DEHP adds an online-RL horizon predictor to frozen chunk policies, yielding higher success on precise and long-horizon robot manipulation by adapting chunk length to task stage.
hub Canonical reference
Mixture of horizons in action chunking
Canonical reference. 83% of citing Pith papers cite this work as background.
15
Pith papers citing it
Background
83% of classified citations
abstract
Vision-language-action (VLA) models have shown remarkable capabilities in robotic manipulation, but their performance is sensitive to the $\textbf{action chunk length}$ used during training, termed $\textbf{horizon}$. Our empirical study reveals an inherent trade-off: longer horizons provide stronger global foresight but degrade fine-grained accuracy, while shorter ones sharpen local control yet struggle on long-term tasks, implying fixed choice of single horizons being suboptimal. To mitigate the trade-off, we propose a $\textbf{mixture of horizons (MoH)}$ strategy. MoH rearranges the action chunk into several segments with different horizons, processes them in parallel with a shared action transformer, and fuses outputs with a light linear gate. It has three appealing benefits. 1) MoH exploits long-term foresight and short-term precision jointly within a single model, improving both performance and generalizability to complex tasks. 2) MoH is plug-and-play for full-attention action modules with minimal training or inference overhead. 3) MoH enables dynamic inference with adaptive horizons, which selects stable actions through cross-horizon consensus, achieving 2.5$\times$ higher throughput than baselines while preserving superior performance. Extensive experiments over flow-based policies $\pi_0$, $\pi_{0.5}$, and one-step regression policy $\pi_{\text{reg}}$ demonstrate that MoH yields consistent and significant gains on both simulations and real-world tasks. Notably, under mixed-task setting, $\pi_{0.5}$ with MoH reaches a new state-of-the-art with 99$\%$ average success rate on LIBERO after only $30k$ training iterations. Project page: https://timsty1.github.io/moh/
hub tools
citation-role summary
background 5
baseline 1
citation-polarity summary
years
2026 15representative citing papers
DVAC uses denoising variance as an intrinsic signal to adaptively chunk actions in flow-based robot policies, improving success rates and cutting replans on LIBERO, RoboTwin, CALVIN, and real-world tasks.
A3 reframes dynamic action chunk commitment in VLA models as self-speculative prefix verification, accepting the longest continuous sequence of actions that satisfies consensus-ordered conditional invariance and prefix-closed sequential consistency.
VLA-Corrector adds a detect-and-correct inference layer using a latent vision monitor and online gradient guidance to enable adaptive action horizons in chunked VLA policies.
AutoSpeed optimizes visuomotor policies over candidate trajectories at varying speeds using a composite cost of prediction error versus horizon length, with DCT-based modulation, yielding shorter execution times and higher success rates while producing speeds that align with task stages.
Adding recurrent memory tokens to VLA models raises success rates on partially observable manipulation tasks from 0.42 to 0.84 on training and 0.07 to 0.23 on held-out tasks while preserving performance under full observability.
VLA models with inference-time steering mitigate action leakage in implicit human-robot collaboration, supporting longer horizons and yielding faster, more reliable assembly than shorter-horizon baselines in a 16-person study.
TempoVLA learns a single VLA policy with controllable execution speed via variable-speed trajectory augmentation and explicit speed conditioning.
LIBERO and CALVIN fail multiple proposed diagnostics for shortcut solvability, statistical significance, overfitting, and data dependence, while a tiny 0.09B probe reaches near-SOTA on LIBERO.
PACE dynamically selects execution horizons for action chunks in robot policies by detecting low-speed transition points in predicted speed profiles, raising success rates from 57.8% to 64.2% on 50 simulation tasks and from 50.7% to 70.4% in real-robot tests.
Learns state-conditioned commitment depth in a 7B vision-language policy that jointly predicts actions and replan intervals, outperforming fixed-depth baselines and larger models on Sliding Puzzle and Sokoban while providing a theoretical dominance result.
A verifier called Future Forward Dynamics Causal Attention enables adaptive action execution in World Action Models, reducing model inferences by 69% and improving success rates in robotic tasks.
AsyncShield restores VLA geometric intent from latency via kinematic pose mapping and uses PPO-Lagrangian to balance tracking with LiDAR safety constraints in a plug-and-play module.
SV-VLA uses infrequent heavy VLA planning of action chunks plus a lightweight closed-loop verifier to achieve both efficiency and robustness in dynamic robot control.
LingBot-VA combines video world modeling with policy learning via Mixture-of-Transformers, closed-loop rollouts, and asynchronous inference to improve robot manipulation in simulation and real settings.
citing papers explorer
-
When to Re-Commit: Temporal Abstraction Discovery for Long-Horizon Vision-Language Reasoning
Learns state-conditioned commitment depth in a 7B vision-language policy that jointly predicts actions and replan intervals, outperforming fixed-depth baselines and larger models on Sliding Puzzle and Sokoban while providing a theoretical dominance result.