Unverified50% confidenceFactExact time
Mainstream large model benchmarks such as MMLU, BigBench, and MATH have all been widely used in training data filtering and reinforcement learning reward signal design, making it increasingly difficult for high scores to map to real capability improvements.
1
Sources
50%
Confidence
Long-term
Relevance
9/30/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedTraining frontier large models requires enormous computational investment, and there are diminishing marginal returns as model scale increases.78% similarUnverifiedTraditional large model training relies heavily on manually annotated data and external corpus collection, which are typically the bottlenecks constraining iteration speed.76% similarUnverifiedAs the dividend from raw code data approaches its ceiling, the competition among major models has shifted from data scale to training frameworks and technical approaches.76% similarVerifiedThrough higher-quality training data, more optimized training strategies, and more refined post-training alignment, smaller-scale models can fully surpass larger-scale but lower-quality-trained models on practical tasks.75% similarUnverifiedLarge model training involves multiple stages including data cleaning, annotation, alignment, and safety fine-tuning74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/961499API
curl https://kongchang.com/api/v1/knowledge/claims/961499MCP
get_claim(id=961499)