Unverified50% confidenceOpinionExact time
Another hypothesis holds that progress in non-verifiable domains comes solely from the continuous injection of large-scale new human data during training, and that these capabilities will hit a ceiling if high-quality human data is exhausted.
1
Sources
50%
Confidence
Long-term
Relevance
10/9/2026
First Seen
Sources
Related Claims
VerifiedThe probability of hallucination increases significantly when contradictory information exists in the training data or when the model is asked about domains that are sparsely represented in the training data.77% similarVerifiedLarge models output the most likely next word rather than precise retrieval of facts, which leads to hallucinations in domains with sparse coverage in the training data.76% similarUnverifiedHumanEval has a limited sample size and has been extensively covered by training data, leading to suspicions of test set leakage in some models75% similarUnverifiedThe understanding of text-to-image models is essentially large-scale statistical association, and generation quality drops significantly when dealing with long-tail concepts that appear rarely in the training set.73% similarUnverified数据规模的增长并不等同于有效学习信号的增长，在模型能力提升到一定阈值后简单堆砌样本的收益会急剧下降73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/991130API
curl https://kongchang.com/api/v1/knowledge/claims/991130MCP
get_claim(id=991130)