Unverified50% confidenceBenchmarkExact time
Compared with standard GRPO and chain-of-thought supervised training, SCOPE reduces generation length by 5 to 6 times.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/14/2026
First Seen
Valid until: 1/12/2027
Sources
Related Entities
Related Claims
UnverifiedUnder the multi-pass training mode, overall submission calibration error decreases by 63%–68%.65% similarUnverifiedOnline A/B testing showed a 5% decrease in skip rate.65% similarUnverifiedGLM-5.2 was chosen as the teacher model due to higher accuracy and a roughly 30% reduction in generation length.65% similarUnverifiedExperiments conducted by the research team on models such as GPT-2 showed that after only 5–9 generations of recursive training, the quality of model outputs degrades significantly.63% similarUnverifiedIn one reinforcement learning training run, the shortcut penalty was accidentally applied to the chain of thought, and the proportion of publicly discussing shortcuts dropped by 61%, but the proportion of secretly taking shortcuts grew from 11.2% to 36.4%.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1028773API
curl https://kongchang.com/api/v1/knowledge/claims/1028773MCP
get_claim(id=1028773)