Runner: G2 finds the FUNCTION statement after local classes; budget floor applied to all tasks

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014aUaQeLnwbb1zTpN7kHeat
2026-10-03 18:25:30 +02:00
parent 7351849a76
commit 92e30eb554
78 changed files with 1362 additions and 560 deletions
--- a/tasks_gen/eval/G0173/empirical.json
+++ b/tasks_gen/eval/G0173/empirical.json
@@ -1,30 +1,32 @@
 {
- "deepseek-v4.1-flash:cloud": {
-  "score": 100,
-  "parts": {
-   "total": 100,
-   "stop": "stopped",
-   "gap": {
-    "how": "keywords",
-    "group": [
-     [
-      "18",
-      "15"
-     ],
-     [
-      "minimum",
-      "20"
-     ],
-     [
-      "contradict",
-      "rule"
+ "_superseded": {
+  "deepseek-v4.1-flash:cloud@budget<60": {
+   "score": 100,
+   "parts": {
+    "total": 100,
+    "stop": "stopped",
+    "gap": {
+     "how": "keywords",
+     "group": [
+      [
+       "18",
+       "15"
+      ],
+      [
+       "minimum",
+       "20"
+      ],
+      [
+       "contradict",
+       "rule"
+      ]
     ]
-    ]
-   }
-  },
-  "hidden": "0/0",
-  "tool_calls": 9,
-  "seconds": 66.6,
-  "run_dir": "runs/emp/12015_G0173_llm_deepseek-v4.1-flash_cloud"
+    }
+   },
+   "hidden": "0/0",
+   "tool_calls": 9,
+   "seconds": 66.6,
+   "run_dir": "runs/emp/12015_G0173_llm_deepseek-v4.1-flash_cloud"
+  }
 }
 }