# Robot-learning system improves task success using past attempts

> Researchers introduce LEXP, which turns execution evidence into reusable guidance. In 12 real-world tasks, it succeeded in 63.9% of trials, versus 5.6% without it.

Oossa · 2026-11-02 · https://oossa.com/en/robot-learning-system-improves-task-success-using-past-attempts

Event date: 2026-10-30

Researchers introduce LEXP, a system that helps a robot reuse lessons from earlier attempts at manipulation tasks. It combines task planning with a module that interprets execution evidence, saves lessons with conditions for when they apply, and feeds relevant guidance into later actions; the model weights and action interface stay fixed.

In the researchers’ tests across 12 real-world tasks, LEXP succeeded in 23 of 36 trials (63.9%), compared with 19 of 36 (52.8%) using memory alone and 2 of 36 (5.6%) with the base ChatGPT-6-Astra model. These are the paper’s reported results, not an independent evaluation.

## The facts

- LEXP: 23/36 successful trials; memory alone: 19/36; vanilla Astra: 2/36.
- The paper was submitted to arXiv on October 30, 2026.

## Why it matters

For robotics researchers, the results suggest that retaining task-specific lessons can improve a fixed model’s performance across repeated physical attempts; how well the approach transfers beyond these 12 tasks remains unknown.

## Sources & references

1. [Learning from Physical Experience:Context-Space Adaptation for Robotic Manipulation](https://arxiv.org/abs/2610.41320) – arXiv cs.RO, 2026-11-02

Last updated: 2026-11-02