The authors introduce a causal decomposition framework that represents a composite task as interconnected unit tasks with explicit prerequisite dependencies.
Controlled prerequisites reveal why multimodal models fail composite tasks
CADET distinguishes intrinsic capability deficits from errors inherited from perception, spatial, temporal, or cognitive prerequisites.
Big Tech
Xia Hu · Brian Potetz · Chun-Ta Lu · Huanfen Yao · Leonidas Guibas · Zhicheng Wang · +3 more
Google Research · Stanford University
Research Digest··3 min read
Hu and colleagues evaluate multimodal large language models by intervening on the prerequisite steps of composite tasks, rather than relying only on final-answer accuracy.
Why this paper
From Google Research and Stanford University
In one line
Causal task decomposition isolates intrinsic capability failures from cascading prerequisite errors in MLLMs.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§