Unverified50% confidenceBenchmarkExact time
The combination of multi-candidate queries plus majority voting plus a judge model appeared to yield an improvement of about +3 points on half the questions, but on the held-out half the gain was exactly 0, while the cost was 5.5 times that of a single call.
1
Sources
50%
Confidence
Long-term
Relevance
10/24/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe method relies on enumerating candidate L1s, and the computational cost of probability scoring grows linearly with the number of candidates.72% similarUnverified在歧义点的四类行为中，'先搜再问'成功率最高，平均通过率达93.4%，远高于'直接猜'的56.5%，'搜多次再猜'为51.5%70% similarUnverifiedThe winner determination logic is to first sort by average score from highest to lowest; if average scores are tied, completion rate is compared; if still tied, the one with fewer steps wins.70% similarUnverifiedPapers that were randomly boosted with initial selections saw their probability of subsequent selection increase by 45.55 percentage points (95% confidence interval: 41.20 to 49.90).68% similarUnverifiedDemo three introduces a decision rule: the top option must reach at least 80% and exceed the second-best option by 20 percentage points; otherwise, the output is "request clarification."68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1092537API
curl https://kongchang.com/api/v1/knowledge/claims/1092537MCP
get_claim(id=1092537)