Unverified50% confidenceBenchmarkExact time
On discharge summaries, both stackings led Stanford/Penn by more than 20 percentage points in recall (recall-first reached 95.3%), with the gain coming almost entirely from generated untyped names; on typed labels, all three exceeded 99%.
1
Sources
50%
Confidence
Long-term
Relevance
11/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedOn nursing notes, recall-first stacking improved recall by 2.8 percentage points relative to Stanford/Penn (reaching 98.9%), but precision was 9.3 percentage points lower.65% similarUnverifiedIn the specific experimental case, the overlapping-anchor scheme (n=912) showed the model at 55.2% versus the majority class at 51.1%, a net advantage of +4.1pp, about 2.4 standard errors; the non-overlapping-anchor scheme (n=163) showed 59.5% versus 54.0%, a net advantage of +5.5pp, about 1.4 standard errors.60% 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%60% similarUnverifiedIn a manually annotated pool of 16 candidates, using oracle selection can raise the full-fact success rate from 26.7% to 90.0%, an improvement of more than 3 times.59% similarUnverifiedThe paired 95% confidence interval for the strict score gain from evidence replacement is [8.77, 10.43].59% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1154380API
curl https://kongchang.com/api/v1/knowledge/claims/1154380MCP
get_claim(id=1154380)