arXiv:2511.16655v1 Announce Type: cross 
Abstract: Cambrian-S aims to take the first steps towards improving video world models with spatial supersensing by introducing (i) two benchmarks, VSI-Super-Recall (VSR) and VSI-Super-Counting (VSC), and (ii) bespoke predictive sensing inference strategies tailored to each benchmark. In this work, we conduct a critical analysis of Cambrian-S across both these fronts. First, we introduce a simple baseline, NoSense, which discards almost all temporal structure and uses only a bag-of-words SigLIP model, yet near-perfectly solves VSR, achieving 95% accuracy even on 4-hour videos. This shows benchmarks like VSR can be nearly solved without spatial cognition, world modeling or spatial supersensing. Second, we hypothesize that the tailored inference methods proposed by Cambrian-S likely exploit shortcut heuristics in the benchmark. We illustrate this with a simple sanity check on the VSC benchmark, called VSC-Repeat: We concatenate each video with itself 1-5 times, which does not change the number of unique objects. However, this simple perturbation entirely collapses the mean relative accuracy of Cambrian-S from 42% to 0%. A system that performs spatial supersensing and integrates information across experiences should recognize views of the same scene and keep object-count predictions unchanged; instead, Cambrian-S inference algorithm relies largely on a shortcut in the VSC benchmark that rooms are never revisited. Taken together, our findings suggest that (i) current VSI-Super benchmarks do not yet reliably measure spatial supersensing, and (ii) predictive-sensing inference recipes used by Cambrian-S improve performance by inadvertently exploiting shortcuts rather than from robust spatial supersensing. We include the response from the Cambrian-S authors (in Appendix A) to provide a balanced perspective alongside our claims. We release our code at: https://github.com/bethgelab/supersanity

يهدف Cambrian-S إلى تحسين نماذج العالم الفيديو من خلال الحساسية المكانية عن طريق تقديم معيارين، VSI-Super-Recall (VSR) وVSI-Super-Counting (VSC)، بالإضافة إلى استراتيجيات استدلال تنبؤية مصممة خصيصًا. تكشف التحليلات النقدية أن نموذجًا أساسيًا، NoSense، يحقق دقة تصل إلى 95% في VSR دون الحاجة إلى الإدراك المكاني، مما يشير إلى أن هذه المعايير يمكن حلها تقريبًا دون تقنيات الحساسية المكانية المتقدمة. كما يتساءل البحث عن فعالية طرق الاستدلال المخصصة المقترحة من قبل Cambrian-S.

Cambrian-S busca mejorar los modelos de mundo en video mediante la supersensibilidad espacial al introducir dos referencias, VSI-Super-Recall (VSR) y VSI-Super-Counting (VSC), junto con estrategias de inferencia predictiva personalizadas. Un análisis crítico revela que un modelo base, NoSense, logra un 95% de precisión en VSR sin cognición espacial, sugiriendo que tales referencias pueden resolverse casi sin técnicas avanzadas de supersensibilidad espacial. El estudio también cuestiona la efectividad de los métodos de inferencia personalizados propuestos por Cambrian-S.

Cambrian-S vise à améliorer les modèles du monde vidéo grâce à la supersensibilité spatiale en introduisant deux références, VSI-Super-Recall (VSR) et VSI-Super-Counting (VSC), ainsi que des stratégies d'inférence prédictive sur mesure. Une analyse critique révèle qu'un modèle de base, NoSense, atteint 95 % de précision sur VSR sans cognition spatiale, suggérant que ces références peuvent être presque résolues sans techniques avancées de supersensibilité spatiale. L'étude remet également en question l'efficacité des méthodes d'inférence sur mesure de Cambrian-S.

Cambrian-S aims to enhance video world models through spatial supersensing by introducing two benchmarks, VSI-Super-Recall (VSR) and VSI-Super-Counting (VSC), along with tailored predictive sensing strategies. A critical analysis reveals that a baseline model, NoSense, achieves 95% accuracy on VSR without spatial cognition, suggesting that such benchmarks can be nearly solved without advanced spatial supersensing techniques. The study also questions the effectiveness of Cambrian-S's tailored inference methods.

Solving Spatial Supersensing Without Spatial Supersensing

Was this article worth reading? Share it

LucidQuery AI

ShareSpeak

Attentive AI

VSDECO

The Visualizer

SERPView

Ready to build your own newsroom?