Arborformer
An experimental language model project built entirely on decision trees, intended to explore non-Transformer language modeling approaches, with an emphasis on interpretability.
Timeline (last 90 days)
Arborformer is a language model experimental project built entirely on decision trees, aimed at replacing the Transformer architecture.
The author of Arborformer emphasizes that it is "100% decision trees," meaning the entire reasoning path consists of tree nodes with no hidden neural network layers.
The original poster explicitly states "seeking feedback," indicating the project is still at an early stage.
The Reddit community's doubts about Arborformer center on expressiveness, scalability, and the lack of public evaluation results comparing it with Transformers.
Arborformer currently looks more like a proof of concept than a mature product.
All Facts (5)
Arborformer is a language model experimental project built entirely on decision trees, aimed at replacing the Transformer architecture.
50%UnverifiedThe author of Arborformer emphasizes that it is "100% decision trees," meaning the entire reasoning path consists of tree nodes with no hidden neural network layers.
50%UnverifiedThe original poster explicitly states "seeking feedback," indicating the project is still at an early stage.
50%UnverifiedThe Reddit community's doubts about Arborformer center on expressiveness, scalability, and the lack of public evaluation results comparing it with Transformers.
50%UnverifiedArborformer currently looks more like a proof of concept than a mature product.
50%