Understand this story
On October 30, 2026, Rohan Paul introduced the method RL with Neural Feedback (RLNF). According to his retweet of Rahul Chhabra's introduction, researchers say they have found a reliable way to train AI models to respond to human preferences using only brain signals; in the example, a face image automatically becomes a version more attractive to the subject while the subject gazes at it, with no ratings, clicks, or prompts needed. In another introduction, Rohan Paul explained that the method reads arousal signals directly from the brain while the viewer looks at generated images, using simple embeddings plus k-NN scoring as the reinforcement learning reward. He considers this direction very natural: human preferences already exist in the brain, RLHF converts them into clicks, while RLNF feeds brain signals directly into the reward. Currently, this introduction remains at the level of method description and examples, with no further progress yet.
AI-generated from public reports
Latest development · Report
Rohan Paul reviews RLNF: using brain signals as rewardsDevelopments
1 confirmed developments
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Oct 30
Rohan Paul
Rohan Paul reviews RLNF: using brain signals as rewards
Rohan Paul introduces RL with Neural Feedback (RLNF): while viewers look at generated images, arousal signals are read directly from the brain, and a simple embedding plus k-NN scoring is used as the reinforcement learning reward. He considers this direction very natural—human preferences already exist in the brain; RLHF turns them into clicks, while RLNF connects brain signals directly to the reward.
Rohan Paul
RLNF: AI reads brain signals to generate images you prefer
Rohan Paul shares Rahul Chhabra's introduction of RL with Neural Feedback (RLNF): researchers claim to have found a reliable way to train AI models to respond to human preferences purely from brain signals. In the example, a picture of a face automatically transforms into a version more attractive to the subject while they are looking at it, with no ratings, clicks, or prompts needed.