Unverified50% confidenceBenchmarkExact time
When sampling eight candidates without constraints from the model, 10.2% of inputs cannot produce any valid output under limited training data, and this proportion drops to 0.4% when using the full training data
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn CLM training, pretraining alone scores 52% on the hard negative sample test; adding the hard negative sample stage jumps it to 69%; training with hard negative samples from day one peaks at only 62% and overfits quickly; omitting replay causes the hard negative sample score to drop from 69% back to 56%.78% similarUnverifiedThe quality ceiling of weak supervision labels depends on the accuracy of the zero-shot model itself; if the model makes systematic errors, the training set will inherit that bias.78% similarUnverifiedAnthropic's research shows that even mild alignment training can cause a 2–8 percentage point decline in model performance on certain reasoning-intensive tasks.77% similarVerifiedWhen fine-tuning with LoRA, the actual number of parameters that need to be trained is only 0.1%-1% of the original.77% similarVerifiedFor a 7-billion-parameter model, LoRA typically requires training less than 1% of the total parameters, and GPU memory requirements can be reduced by more than 60%.76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1002039API
curl https://kongchang.com/api/v1/knowledge/claims/1002039MCP
get_claim(id=1002039)