Unverified50% confidenceFactExact time
Cases in which two different configurations gave opposite-sign judgments on the same action within the same day accounted for about 4%.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/30/2026
First Seen
Valid until: 1/28/2027
Sources
Related Entities
Related Claims
UnverifiedThe same prompt produces noticeably different results across generation rounds, so at least three to five versions should be generated in batch.65% similarUnverifiedThe release notes state that Decision-1 changes its decision on an average of 1.3% of perturbations, and shows zero flips when option descriptions are reworded, reversed, or shuffled.65% similarUnverifiedRewriting one criterion in a prompt causes the pass rate of other criteria within the same prompt to fluctuate by 5 to 10 percentage points.64% similarUnverifiedThe experimental results show that even with identical inputs, the raw Recall@1 between latent representations produced by different schemas is only 5.9%–17.7%.64% similarUnverifiedAmong the four judges that also see the artifact, the stronger the model, the smaller the gain from claimcheck, and the strongest judge is already close to the level of 'informed single call'.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1113938API
curl https://kongchang.com/api/v1/knowledge/claims/1113938MCP
get_claim(id=1113938)