Unverified50% confidenceFactExact time
The study systematically compared expert pruning, FFN pruning, and layer pruning on four categories of math and coding benchmarks.
1
Sources
50%
Confidence
Long-term
Relevance
11/3/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe research team evaluated over 35 compressed checkpoints across six open-weight models, covering the two mainstream method categories of quantization and pruning.68% similarUnverifiedThe orchestration layer contains four key components: task decomposition, model routing, result fusion, and quality verification.63% similarUnverifiedThe research team compared MoFlow against six strong baseline methods on six benchmarks covering three types of tasks: mathematics, code, and question answering63% similarUnverifiedThe paper's controlled experiments cover 5 foundation models, 5 base classifiers and 3 scales, and the results show that the gains hold across different encoders, with larger improvements observed at smaller scales.63% similarUnverifiedThe study proposes a claim-level analysis framework that forms four quadrants by crossing correct/incorrect answers with complete/absent evidence, which can reveal changes masked by aggregate metrics.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1129100API
curl https://kongchang.com/api/v1/knowledge/claims/1129100MCP
get_claim(id=1129100)