Long-horizon agentic task on Healthbench Hard
28.06PerformanceAggAgent
Evaluation Results
| Method | Links | |
|---|---|---|
| AggAgentBackbone=Qwen3.5-122B, Aggregation Category=LLM-based Aggregation, Number of Rollouts (K)=82026.04 | 28.06 | |
| AggAgentBackbone=GLM-4.7-Flash, Aggregation Category=LLM-based Aggregation, Number of Rollouts (K)=82026.04 | 27.99 | |
| Solution AggregationBackbone=Qwen3.5-122B, Aggregation Category=LLM-based Aggregation, Number of Rollouts (K)=82026.04 | 26.3 | |
| AggAgentBackbone=MiniMax-M2.5, Aggregation Category=LLM-based Aggregation, Number of Rollouts (K)=82026.04 | 24.46 | |
| Summary AggregationBackbone=Qwen3.5-122B, Aggregation Category=LLM-based Aggregation, Number of Rollouts (K)=82026.04 | 23 | |
| Solution AggregationBackbone=MiniMax-M2.5, Aggregation Category=LLM-based Aggregation, Number of Rollouts (K)=82026.04 | 21.84 | |
| Summary AggregationBackbone=MiniMax-M2.5, Aggregation Category=LLM-based Aggregation, Number of Rollouts (K)=82026.04 | 16.92 | |
| Solution AggregationBackbone=GLM-4.7-Flash, Aggregation Category=LLM-based Aggregation, Number of Rollouts (K)=82026.04 | 15.72 | |
| Best-of-NBackbone=Qwen3.5-122B, Aggregation Category=Heuristic Aggregation, Number of Rollouts (K)=82026.04 | 13.01 | |
| Pass@1Backbone=Qwen3.5-122B, Aggregation Category=Single Rollout, Number of Rollouts (K)=12026.04 | 12.87 | |
| Fewest Tool CallsBackbone=Qwen3.5-122B, Aggregation Category=Heuristic Aggregation, Number of Rollouts (K)=82026.04 | 12.83 | |
| Best-of-NBackbone=GLM-4.7-Flash, Aggregation Category=Heuristic Aggregation, Number of Rollouts (K)=82026.04 | 9.91 | |
| Pass@1Backbone=MiniMax-M2.5, Aggregation Category=Single Rollout, Number of Rollouts (K)=12026.04 | 9.67 | |
| Fewest Tool CallsBackbone=GLM-4.7-Flash, Aggregation Category=Heuristic Aggregation, Number of Rollouts (K)=82026.04 | 8.9 | |
| Best-of-NBackbone=MiniMax-M2.5, Aggregation Category=Heuristic Aggregation, Number of Rollouts (K)=82026.04 | 8.79 | |
| Pass@1Backbone=GLM-4.7-Flash, Aggregation Category=Single Rollout, Number of Rollouts (K)=12026.04 | 8.67 | |
| Summary AggregationBackbone=GLM-4.7-Flash, Aggregation Category=LLM-based Aggregation, Number of Rollouts (K)=82026.04 | 7.35 | |
| Fewest Tool CallsBackbone=MiniMax-M2.5, Aggregation Category=Heuristic Aggregation, Number of Rollouts (K)=82026.04 | 5.34 |