference expensive due to repeated prefill and decoding. Enter ODB-dLLM, a dual-boundary framework with adaptive prefill length prediction and dLLM-specific jump-share speculative decoding.
-
Stanford’s Coordination Layer Theory Challenges LLM AGI Bottleneck
By
–
Are LLMs a dead end for AGI? Stanford's Edward Y. Chang argues the bottleneck isn't the model, but a missing "coordination layer." The new theory (UCCT) sees reasoning as a phase transition, moving from simple pattern-matching to goal-directed thinking. He built MACI, a
-
PosterCopilot: AI Framework for Professional Poster Design
By
–
What if AI could design professional posters, not just generic images?
— 机器之心 JIQIZHIXIN (@jiqizhixin) 19 décembre 2025
Enter PosterCopilot — a new framework that teaches AI the geometry and aesthetics of real layout design.
It uses a 3-stage training strategy to align visual output with professional standards, enabling… pic.twitter.com/ypjhhBXCBQWhat if AI could design professional posters, not just generic images? Enter PosterCopilot — a new framework that teaches AI the geometry and aesthetics of real layout design. It uses a 3-stage training strategy to align visual output with professional standards, enabling
-
Robot Learning from Success: SRPO Method for Vision-Language Models
By
–
What if a robot could learn from its own successes, not just its failures? This research introduces SRPO, a method that lets vision-language-action models use their own best attempts as a reference to assign smarter rewards. This self-referential approach boosted a robot's
-
OpenAI’s Deep Research Leader Joins Tencent as Chief AI Scientist
By
–
Big AI leadership news @ShunyuYao12 Shunyu Yao (姚顺雨), a rising star in AI agents and one of the key minds behind OpenAI’s Deep Research and Computer-Using Agent (CUA), has just been appointed Chief AI Scientist at Tencent. Tencent builds WeChat, China’s super-app used
-
Elastic Reasoning: Cognitive-Inspired Approach for Large Language Models
By
–
Beyond Fast and Slow: Cognitive-Inspired Elastic Reasoning for Large Language Models Paper: https://
arxiv.org/abs/2512.15089