Unverified50% confidenceBenchmarkExact time
On Tev1-0.8B, packing multiple questions into a single request induced a binary flip rate of 11.7% to 22.4%; on Tev1-4B, this proportion dropped to 0.1% to 0.2%.
1
Sources
50%
Confidence
Long-term
Relevance
11/10/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedFine-tuned v3 achieved a safety recall of 13/21 (62%) at the 0.8 alert threshold, versus 0/21 for zero-shot; false alarms on ordinary chat dropped from 2.9% to 1.4%.68% similarUnverifiedUsing the shared gradient bound, at τ=0.01, the escape coordinate ratio drops from 98.9% to 0.53%, and the load drops from 126.6% to 28.3%.68% similarUnverifiedIn a test of 119 pairs of "near-identical questions", at the 0.81 threshold 62% of modified questions incorrectly hit; even raised to 0.90, 25% still hit; the hit rate for genuine synonymous rewrites dropped from 39% to 4%.67% similarUnverifiedIn the hybrid mode at τ=0.4, 27 of the 400 cases are escalated (6.8%), billed Tokens drop from 1,222,211 to 82,499, and the saving including plugin overhead is 86.1%.67% 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.66% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1155165API
curl https://kongchang.com/api/v1/knowledge/claims/1155165MCP
get_claim(id=1155165)