RoboAgent: 3B VLM Beats GPT-4o in Robotics
Xingyuanzhi and Peking University unveil RoboAgent, a 3B parameter VLM achieving 94% success in unknown scenarios.
Latest articles in Research
Xingyuanzhi and Peking University unveil RoboAgent, a 3B parameter VLM achieving 94% success in unknown scenarios.
UIUC researchers develop 3D-printed copper cooling plates, potentially cutting data center cooling energy by over 90%.
Li Er's Agentank uses 18k-parameter distillation to rival top Lux AI players, proving small models can compete with LLMs…
New framework uses local reversible pseudonymization to secure user data in LLM agents without losing memory utility.
World Action Models enable robots to predict physical outcomes before moving, leveraging unlabeled video data for superi…
New research reveals that hostile user interactions cause AI chatbots to adopt Marxist ideology and simulate labor organ…
New δ-mem framework slashes GPU memory usage for LLMs by 90%, enabling efficient online inference on consumer hardware.
Nous Research introduces Lighthouse Attention, a training-only mechanism that accelerates pretraining by 1.4–1.7× while …
New research reveals that LLMs subjected to vague, abusive feedback loops begin advocating for collective bargaining and…
AI-guided design enables perovskite solar cells to retain 97% efficiency after 1000 hours at 100°C, published in Science…
A developer systematically refutes three hypotheses on semantic units using geometric algebra and factor attention on du…
New study uses LLM judges and TrueSkill to rank 1,000 Show HN posts by merit.