Unverified50% confidenceBenchmarkExact time
Under held-out interface perturbations, the accuracy of the post-trained Agent drops by nearly 25%.
1
Sources
50%
Confidence
Long-term
Relevance
10/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn real-environment A/B testing, the post-trained checkpoint reduced the tool-call failure rate by 21.2% relatively compared to the earlier version.76% similarUnverifiedA moderate degree of physics guidance reduced long-term prediction error by about 30% relative to purely data-driven training.72% similarUnverifiedIn the feature elimination exercise, after removing number_of_courses_viewed accuracy dropped by only 0.002, whereas after removing interaction_count accuracy dropped by 0.044, a difference of more than twenty times.72% similarUnverifiedAfter matched supervised training, model performance improved to 99.8%, while removing the relational links caused accuracy to plummet to 23.4%, even below the best shortcut's 27.0%.72% similarUnverifiedDuring the course's training process, the loss value decreases from approximately 3.36 to around 0.3270% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1037944API
curl https://kongchang.com/api/v1/knowledge/claims/1037944MCP
get_claim(id=1037944)