# Anvil：通过知识再循环扩展具身空间智能

> 原文链接：https://robotics.hellobike.com/papers/2026-10-15-2610-21882
> 频道：papers | 更新日期：2026-10-15 | 标签：arxiv, cs.RO
> 核心摘要：Anvil通过知识循环统一数据生成、训练与评测，提升具身空间智能，并在七个公开基准上验证。

# Anvil：通过知识再循环扩展具身空间智能

::: v-pre
> 原文链接：[Anvil: Scaling Embodied Spatial Intelligence through Knowledge Recirculation](https://arxiv.org/abs/2610.21882)（arXiv 2610.21882 · 发布日期：2026-10-15）

**Anvil: Scaling Embodied Spatial Intelligence through Knowledge Recirculation**

Anvil通过知识循环统一数据生成、训练与评测，提升具身空间智能，并在七个公开基准上验证。

## 核心要点

- 动机是具身空间智能需定位实体、推理关系并表达为行动相关形式，而联合训练易混淆不同阶段知识。
- Anvil-Gen以异构数据飞轮结合形式特定生成、跨阶段一致性验证和失败驱动再收集，覆盖11种Ground--Relation--Act形式。
- Anvil-Ben提供场景对齐监督与场景关联评估，以区分阶段能力和错误传播。
- Anvil-MOPD通过专家教师蒸馏整合异构空间知识，在七个公开基准平均优于Qwen3.5-9B和RynnBrain-8B。

**验证范围与局限**：验证范围为基于图像的空间记录、保留评测集以及七个公开基准；摘要未说明真机实验。

## 英文摘要 (Abstract)

Embodied spatial intelligence requires more than visual recognition: models must localize entities, reason about their relations, and express this knowledge in action-relevant forms. Yet joint training can conflate the knowledge needed at different stages because supervision is fragmented across incompatible geometric and semantic outputs. We introduce Anvil, a unified framework for data generation, training, and benchmarking for embodied spatial intelligence. (1) Anvil-Gen is a heterogeneous data flywheel that combines form-specific generation, cross-stage consistency verification, and failure-driven recollection across 11 Ground--Relation--Act forms. (2) Anvil-Ben provides scene-aligned supervision for learning diverse embodied spatial capabilities and scene-linked assessments that distinguish stage competence from error propagation. (3) Anvil-MOPD is a 9B student that consolidates heterogeneous spatial knowledge through distillation from specialist teachers. Semantic-first-fault routing targets teacher supervision to the student's earliest decisive failure, while stage normalization controls the loss scale across native outputs of different lengths. We construct 5.8M source-grounded records spanning 160K images, reserving 580K spatial records for held-out evaluation and using the remaining 5.22M records for training. On average across seven public benchmarks, Anvil-MOPD outperforms Qwen3.5-9B and RynnBrain-8B by 9.29 and 11.03 percentage points, respectively, while reaching 80.67% on VABench-Point. Controlled studies support targeted coverage, selective transfer, and feedback-guided repair as complementary framework components.

作者：Weilin Ruan, Bohan Zhou, Jinghao Li, Yangkai Wei, Wenqian Li, Peiran Dong, Yingjie Xu, Leyi Wu, Haojian Huang, Binghui Xie
:::

## 原始参考来源 (Sources)

- [arXiv 2610.21882：Anvil: Scaling Embodied Spatial Intelligence through Knowledge Recirculation](https://arxiv.org/abs/2610.21882)