Unverified50% confidenceFactExact time
The pretrained BGE-small, using only query information under the P=5 setting, achieves a target exposure rate of 100% on both datasets.
1
Sources
50%
Confidence
Long-term
Relevance
11/4/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedUnder naive training, the network is first trained to 100% accuracy on Task A, then trained only on Task B data, after which Task A accuracy drops from 100% to 64% while Task B reaches 100%.69% similarUnverified在小样本LoRA训练中,rank值设置存在偏差-方差权衡,实践中rank=4至32是常见的安全区间67% similarUnverifiedIn reduced experiments using only 20%, 40%, and 80% of the training data, the performance gains persist stably, confirming the reliability of this method in low-resource settings.66% similarUnverifiedTraditional benchmarks such as MMLU and HumanEval have experienced benchmark saturation due to training data contamination; when GPT-4 was released, its MMLU score was approximately 86%, whereas multiple open-source models have now surpassed 90%66% similarUnverified实验采用8:2比例划分数据,20%作为内部测试集,训练集内部使用5折交叉验证,并移除重复记录以避免数据泄漏65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1133892API
curl https://kongchang.com/api/v1/knowledge/claims/1133892MCP
get_claim(id=1133892)