Unverified50% confidenceBenchmarkExact time
The general-purpose policy model π_{0.5} has a grasping success rate of 0% in real deployment scenarios without adaptation.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAcross all 68 task-paradigm combinations, in 35.3% of them PPG foundation models show no meaningful advantage over the random/constant baseline.72% similarUnverifiedWhen the success rate is extremely low, almost all sample rewards are identical, resulting in extremely high policy gradient variance and very low signal-to-noise ratio, preventing the model from obtaining effective update directions.69% similarUnverified传统的成功率指标难以全面反映策略的鲁棒性，一个策略可能在90%的常规任务中表现出色却在剩余10%的关键场景中出现灾难性失败68% similarUnverifiedBenchmark scores measure the average of a model's success consistency and failure frequency, not its peak capability.68% similarUnverifiedOn performance model code construction tasks, the average pass rate of all model configurations other than GPT-5.6 Sol was below 15%, with extremely large fluctuations across multiple runs.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1002338API
curl https://kongchang.com/api/v1/knowledge/claims/1002338MCP
get_claim(id=1002338)