{"data":[{"project":"default","id":"cap.cardinality-cases-kai-1.4.c1856a77-kai","benchmark":"cap.cardinality","split":"cases","system":"kai","version":"kai-1.4@c1856a77730f33781dd4ca662a629051d445b50a52d3ea012c9492395461f52a","baseline":false,"study":"decision-suite","embedder":"","reader":"","k":null,"temperature":null,"max_tokens":null,"prompt":"","prompt_digest":"","dataset_digest":"","store_digest":"","facts_digest":"","commit":"9492aa5","questions":1200,"answered":1200,"when":"","ended":"2026-10-16T13:32:12Z","by":"evo","completion":"complete","key":"49201bb3cdf0355277cbb9770916ebbb96e4bf66122b0d416c35efe1827bce9c","measures":[{"category":"all","metric":"k4.accall","value":0.9425,"lo":null,"hi":null,"n":377,"of":400},{"category":"all","metric":"k4.accuracy","value":0.9425,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k4.brier","value":0.0953,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k4.ece","value":0.0475,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k4.nll","value":0.426,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k4.recall1","value":0.9425,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k4.recall5","value":1,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k4.recall20","value":1,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k4.rejected","value":0,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k4.unanswered","value":0,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k16.accall","value":0.92,"lo":null,"hi":null,"n":368,"of":400},{"category":"all","metric":"k16.accuracy","value":0.92,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k16.brier","value":0.1372,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k16.ece","value":0.0653,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k16.nll","value":0.6162,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k16.recall1","value":0.92,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k16.recall5","value":0.9875,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k16.recall20","value":1,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k16.rejected","value":0,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k16.unanswered","value":0,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k77.accall","value":0.88,"lo":null,"hi":null,"n":352,"of":400},{"category":"all","metric":"k77.accuracy","value":0.88,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k77.brier","value":0.1987,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k77.ece","value":0.0936,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k77.nll","value":0.7215,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k77.recall1","value":0.88,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k77.recall5","value":0.9675,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k77.recall20","value":0.9875,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k77.rejected","value":0,"lo":null,"hi":null,"n":null,"of":null},{"category":"all","metric":"k77.unanswered","value":0,"lo":null,"hi":null,"n":null,"of":null}],"notes":"device cpu; dtype f32-q8_0; host evo; Kai: kai serve --device cpu --reach 32768 (kai 0.4.8, sha256 949bac5535016a23...) over the kai-1.4 folder (weights c1856a77, kai.json 8e3f977c, calibration cal_68c2d5e983f2705d), three replicas of four threads on evo's x86-64 CPU, production's CPU family (evo-suite.sh, run.log); kai-1.4 holds no reader and no capability, so its own head answers every question. Cases: capability/cardinality.py's, Banking77 test, 400 a size (at 4 and 16 the gold intent and its nearest by word overlap, at 77 every intent). 150 options (CLINC150) was not read: the reader's host had no `datasets` (cardinality/card.0.log). hanzo-inc/kai results/kai-1.4/unified/suite at 9492aa5; scored by hanzoai/benchmarks decision/ at a3c7859. ","revision":"original","canonical":true,"visibility":"public","trainable":false,"publishable":true}],"total":1,"completion":{"complete":1,"partial":0,"inconsistent":0,"unknown":0},"gaps":[{"field":"dataset_digest","runs":1},{"field":"embedder","runs":1},{"field":"facts_digest","runs":1},{"field":"k","runs":1},{"field":"max_tokens","runs":1},{"field":"prompt","runs":1},{"field":"prompt_digest","runs":1},{"field":"reader","runs":1},{"field":"store_digest","runs":1},{"field":"temperature","runs":1},{"field":"when","runs":1}]}