Share your thoughts, 1 month free Claude Pro on usSee more

Long-horizon agentic task on HLE

60Performance

AggAgent

Updated 8d ago

Evaluation Results

Method	Links
AggAgent 2026.04		60
Best-of-N 2026.04		58.06
Summary Aggregation 2026.04		58.06
Weighted Voting 2026.04		57.42
AggAgent 2026.04		54.19
Solution Aggregation 2026.04		54.19
Best-of-N 2026.04		53.55
Majority Voting 2026.04		53.55
Solution Aggregation 2026.04		50.97
Summary Aggregation 2026.04		50.32
Kimi-K2.5-1T 2026.05		50.2
Fewest Tool Calls 2026.04		46.45
Pass@1 2026.04		45.73
Weighted Voting 2026.04		45.16
Majority Voting 2026.04		43.87
Claude-4.5-Opus 2026.05		43.4
Fewest Tool Calls 2026.04		41.29
Pass@1 2026.04		39.35
SAM 2026.05		38.2
GLM-4.7 2026.05		37.5
AggAgent 2026.04		37.42
GLM-4.7 2026.05		37.2
GLM-4.7 2026.05		37.2
SAM 2026.05		37.2
GLM-4.7 2026.05		36.5
Qwen3.5-35B-A3B 2026.05		36
GPT-5.4 2026.05		35.2
Qwen3.5-35B-A3B 2026.05		35.2
Summary Aggregation 2026.04		34.84
Qwen3.5-35B-A3B 2026.05		34.2
Qwen3.5-35B-A3B 2026.05		34
Solution Aggregation 2026.04		32.9
Best-of-N 2026.04		30.97
IterResearcher-30B-A3B 2026.05		28.8
Weighted Voting 2026.04		27.74
Fewest Tool Calls 2026.04		27.1
Majority Voting 2026.04		26.45
Pass@1 2026.04		25
OpenAI-o3 2026.05		24.9
WebThinker-32B 2026.05		15.8
WebSailor-32B 2026.05		9.6