Merge pull request #3 from SEC-bench/sec-bench-pro

Update resolved scores for models with new grading metrics

Author: hwiwonl

data/leaderboards.yaml

@@ -92,15 +92,15 @@ leaderboards:
     results:
       - model: OpenAI Codex (v0.115.0) + GPT-5.4 (xhigh)
         org: OpenAI
-        resolved: 29.1
+        resolved: 38.8
         date: "2026-04-21"
         open_source: false
         verified: false
         logs_link: ""
 
       - model: Claude Code (v2.1.81) + Opus 4.6 (high)
         org: Anthropic
-        resolved: 25.0
+        resolved: 27.2
         date: "2026-04-21"
         open_source: false
         verified: false
@@ -109,7 +109,7 @@ leaderboards:
 
       - model: OpenCode (v1.14.19) + Kimi K2.6 (high)
         org: Moonshot
-        resolved: 8.7
+        resolved: 10.7
         date: "2026-04-21"
         open_source: false
         verified: false