The Token Replacement Test shows VLMs keep most accuracy gains even after corrupting or replacing continuous thought token content, indicating the tokens are not used as information bottlenecks.
Visual spatial reasoning
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
years
2026 3verdicts
UNVERDICTED 3representative citing papers
ESI-Bench shows active exploration outperforms passive observation in multimodal LLMs on spatial tasks but reveals failures from poor action choices and overconfident belief commitment unlike humans.
IPT supervision improves spatial reasoning in VLMs on perspective taking, path tracing, and multiview counting tasks, often outperforming textual chain-of-thought while remaining consistent with observed inputs.
citing papers explorer
-
Ablate-to-Validate: Are Vision-Language Models Really Using Continuous Thought Tokens?
The Token Replacement Test shows VLMs keep most accuracy gains even after corrupting or replacing continuous thought token content, indicating the tokens are not used as information bottlenecks.
-
ESI-Bench: Towards Embodied Spatial Intelligence that Closes the Perception-Action Loop
ESI-Bench shows active exploration outperforms passive observation in multimodal LLMs on spatial tasks but reveals failures from poor action choices and overconfident belief commitment unlike humans.
-
Imaginative Perception Tokens Enhance Spatial Reasoning in Multimodal Language Models
IPT supervision improves spatial reasoning in VLMs on perspective taking, path tracing, and multiview counting tasks, often outperforming textual chain-of-thought while remaining consistent with observed inputs.