← All threads

Generalist Robot Control

Investigates how generalist AI models can control robotic manipulators reliably without task-specific training or demonstrations.

8 papers

Where this stands

The written synthesis of this thread is for subscribers. Subscribe.

Results across this thread

6 reported results from the papers in this thread.

PaperBenchmarkMetricResultConditionsCode

The table is for subscribers. Subscribe to see every reported number side by side.

How this thread developed

  1. Nov 2026 · Apple

    Robots controlled by generalist AI outperform specialists without task-specific training

    demonstrates that generalist AI models can control robot manipulators more reliably than specialists without task-specific training or demonstrations.

  2. Nov 2026 · Northwestern University, Carnegie Mellon University

    Coding agents control simulated robots less reliably under realistic constraints

    Tests coding agents on simulated robotics tasks under realistic constraints, revealing that agents struggle to control robots reliably.

  3. 3 further papers

    Nov 2026 · Amazon, TU Dresden

    AI agents built without training trail ordinary programs on games

    Tests direct compilation of game rules into neural network weights without training, a form of zero-shot control for game agents on held-out tasks.

  4. 1 further paper

    Nov 2026 · Amazon, Rice University

    Humanoid robots open round doorknobs without real training

    Enables zero-shot transfer of manipulation skills from simulation to real humanoid without real training data

4 of 8 papers shown