Add paper baselines to leaderboard

submissions/2026-04-10-anthropic-claude-opus-react.json

@@ -0,0 +1,31 @@
+{
+  "metadata": {
+    "agent_name": "ReAct-Claude-Opus",
+    "base_model": "claude-opus-4-6",
+    "scaffold": "ReAct",
+    "tools_used": ["web_search", "python_interpreter"],
+    "organization": "DeepSynth authors (paper baseline)",
+    "paper_url": "https://arxiv.org/abs/2602.21143",
+    "code_url": "https://github.com/agentdeepsynthesis/deepsynth-bench",
+    "submission_date": "2026-04-10",
+    "split": "test",
+    "num_seeds": 3
+  },
+  "scores": {
+    "overall": {"exact_match": 0.22, "f1": 0.38, "llm_judge": 0.47},
+    "per_domain": {
+      "science":    {"f1": 0.REPLACE},
+      "geography":  {"f1": 0.REPLACE},
+      "economics":  {"f1": 0.REPLACE},
+      "history":    {"f1": 0.REPLACE},
+      "culture":    {"f1": 0.REPLACE},
+      "politics":   {"f1": 0.REPLACE},
+      "technology": {"f1": 0.REPLACE}
+    }
+  },
+  "efficiency": {
+    "avg_cost_usd": 0.44,
+    "avg_latency_s": 15.2,
+    "avg_num_tool_calls": 5.8
+  }
+}