Saved in:
| Main Authors: | , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2606.01810 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914622120919040 |
|---|---|
| author | Lu, Zheng Gao, Mingqi Xie, Qinlei Zhong, Wanqi Cui, Hanwen Cao, Heng Song, Zirui Yang, Yifan Luo, Chong Liu, Bei Li, Yiming |
| author_facet | Lu, Zheng Gao, Mingqi Xie, Qinlei Zhong, Wanqi Cui, Hanwen Cao, Heng Song, Zirui Yang, Yifan Luo, Chong Liu, Bei Li, Yiming |
| contents | Current benchmarks for embodied vision-language planning often favor linguistic next-token prediction over physically grounded next-state reasoning. This rewards models that mimic statistical language priors rather than track causal dependencies, reducing physical planning to shallow sequence modeling. We argue that reliable physical autonomy requires a shift from linguistically grounded token prediction toward physically grounded causal reasoning. To this end, we introduce Causal-Plan-Bench, a high-fidelity diagnostic suite curated through multi-stage verification to evaluate embodied planning across four causal dimensions. We also construct Causal-Plan-1M, a million-scale corpus of explicit reasoning traces produced by a four-stage annotation pipeline over egocentric videos. Extensive evaluation shows that leading models still struggle to demonstrate genuine physical agency, with Gemini 3 Pro reaching only 38.18 on our benchmark. In contrast, our training recipe enables Causal Planner, built on Qwen3-VL-8B, to internalize physical logic for more accurate next-state estimation. The model achieves strong in-domain performance and cross-benchmark generalization, and reveals a Causal Scaling Law: scaling causal training data to one million instances yields a 36.3% relative gain, from 33.22 to 45.28. Overall, our work provides a concrete step toward turning agents from superficial token predictors into physically grounded causal reasoners. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2606_01810 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Token Predictors Are Not Planners: Building Physically Grounded Causal Reasoners Lu, Zheng Gao, Mingqi Xie, Qinlei Zhong, Wanqi Cui, Hanwen Cao, Heng Song, Zirui Yang, Yifan Luo, Chong Liu, Bei Li, Yiming Artificial Intelligence Current benchmarks for embodied vision-language planning often favor linguistic next-token prediction over physically grounded next-state reasoning. This rewards models that mimic statistical language priors rather than track causal dependencies, reducing physical planning to shallow sequence modeling. We argue that reliable physical autonomy requires a shift from linguistically grounded token prediction toward physically grounded causal reasoning. To this end, we introduce Causal-Plan-Bench, a high-fidelity diagnostic suite curated through multi-stage verification to evaluate embodied planning across four causal dimensions. We also construct Causal-Plan-1M, a million-scale corpus of explicit reasoning traces produced by a four-stage annotation pipeline over egocentric videos. Extensive evaluation shows that leading models still struggle to demonstrate genuine physical agency, with Gemini 3 Pro reaching only 38.18 on our benchmark. In contrast, our training recipe enables Causal Planner, built on Qwen3-VL-8B, to internalize physical logic for more accurate next-state estimation. The model achieves strong in-domain performance and cross-benchmark generalization, and reveals a Causal Scaling Law: scaling causal training data to one million instances yields a 36.3% relative gain, from 33.22 to 45.28. Overall, our work provides a concrete step toward turning agents from superficial token predictors into physically grounded causal reasoners. |
| title | Token Predictors Are Not Planners: Building Physically Grounded Causal Reasoners |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2606.01810 |