Unverified50% confidenceFactExact time
Candidate-set drift (especially the addition of distractor tools) caused most of the performance loss, whereas merely modifying tool descriptions or parameter schemas was comparatively less harmful.
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe magnitude of performance degradation varies enormously across different perturbation axes; some hardware changes barely affect performance, while others cause catastrophic drops.77% similarUnverifiedFor models that do not support tool calling, it is generally necessary to simulate tool-calling behavior through carefully designed prompts, though stability and accuracy will decline noticeably.76% similarUnverifiedThe quality of the description directly affects the accuracy of tool invocation; vague or ambiguous descriptions may cause the model to trigger a tool incorrectly or fail to call it when it should.75% similarUnverifiedThe weighting strategy multiplies quality scores onto the behavior cloning loss, making the gradient contribution of low-quality demonstrations smaller.75% similarUnverifiedNewly launched features often lack sufficient adversarial testing, and security reviews may only cover the main flow, leaving edge cases, abnormal inputs, and permission intersection zones easily blank, so the duplicate rate is lower and the probability of being first is higher.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1069492API
curl https://kongchang.com/api/v1/knowledge/claims/1069492MCP
get_claim(id=1069492)