Update strict score bounds to 0.01 and 0.99 to avoid float rounding bugs

Author: PathikSingh

inference.py

@@ -9,7 +9,7 @@
 from dataclasses import dataclass, field
 from typing import Any, Callable, Dict, Iterable
 
-_SCORE_EPS = 0.001
+_SCORE_EPS = 0.01
 
 
 def _clamp_score(value: object) -> float:

server/environment.py

@@ -30,7 +30,7 @@
 from .task_bank import Task, get_random_task
 from .tool_registry import ToolRegistry
 
-STRICT_SCORE_EPSILON = 0.001
+STRICT_SCORE_EPSILON = 0.01
 
 
 def clamp_strict_score(value: float) -> float:

server/grader.py

@@ -29,7 +29,7 @@ class Grader:
     def _strict_score(value: float) -> float:
         if not isfinite(value):
             return 0.5
-        return round(max(0.001, min(value, 0.999)), 3)
+        return round(max(0.01, min(value, 0.99)), 3)
 
     def evaluate(
         self,