RICE (Mixture-of-experts routing): superseded — cited as a baseline and beaten by newer methods. 2 paper(s) critique it, 1 beat it on benchmarks — #59 of 1370 most-superseded. Sub-problem: cluster led by SteerMoE. Newer alternatives in the same sub-problem include PADD, MESA, PR2, RouteHijack, MASCing.

Superseded baseline#59 of 1,370 most-superseded

RICE

Mixture-of-experts routing

superseded — cited as a baseline and beaten by newer methods

2 papers critique it · 1 beat it on benchmarks

What papers say

Verbatim critique sentences, each from a paper that cites RICE as a baseline.

“Yet RICE has drawbacks: it relies on the presence of an explicit $<\!think\!>$ token, making it unsuitable in other settings; it only amplifies experts”
— Steering MoE LLMs via Expert (De)Activation
“RICE~wang2025two recently introduces a steering method that focuses on thinking experts. However, this approach is currently limited to extremely large reasoning models, such as DeepSeek-R1~deepseekai2025deepseekr1incentivizingreasoningcapability and Qwen3-235B~qwen3technicalreport.”
— Do Domain-specific Experts exist in MoE-based LLMs?

Beaten on benchmarks

Head-to-head results where a newer method reports beating RICE. Values are copied from the source paper's tables — verify against the cited paper.

DeepSeekMoE beats RICE · Average [Qwen3-30B-Instruct]
83.3 vs 82.4
Do Domain-specific Experts exist in MoE-based LLMs?
DeepSeekMoE beats RICE · Average [GPT-OSS-120B]
84.4 vs 82.1
Do Domain-specific Experts exist in MoE-based LLMs?
DeepSeekMoE beats RICE · Average [Qwen3-30B-Thinking]
69.6 vs 66.9
Do Domain-specific Experts exist in MoE-based LLMs?
DeepSeekMoE beats RICE · Average [GPT-OSS-20B]
67.1 vs 61.0
Do Domain-specific Experts exist in MoE-based LLMs?
DeepSeekMoE beats RICE · AIME 24 [Qwen3-30B-Instruct]
70.0 vs 63.3
Do Domain-specific Experts exist in MoE-based LLMs?
DeepSeekMoE beats RICE · AIME 25 [Qwen3-30B-Instruct]
60.0 vs 46.7
Do Domain-specific Experts exist in MoE-based LLMs?
DeepSeekMoE beats RICE · AIME 24 [GPT-OSS-20B]
77.3 vs 56.7
Do Domain-specific Experts exist in MoE-based LLMs?
DeepSeekMoE beats RICE · AIME 25 [GPT-OSS-20B]
78.6 vs 46.7
Do Domain-specific Experts exist in MoE-based LLMs?

Newer alternatives

Recent methods in the same sub-problem, not yet superseded in the knowledge base.