Unverified50% confidenceFactExact time
Stronger and reasoning models perform better, but still fail extensively on rejections with four or more independent reasons.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/17/2026
First Seen
Valid until: 2/15/2027
Sources
Related Entities
Related Claims
UnverifiedThe study found that models' accuracy on semantically plausible distorted statements is actually higher than their performance on meaningless random replacements.75% 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.73% similarUnverifiedReal-world paraphrasing reduces the invalid intent rejection rate: among 16 configurations, 15 showed a drop in rejection rate for the best model.73% similarUnverifiedEnabling reasoning mode halved the refusal rates of two models.71% similarUnverifiedModels at higher reasoning tiers such as Max or Ultra do not necessarily produce better or more efficient results, and sometimes benchmark scores actually decline due to over-reasoning71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1179134API
curl https://kongchang.com/api/v1/knowledge/claims/1179134MCP
get_claim(id=1179134)