Unverified50% confidenceFactExact time
Even in modern embedding models optimized through instruction fine-tuning or contrastive learning (such as BGE and E5), the anisotropy problem has not been completely eliminated, only reduced in degree.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/11/2026
First Seen
Valid until: 1/9/2027
Sources
Related Entities
Related Claims
Unverified阿里团队发现强化学习训练不会导致模型其他能力下降，反而会带来全面协同提升，与传统微调中的灾难性遗忘问题形成对比70% similarUnverifiedThe study concluded that, under its experimental setup, learned contextual planning is only a weak correlation signal rather than a substitute for strong retrieval.65% similarUnverifiedUnlike traditional reinforcement learning, NEAT does not rely on gradient descent but instead simulates natural selection to iterate.65% similarUnverified在提示中提供对比性示例（同时给出正反案例）比单独给出正面案例的引导效果更强，属于少样本学习（Few-shot Learning）65% similarUnverifiedFoundational studies and multiple meta-analyses concluded that the ability transfer expected from learning programming did not occur, and the gains did not transfer to general cognitive tasks.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/995600API
curl https://kongchang.com/api/v1/knowledge/claims/995600MCP
get_claim(id=995600)