Will Hierarchical Reasoning Models scale?
4
Ṁ442026
41%
chance
1D
1W
1M
ALL
Resolves YES if someone makes a Hierarchical Reasoning Models of the type shown in this paper https://arxiv.org/html/2506.21734v1 that is >~100x the size of the 27 million parameter model in the paper, and that has much better performance (comparable to the leap between GPT-2 and GPT-3). Resolves NO if this doesn't happen by the start of 2026.
The scaling doesn't have to happen to the parameter count, if I still judge it to be 100x bigger.
I will not trade in this market, and I will throw away mana on resolution such that I get the same amount of mana from resolving either way.
Get Ṁ1,000 play money
Related questions
Related questions
Will OpenAI release a model which generates images using reasoning / inference-time scaling before 2026?
48% chance
Is scale unnecessary for intelligence (<10B param human-competitive STEM model before 2030)?
72% chance
Is scale unnecessary for intelligence (<10B param human-competitive STEM model before 2027)?
25% chance
Will scaling transformers lead to a 60% score on ARC-AGI-2?
31% chance