Unverified50% confidenceCautionExact time
Without a tool allowlist, the model may fail frequently because many legitimate calls are denied.
1
Sources
50%
Confidence
Long-term
Relevance
11/7/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedWhen models fail in tool-orchestration tasks with side effects, it is often not that they "don't know how to call tools," but that they lack explicit modeling of preconditions, transaction boundaries, and idempotency.75% similarUnverifiedIn the experiments, even when the system correctly suggested the next tool that should be invoked, many tasks still failed, with failure modes including directly giving up and not initiating any tool call.75% similarUnverified轻量模型在高难度任务上的失败往往是结构性的,切换模型比优化提示词更务实75% similarUnverifiedIn standard evaluations, weak models usually lag significantly behind frontier models on tasks such as reasoning and tool-calling accuracy.74% similarUnverifiedAmong AI application failure cases, a considerable proportion stem from using the wrong tool rather than insufficient model capability, but this proportion lacks authoritative statistics, with figures varying widely across sources.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1144934API
curl https://kongchang.com/api/v1/knowledge/claims/1144934MCP
get_claim(id=1144934)