Unverified60% confidenceFactExact time
Research has found that large language models tend to be overconfident when not specifically optimized, meaning they output high confidence scores but have a relatively high actual error rate.
2
Sources
60%
Confidence
Long-term
Relevance
9/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe overconfidence of a model answering directly while skipping retrieval is technically referred to as miscalibration, meaning there is a systematic deviation between confidence and actual accuracy.80% similarUnverifiedOver-alignment or over-refusal is a common engineering challenge faced by current mainstream large language models, where models are trained to be overly conservative, resulting in a high false-positive rate for reasonable requests.78% similarUnverifiedThere is a gap between the output confidence of language models and their actual accuracy, a phenomenon known in machine learning as "miscalibration."75% similarUnverifiedUncalibrated neural networks typically exhibit overconfidence, which is especially pronounced on out-of-distribution data.74% similarUnverifiedLarge models have higher overall accuracy, but their errors are harder to detect because their output is more fluent and persuasive.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/912208API
curl https://kongchang.com/api/v1/knowledge/claims/912208MCP
get_claim(id=912208)