method · first seen 8 Oct 2026
hindsight reasoning
Using a known solution to reverse-engineer useful intermediate ideas that supervise a model on problems it cannot yet solve.
Mentions per week
PapersCodeDiscussion
Earliest sources
- Learning to Plan by Looking Back: Hindsight Hierarchies for Training Reasoning ModelsPapers · 8 Oct 2026 · Lars Simon, Holger Eble, Manuel Radons
Latest sources
- Learning to Plan by Looking Back: Hindsight Hierarchies for Training Reasoning ModelsPapers · 8 Oct 2026 · Lars Simon, Holger Eble, Manuel Radons