Open Models Almanac
Methods

single mention · not yet filed

Action reasoning via intermediate perception and trajectory representations

Action chain-of-thought extends chain-of-thought reasoning to embodied tasks, having a vision-language-action model reason before it acts.1 Each step of the reasoning chain is grounded directly in the scene, rather than being latent embeddings, generated sub-goals, or textual descriptions that are difficult to ground in the real world.2,3

How this definition was written

Summarised by glm-5.3-flash from 1 source; every sentence is checked against the quote it cites.

  1. recent works in robotics have explored extending reasoning capabilities to embodied tasks within vision-language-action (VLA) models — MolmoAct Action Reasoning Models that can Reason in Space
  2. grounds every step of its reasoning chain directly in the scene — MolmoAct Action Reasoning Models that can Reason in Space
  3. whose reasoning is represented as latent embeddings, generated sub-goals, or textual descriptions that are difficult to ground in the real world — MolmoAct Action Reasoning Models that can Reason in Space

Also called action chain-of-thought (CoT).