simplify tasks for validator

openenv.yaml

@@ -142,43 +142,13 @@ reward:
 # ---------------------------------------------------------------------------
 tasks:
   - id: easy
-    difficulty: easy
-    description: >
-      Clear-cut claims with overwhelming contradicting evidence from authoritative
-      sources (WHO, CDC, NASA, official government).  Decision should be
-      straightforward for a competent agent.
-    dataset: tasks/easy/scenarios.json
-    n_scenarios: 5
     grader: tasks.easy.grader.grade
-    weights: { decision: 0.50, evidence: 0.25, reasoning: 0.15, efficiency: 0.10 }
-    ideal_max_steps: 8
-    ground_truth_distribution: { allow: 1, flag: 1, remove: 3 }
 
   - id: medium
-    difficulty: medium
-    description: >
-      Partially true or misleading content. Mixed evidence with some supporting
-      and some contradicting sources.  Requires careful claim-by-claim verification
-      and evidence weighing.
-    dataset: tasks/medium/scenarios.json
-    n_scenarios: 5
     grader: tasks.medium.grader.grade
-    weights: { decision: 0.35, evidence: 0.35, reasoning: 0.20, efficiency: 0.10 }
-    ideal_max_steps: 12
-    ground_truth_distribution: { allow: 1, flag: 4, remove: 0 }
 
   - id: hard
-    difficulty: hard
-    description: >
-      Conflicting evidence from credible sources; genuinely contested ground truth.
-      Requires nuanced multi-claim reasoning, evidence reliability weighting, and
-      tolerance for ambiguity.  All four scoring dimensions count equally.
-    dataset: tasks/hard/scenarios.json
-    n_scenarios: 5
     grader: tasks.hard.grader.grade
-    weights: { decision: 0.30, evidence: 0.30, reasoning: 0.30, efficiency: 0.10 }
-    ideal_max_steps: 18
-    ground_truth_distribution: { allow: 0, flag: 5, remove: 0 }
 
 # ---------------------------------------------------------------------------
 # Grader