ARC: A Reasoning Recipe for Robot Foundation Models

Gokul Puthumanaillam; Tao Sun; Elie Aljalbout; Moritz Reuss; Zhaoshuo Li; Fabio Ramos; Ankit Goyal; Jenai Xuning Yang

cs.ROOct 8, 2026

ARC: A Reasoning Recipe for Robot Foundation Models

Authors: Gokul Puthumanaillam, Tao Sun, Elie Aljalbout, Moritz Reuss, Zhaoshuo Li, Fabio Ramos, Ankit Goyal, Jenai Xuning Yang

Organizations: University of Illinois Urbana-Champaign · NVIDIA · Stanford University · University of Sydney · Proception

Abstract

The prevailing approach to improving robot foundation models (RFMs) relies on larger models, more robot demonstrations, and costly training at scale. We show that there exists an effective and efficient complementary approach: the right reasoning recipe can substantially improve the zero-shot task performance of existing state-of-the-art RFMs. We refer to this recipe as ARC. It consists of three key ingredients: a reasoning trace, a scalable automatic labeling pipeline, and a strategy for adapting pretrained RFMs to use these traces for control. First, we find that effective reasoning traces should be grounded in the robot's next action and explain its causal structure: why the action is appropriate and what effect it should produce. Second, we show that these traces can be generated automatically from existing demonstrations, enabling us to construct ARC-Trace-DROID from DROID without collecting new robot data. Third, we show how state-of-the-art VLAs such as $π_{0.5}$ and WAMs such as Cosmos3-Nano-Policy can learn to use these traces for control, with fine-tuning and inference tailored to each model's architecture and capabilities. Using ARC, we obtain gains in zero-shot RFM performance that, to our knowledge, are unprecedented without additional robot demonstrations or foundation-scale training. The adapted models establish a new state of the art on RoboLab-120 and MolmoSpaces, with gains of up to 50 percentage points on RoboLab-Reasoning-50. On real robots, ARC improves $π_{0.5}$ 's task success by 82.2 percentage points. Project website: https://arc-robot-reasoning.github.io/

Figures & tables

Figure 1 : Overview of Arc : Arc -Trace automatically relabels DROID with action-grounded causal traces to produce Arc -Trace-DROID. We show how to fine-tune VLAs and WAMs with this supervision, achieving benchmark-leading performance without additional robot demonstrations.

Figure 2 : Reasoning traces and action-expert attention.

Figure 3 : Overview of the Arc -Trace relabeling pipeline, illustrated with an example task.

Table 4

Policy	SR
Cosmos3-Nano + ARC	61.0% +50.0
$\pi_{0.5}$ + ARC	57.0% +46.8
$\pi_{0.5}$ + explicit subtasks	21.0%
Cosmos3-Nano-Policy	11.0%
$\pi_{0.5}$	10.2%

Table 3 : Reasoning-50 and Reasoning-Hardware results.

Figure 4 : Hardware rollout of $\pi_{0.5}$ + Arc on RoboLab-Reasoning-Hardware. The VLM dynamically refreshes the reasoning traces ( $\sim 1\,\mathrm{Hz}$ in this episode), while robot control runs at $15\,\mathrm{Hz}$ .

Figure 5 : Illustrative hardware rollout on a long-horizon, reasoning-intensive task in a dynamic environment.

Figure 6 : Instruction robustness on RoboLab-120.

Figure 7 : Inference-time efficiency on RoboLab-120 (default).

Figure 8 : Training efficiency on RoboLab-120 (specific) for Cosmos3-Nano-Policy with and without Arc .

Figure 9 : Key ablations of ARC on RoboLab-120.

Appendix figures & tables18 assets

Supplementary material from the paper’s appendix.

Appendix

Figure 10 : Full natural-language reasoning annotations for “Put the marker in the pot,” with synchronized external- and wrist-camera views. Part 1: labeled keyframes 000–005.

Figure 11 : Full natural-language reasoning annotations for “Put the marker in the pot,” continued. Part 2: labeled keyframes 006–011.

Figure 12 : Arc fine-tuning architecture for $\pi_{0.5}$

Figure 13 : Arc fine-tuning architecture for Cosmos3-nano-policy

Figure 14 : Real-time deployment strategy

RoboLab-Reasoning-50 50 tasks
Task	Category
“The plate should only have edible items”	Simple
“Everything made of wood goes in the bowl”	Simple
“Put the smallest item in the bowl”	Simple
“Put the second tallest object in the bowl”	Simple
“Put the item that matches the bowl’s color into it”	Simple

Appendix

Table 4 : RoboLab-Reasoning-50 tasks and categories. Tasks assigned to two categories retain both labels.

