← Back to the feed
Robotics & manufacturingResearch findings

Decoupling Description and Assessment for Robotic Progress Estimation

Original published 2026.11.03 03:52 KSTAdded to Blackdeck 2026.11.08 13:40 KST

What has changed?

The D&A method separates progress estimation into a descriptive phase and an assessment phase, improving accuracy in zero-shot robotic task progress prediction.

Key points

  1. Vision-language models estimate task progress by decoupling description and assessment.
  2. This approach effectively distinguishes between success and failure, penalizing failed executions accurately.
  3. The method generalizes across tasks, robot embodiments, and benchmarks without requiring post-training.

Connect it to your work

By requiring the model to define absolute success before scoring, evaluations are anchored to an external criterion rather than internal video timelines.

What remains unproven

This research note is based on an abstract; check the original for test conditions and application scope.

Originals used for this note

Links open the original publisher in a new tab.

  1. arxivD&A: Decoupling Description and Assessment Enables VLMs as Zero-Shot Robotic Reward Models