ResearchPod Summary
This paper explores whether general-purpose reasoning models, similar to computer-use agents, can be trained to control robots natively by observing scenes, reasoning in a chain of thought, and outputting end-effector poses. Specifically, the authors aim to bridge the performance gap between frontier proprietary models and open-weight models without relying on task-specific harnesses or human demonstrations.
To solve this, the authors built a Post-training Engine for Agentic Robotics (PEAR). They trained a 27B open-weight model using robot-use reinforcement learning (RURL) entirely in simulation, drawing on techniques from computer-use agents. The agent receives camera images and the robot's current state, produces a chain of thought, and outputs a target pose. Inverse kinematics then handles the conversion to joint commands. Training followed a multi-stage curriculum, starting with easier tasks in the LIBERO suite and progressing to more complex tasks in RoboDojo, allowing the agent to learn transferable spatial and manipulation skills.
The model shows remarkable zero-shot transfer capabilities. After training on simulation benchmarks (LIBERO and RoboDojo), the model achieved a 60% success rate on a real-world dual-arm robot that it had never encountered during training. Furthermore, it demonstrated significant robustness to task perturbations and novel instructions, outperforming established vision-language-action (VLA) models on benchmarks like LIBERO-Pro. Notably, this specialized training resulted in no measurable degradation of the model's performance on general reasoning, coding, and vision benchmarks, suggesting that robot-use training can be seamlessly integrated into a foundation model's post-training regimen.
This work provides a simple, scalable recipe for building robot-use agents that do not require fragile, task-specific software harnesses. By proving that robotic control is a skill that can be acquired by reasoning models through simulation-based reinforcement learning, the authors open a path toward creating truly general-purpose robotic agents that can handle diverse, real-world tasks simply by following natural language instructions.
Sam: Practice in simulation helped a reasoning model operate a real robot without demonstrations. In this study, it completed sixty percent of the real-world trials, while the starting model completed none.
Alex: That sounds like a substantial transfer. Are we talking about a robot following a rehearsed routine, or a model deciding what to do?
Sam: The model decides each move from camera images and the task instruction. This is a methods preprint by Jiaming Tang and colleagues, called “Training Reasoning Models in Simulation to Natively Use Real Robots.” The real-world result is promising, but narrow.
Alex: So for researchers, the question is whether simulated practice teaches something portable, rather than just another simulator-specific policy.
Sam: And whether that could become part of training general-purpose models. We’ll unpack the shared interface that makes transfer possible, the strongest comparison, and why these trials don’t yet establish broad robot reliability.
Alex: Start with what already existed. What makes this more than putting a language model in charge of an arm?
Sam: The authors build on robot-use agents: models that observe a robot’s surroundings and choose its actions. Strong proprietary models already performed well this way. Open-weight models, whose trained weights are available, were substantially less capable.
Alex: Other systems give those models tools, though. Does this work remove the tools, or just train the model to use them better?
Sam: It removes the added perception tools, motion planners, and libraries of predefined skills. Those supports are what the authors call a harness. Their model must locate objects, decide how to manipulate them, and notice failed moves itself.
Alex: “No harness” could sound like no robotics software underneath. Surely something still has to turn its requests into motor movements.
Sam: That distinction matters. The model outputs a target for the end effector, the working tip of the arm. Inverse kinematics converts that target position and orientation into joint positions. A conventional controller then moves the arm.
AI-generated third-party summary by ResearchPod. Not official content or an endorsement by the paper authors or affiliated organizations.
Alex: So the model chooses where the gripper should go, rather than controlling every motor. That creates a boundary between reasoning and the robot’s mechanics.
Sam: The same interaction protocol works across different robots. The prompt describes the robot, its coordinates, and its cameras. Its own control software handles execution. After each move, the model gets fresh images and robot state, with earlier exchanges still available.
Alex: Make that concrete. What does an actual task look like through this interface?
Sam: In a real-world task, the instruction asks the robot to use a set square to push blocks into an aligned row. It must then release the tool and reset the arm. The model repeatedly observes, reasons, requests a move, and checks what happened.
