Tag: Reasoning
-
A quote from Boaz Barak, Gabriel Wu, Jeremy Chen and Manas Joglekar
One way to think of confessions is that we are giving the model access to an “anonymous tip line” where it can turn itself in by presenting incriminating evidence of misbehavior. But unlike real-world tip lines, if the model acted badly in the original task, it can collect the reward for turning itself in while…
-
The Goon Squad, by Daniel Kolitz
From these companies’ perspective, the ideal consumer would do literally nothing but goon, lose at gambling, and maybe watch other people play video games. You can try to fight this. You can read a book, pet a dog, buy a stupid box to lock away your phone. You can make a joke about the box,…
-
The hard part: A quote from Jason Gorman
The hard part of computer programming isn’t expressing what we want the machine to do in code. The hard part is turning human thinking — with all its wooliness and ambiguity and contradictions — into computational thinking that is logically precise and unambiguous, and that can then be expressed formally in the syntax of a…
-
Interleaved Thinking
Interleaved thinking is essential for LLM agents: it means alternating between explicit reasoning and tool use, while carrying that reasoning forward between steps.This process significantly enhances planning, self‑correction, and reliability in long workflows. — Read on simonwillison.net/2025/Nov/3/minimax/