Unverified50% confidenceFactExact time
Lowering min_area from 120 to 40 raises recall under mild occlusion from 0.776 to 0.981 and bud recall from 0.382 to 0.947, but false positives rise from 4 to 49 and total error flips from −20.5% to +21.4%.
1
Sources
50%
Confidence
Long-term
Relevance
10/19/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIn the fixed-vector self-similarity experiment, when the offset goes from 0 to 1024, the mean of the raw scores always fluctuates within ±0.38, while the mean of the absolute values only drops from 3.17 to 3.02, a change of about 4%.72% similarUnverified接受率随位置逐级衰减：GSM8K从位置0的0.92下滑至约0.53，MTBench从0.78急跌至0.13，因误差在连续猜测中累积72% similarUnverifiedOn the hard tier, recall@5 improved from 0.000 to 0.955 (under default blending weights), and reached 1.000 when using the lower-bound variant.72% similarUnverifiedAt the 0.81 threshold, the proportion of cached answers that "fit the question" is 5.8 to 10.7 percentage points lower than fresh answers; converted to all responses, the overall quality loss is about 2 to 3 percentage points; Haiku 5.5 has the largest gap (10.7 percentage points), with a cached-answer pass rate of only 67.9%.70% similarUnverifiedIn the fact-updating task, the success rate of the embedding retrieval method fluctuated between 0.30 and 0.95, with a maximum gap of 0.45 across different random seeds.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1065705API
curl https://kongchang.com/api/v1/knowledge/claims/1065705MCP
get_claim(id=1065705)