Alex: How does it learn that sequence without demonstrations? A failed attempt doesn’t tell it which earlier decision went wrong.
Sam: They use reinforcement learning, meaning learning from rewards for attempted tasks. Here the reward is deliberately sparse: success or failure at the episode’s end. They compare groups of attempts at the same task and starting state, then reinforce responses associated with better outcomes.
Alex: But if the starting model almost never succeeds, that reward gives it very little to learn from. How do they get past that?
Sam: They start with simulated tasks where it already succeeds occasionally. Learning there transfers to harder tasks, which they add once success becomes frequent enough to provide a signal. Their training engine, PEAR, supports these extended interactions at scale.
Alex: Let’s separate learning the training environment from learning something more general. How large was the improvement on the main benchmark?
Sam: On LIBERO, a suite of simulated manipulation tasks, success rose from about two percent to seventy-four percent. That brought the model close to the strongest proprietary comparisons at their maximum reasoning settings. But performance on familiar tasks isn’t the most revealing comparison here.
Alex: Because a robot can score well by repeating the motion associated with a familiar scene. Did they test whether it actually follows a changed instruction?
Sam: They did, using LIBERO-Pro, which alters goals and layouts. Their comparison was a vision-language-action model, connecting images and instructions to robot actions, called pi point five. It had been trained with demonstrations of every original LIBERO task.
Alex: That gives the demonstration-trained model an advantage on familiar tasks. What happens when the scene stays familiar but the requested task changes?
Sam: The reasoning model succeeded about fifty-seven percent of the time. The demonstration-trained model succeeded less than one percent of the time. The authors report that it tended to perform the task originally associated with that scene instead.
Alex: That’s a stronger test of instruction following than merely changing the wording. Does it mean the reasoning model is more robust across the board?
Sam: No. The demonstration-trained model still led when instructions were reworded or objects restyled without changing the required task. It also led on the original benchmark. And those comparison results came from prior reports, not fresh evaluations by this team.
Alex: Now take us out of simulation. How much real-world evidence sits behind the opening result?
Sam: Each model received thirty trials on a real dual-arm robot absent from training. The tasks were object pickup and aligning blocks with the set square. The final model had trained on those task types in simulation, but used no real-robot training data.
Alex: So this tests familiar goals on unfamiliar physical hardware. It doesn’t establish that the robot can handle arbitrary new real-world jobs.
Sam: That’s the key boundary. My assessment is that this is persuasive evidence of transfer within those tasks, not broad deployment readiness. The small real-world evaluation matters more than the headline percentage alone.
Alex: The document also includes drawing and a chemistry task. Do those extend the case beyond the practiced goals?
Sam: They provide additional demonstrations, not a comparable success-rate evaluation. The robot draws a simple version of The Starry Night and selects phenolphthalein to test alkalinity. But the authors say these harder tasks required detailed descriptions of both the robot and the task.
Alex: What else would stop someone treating this as a ready-to-use robot controller?
Sam: Speed and prompt dependence. It reasons at length before every move, and lowering the reasoning effort sharply reduces success. It also cannot reliably plan chunks of actions, so complex movements require repeated decisions and stops. Controller quality still affects accuracy.
Alex: And does learning robot control displace the model’s other abilities? That matters if this is supposed to become general model training.
Sam: Across the tested reasoning, coding, and vision benchmarks, they found no measurable degradation. That supports the proposal, but doesn’t prove every other capability survives. Likewise, the effort comparison links longer reasoning with better performance; it doesn’t isolate why this method generalizes.
Alex: Who should read the full document, and where should they start?
Sam: Robot-learning researchers should start with the robot-use interface, then the robustness and instruction-following comparison. Foundation-model researchers should read the training curriculum and retained-capabilities evaluation. Both should read the sim-to-real section alongside the limitations, especially the narrow task coverage and dependence on detailed prompts.
Alex: And for someone who only needs the central lesson?
Sam: Simulated practice can teach a model skills that survive a change of robot. But transfer is not the same as general reliability.
Alex: Thanks for listening.