Driving · 2026-10-09
Recompose and Refine Latent Reasoning Flows for Vision-Language-Action Models
Hongyu Shi, Sen Zhao, Zuyu Zhang, Lifeng Shen, Ding Zou, Xinyu He, Xu Zhang, Qinghua Zhang
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
FlowMem is a unified vision-language-action model that stores successful latent reasoning as ordered “reasoning flow” fragments. At test time it dynamically retrieves and recomposes compatible latent suffixes across episodes, then refines the resulting reasoning tokens using current visual and proprioceptive evidence before conditioning continuous action generation.
Why it matters
The authors target a gap in latent-reasoning VLA: successful latent computation is usually discarded after execution. By treating reasoning as a reusable, progress-aware trajectory and adapting it to current embodied context, FlowMem aims to improve closed-loop task success and robustness under distribution shifts.
Method
- Store successful latent reasoning as ordered flow fragments with progress and transition metadata (Reasoning Flow Memory).
- Given current embodied context, retrieve anchor-aligned suffixes, deterministically search for a compatible multi-source route that preserves within-source order, then tokenize it for the latent interface.
- Refine retrieved route tokens with a fixed-depth residual operator using current visual/proprioceptive evidence, then inject refined tokens to condition continuous action generation.
Limitation
The paper notes that FlowMem does not lead every perturbation column on LIBERO-Plus, delimiting where the mechanism offers only marginal benefit.
Abstract (from arXiv)
Latent reasoning enables vision-language-action (VLA) models to transform multimodal observations into task-relevant internal states before generating continuous robot actions. While existing methods learn to generate or refine such states for each policy query, they discard successful reasoning after execution and therefore reconstruct similar computation from scratch. We present Reasoning and Flow Memory (FLOWMEM), a unified VLA model that turns successful latent computation into reusable reasoning experience. Rather than appending a fixed retrieved context, FLOWMEM dynamically retrieves and recomposes compatible latent fragments as the embodied context evolves, forming a reasoning route that follows the temporal structure and progress of successful computation. The route is then refined using current visual and proprioceptive evidence before it conditions action generation. Experiments on RoboMME and LIBERO-Plus show that FLOWMEM attains 48.0% and 77.3% success, outperforming memory-free policies by 1.7 and 4.1 percentage points, respectively. These results demonstrate the value of reusing successful latent computation for closed-loop VLA control.
Related papers
- Memory Without Curation: Learning Multi-Stream Latent Memory for Robot Policies
- Reasoning as a Distribution: Generative Plan Latents for Vision-Language-Action Models
- Fast-VLA: You Only Think Once for Robot Manipulation
- Plan First, Act Later: Streaming Reasoning for Executable Memory in Robot Control
- LAW: Unifying Language, Action, and World Modeling for Agentic Physical Intelligence
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First public release: 2026-10-08 · Indexed by Pub-AI: 2026-10-09 UTC
Recompose and Refine Latent Reasoning Flows for Vision-Language-Action Models
Preprint posted: 2026-10-08 · Original record ↗
- arxiv
- 2610.12090
Explicit source versions (1)
- v1 · 2026-10-08
- Preprint available
Source date: 2026-10-08 · First detected: 2026-10-25 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · atom:published (official metadata)
Abstract · arxiv
Latent reasoning enables vision-language-action (VLA) models to transform multimodal observations into task-relevant internal states before generating continuous robot actions. While existing methods learn to generate or refine such states for each policy query, they discard successful reasoning after execution and therefore reconstruct similar computation from scratch. We present Reasoning and Flow Memory (FLOWMEM), a unified VLA model that turns successful latent computation into reusable reasoning experience. Rather than appending a fixed retrieved context, FLOWMEM dynamically retrieves and recomposes compatible latent fragments as the embodied context evolves, forming a reasoning route that follows the temporal structure and progress of successful computation. The route is then refined using current visual and proprioceptive evidence before it conditions action generation. Experiments on RoboMME and LIBERO-Plus show that FLOWMEM attains 48.0% and 77.3% success, outperforming memory-free policies by 1.7 and 4.1 percentage points, respectively. These results demonstrate the value of reusing successful latent computation for closed-loop VLA control.