Figure 15 : Hardware setup for real-world experiments

Figure 16 : RoboLab-Reasoning-Hardware task setups for the simple category. Each row pairs the task instruction with the external hardware-camera view and its simulation setup.

Figure 17 : RoboLab-Reasoning-Hardware task setups for the long horizon category. Task instructions are shown alongside external hardware-camera views and simulation illustrations.

Figure 18 : RoboLab-Reasoning-Hardware task setups for the discovery category. Each row pairs the task instruction with the external hardware-camera view and its simulation setup.

Figure 19 : RoboLab-Reasoning-Hardware task setups for the negation category. Each row pairs the task instruction with the external hardware-camera view and its simulation setup.

Panel of Figure 20 — Figure 20 : Detailed RoboLab-120 benchmark results. Results under vague, default, and specific instructions are shown from top to bottom. Each table lists the top 11 policies, including both ARC variants, ranked by success rate. Green-shaded rows highlight ARC policies; +values indicate success-rate gains over their respective base policies in percentage points. Reported uncertainty is included where available.

Table 5 : Training hyperparameters and model configuration for $\pi_{0.5}$ -DROID + ARC (VLA).

Table 6 : Model configuration and training run settings for Cosmos3-Nano-Policy + ARC (WAM).

Explore similar work

$π_{0.7}$ : a Steerable Generalist Robotic Foundation Model with Emergent Capabilities
Apr 16, 2026Physical Intelligence, Bo Ai, Ali Amin +85Robot Policy Learning Robot Skill Learning
Grounded Action Model: 3D Grounding as a Foundation for Robotics
Sep 20, 2026Gehao Zhang, Weikai Huang, Shailesh Shailesh +3Long-Horizon Robotic Manipulation Language-Conditioned Robot Manipulation
Northwestern University · University of Washington · National University of Singapore +1
G0.5: One Autoregressive Stream for Robot Reasoning and Action
Aug 12, 2026Yicheng Liu, Zibin Dong, Baijun Ye +24Cross-Embodiment Robot Learning VLM Reasoning
Galaxea Team

Apr 16, 2026, cs.LGJ/K move · Enter open · S save

$π_{0.7}$ : a Steerable Generalist Robotic Foundation Model with Emergent Capabilities

Physical Intelligence, Bo Ai, Ali Amin, Raichelle Aniceto+84

Robot Policy Learning Robot Skill Learning Language-Conditioned Robot Manipulation Robot Foundation Models Generalization in Robotic Manipulation Cross-Embodiment Robot Learning+2

Open paper AskPDF arXiv

We present a new robotic foundation model, called

π_{0.7}

, that can enable strong out-of-the-box performance in a wide range of scenarios.

π_{0.7}

can follow diverse language instructions in unseen environments, including multi-stage tasks with various kitchen appliances, provide zero-shot cross-embodiment generalization, for example enabling a robot to fold laundry without seeing the task before, and perform challenging tasks such as operating an espresso machine out of the box at a level of performance that matches much more specialized RL-finetuned models. The main idea behind

π_{0.7}

is to use diverse context conditioning during training. This conditioning information, contained in the prompt, makes it possible to steer the model precisely to perform many tasks with different strategies. It is conditioned not just on a language command that describes what it should do, but on additional multimodal information that also describes the manner or strategy in which it should do it, including metadata about task performance and subgoal images. This enables

π_{0.7}

to use very diverse data, including demonstrations, potentially suboptimal (autonomous) data including failures, and data from non-robot sources. Our experiments evaluate

π_{0.7}

across numerous tasks with multiple robot platforms, on tasks that require speed and dexterity, language following, and compositional task generalization.

Abstract

Explore similar work

π0.7π_{0.7}π0.7​: a Steerable Generalist Robotic Foundation Model with Emergent Capabilities

Grounded Action Model: 3D Grounding as a Foundation for Robotics

G0.5: One Autoregressive Stream for Robot Reasoning and Action

$π_{0.7}$ : a Steerable Generalist Robotic Foundation Model with Emergent Capabilities