A technique for fine-tuning AI models by training them on human preference rankings, commonly used to align language model outputs with desired behavior.
Diogo Almeida left OpenAI two years ago convinced that human language is the wrong output for automation. The model his startup shipped this week returns probabilities, and one team testing it clocked classification 5 to 18 times faster.
Reality
A model that quits is harder to audit than a model that lies, because a refusal leaves no artifact to check. The Xe driver fix shows the quitting verdict can be flatly wrong.
Perspective Coverage
3 publishersReality
A developer ran 200 flagged snippets past two frontier models with identical prompts. One cleared 51% of the false alarms; the other cleared 20% and agreed with 90% of what it saw. The countermeasures are a model property.
Reality
A vendor audited 359,388 edges in its own memory store and found feedback-shaped graphs only where its benchmark harness supplied the feedback. Live tenants got one reinforcement per edge.
Reality
A UNICAMP team tested 21 models against left-, right- and unlabelled users. All of them moved toward the user, which makes any neutrality audit run without a user profile close to useless.
Reality