New benchmark tests multimodal AI models on counterfactual spatial transition reasoning
Researchers introduced CST-Bench, a benchmark evaluating counterfactual spatial transition reasoning in multimodal large language models, paired with a causal transition on-policy self-distillation method, addressing gaps in static-scene-focused benchmarks.
Read the full story at arXiv cs.AI ↗
Timeline · 1 report
- 2026-11-06 05:00 · arXiv cs.AI
CST-Bench: Benchmarking Counterfactual Spatial Transition Reasoning with Causal Transition On-Policy Self-Distillation