Training Strategies for Efficient Embodied Reasoning

TL;DR

The paper tackles the generalization challenge of vision-language-action robot policies by dissecting why embodied chain-of-thought reasoning (ECoT) helps. It introduces ECoT-Lite, a set of lightweight training recipes that isolate representation learning, curriculum, and expressivity mechanisms, aiming to retain performance gains while enabling fast inference. Through extensive LIBERO-90 and BridgeData V2 experiments, the authors show that reasoning pre-training and test-time reasoning dropout can capture much of ECoT's benefit with significantly faster inference, while thinking tokens do not help and full ECoT remains the top performer in some settings. The work provides a structured analysis of CoT in embodied robotics and offers practical guidelines for when to deploy each ECoT-Lite variant, advancing the deployability of reasoning-enabled VLAs in real-world robotics.

Abstract

Robot chain-of-thought reasoning (CoT) -- wherein a model predicts helpful intermediate representations before choosing actions -- provides an effective method for improving the generalization and performance of robot policies, especially vision-language-action models (VLAs). While such approaches have been shown to improve performance and generalization, they suffer from core limitations, like needing specialized robot reasoning data and slow inference speeds. To design new robot reasoning approaches that address these issues, a more complete characterization of why reasoning helps policy performance is critical. We hypothesize several mechanisms by which robot reasoning improves policies -- (1) better representation learning, (2) improved learning curricularization, and (3) increased expressivity -- then devise simple variants of robot CoT reasoning to isolate and test each one. We find that learning to generate reasonings does lead to better VLA representations, while attending to the reasonings aids in actually leveraging these features for improved action prediction. Our results provide us with a better understanding of why CoT reasoning helps VLAs, which we use to introduce two simple and lightweight alternative recipes for robot reasoning. Our proposed approaches achieve significant performance gains over non-reasoning policies, state-of-the-art results on the LIBERO-90 benchmark, and a 3x inference speedup compared to standard robot reasoning.

Figures (15)

• Figure 1: Illustration of our proposed ECoT-Lite approaches. Past robot reasoning policies are performant but slow. By testing numerous hypotheses on why robot reasoning improves policy performance, we find two simple lightweight alternatives for training policies with embodied reasoning data without producing reasonings at test time, boosting performance over non-reasoning VLAs while maintaining fast inference speeds.
• Figure 2: Example intermediate reasoning steps. We use Embodied Chain-of-Thought Reasoning (ECoT Zawalski24-ecot) as a representative robot reasoning approach for this work, and thus indicate which steps it does not use with dashed borders (but they are used in other similar works Hwang24-emmaZhao25-cotvla-visualchainofthoughtreasoning).
• Figure 3: ECoT-Lite training recipes. Blue indicates inputs, orange indicates outputs/generations, dashed border represents absence during test-time (and random drop-out during training). (a): Standard VLA Brohan23-rt2Kim24-openVLA and embodied CoT Zawalski24-ecot training. (b) Pre-train or co-train VLA models with embodied reasoning data. (c): Provide reasoning data as a "scaffolding" in context during training. (d): Train with reasoning dropout, remove reasoning during inference. (e): Introduce non-semantic "thinking tokens" to increase effective model expressivity.
• Figure 4: Example ECoT reasonings for LIBERO and Bridge. See \ref{['fig:example-cots']} for more examples.
• Figure 5: Top: Performance of all methods on LIBERO-90 benchmarks. The most performant approaches are ECoT and the ECoT-Lite reasoning dropout policy, both of which beat past state-of-the-art on the standard LIBERO-90 evaluations (90.8% and 89.4% vs. 88.6% by Mete24-quest). Reasoning pre-training also improves performance significantly. See \ref{['tab:libero-results']} for numerical values and standard errors. Bottom: We replicate the reasoning dropout and pre-training policies in Bridge to validate their real-world effectiveness. Both ECoT-Lite approaches improve on the standard VLA's performance. While full ECoT is the most performant, the ECoT-Lite policies do not generate test-time reasonings, making their inference speeds much faster. See \ref{['tab:bridge-results']} for per-task numerical values and standard errors. Asterisks in legend indicate method appears in top and bottom.