List your company

Scientific Code Generation on ScienceAgentBench (test)

32.8SRMAA
9.9215.8621.827.74Mar 3, 2026Mar 20, 2026Apr 7, 2026Apr 25, 2026May 13, 2026May 31, 2026Jun 18, 2026

Evaluation results

MethodLinks
MAABackbone=GPT-5.42026.0632.8——
SkillOptBackbone=GPT-5.42026.0631.9——
MAABackbone=Qwen3.7-Max2026.0630.7——
SkillOptBackbone=Qwen3.7-Max2026.0630.3——
Reactive UpdateBackbone=GPT-5.42026.0629.2——
Trace2SkillBackbone=GPT-5.42026.0628.6——
Reactive UpdateBackbone=Qwen3.7-Max2026.0628.4——
OpenHands CodeActBase Model=GPT-4o, Knowledge Condition=With Knowledge2026.0327.586.373.5
LCPBase Model=GPT-4o, Knowledge Condition=With Knowledge2026.0327.586.487.3
Trace2SkillBackbone=Qwen3.7-Max2026.0627.2——
LCPBase Model=GPT-4o, Knowledge Condition=Without Knowledge2026.0326.585.190.2
Single-shot (ExpeL-style)Backbone=GPT-5.42026.0626.1——
Single-shot (ExpeL-style)Backbone=Qwen3.7-Max2026.0625.8——
Self-DebugBase Model=GPT-4o, Knowledge Condition=With Knowledge2026.0323.585.671.6
SkillOptBackbone=DeepSeek-V4-Flash2026.0623.4——
Frozen (no memory)Backbone=GPT-5.42026.0622.7——
Self-DebugBase Model=GPT-4o, Knowledge Condition=Without Knowledge2026.0322.684.483.3
SkillOptBackbone=Qwen3.6-Flash2026.0622.6——
Frozen (no memory)Backbone=Qwen3.7-Max2026.0622.1——
MAABackbone=DeepSeek-V4-Flash2026.0621.7——
MAABackbone=Qwen3.6-Flash2026.0620.8——
Trace2SkillBackbone=DeepSeek-V4-Flash2026.0620.4——
Single-shot (ExpeL-style)Backbone=DeepSeek-V4-Flash2026.0619.8——
OpenHands CodeActBase Model=GPT-4o, Knowledge Condition=Without Knowledge2026.0319.683.178.4
Reactive UpdateBackbone=DeepSeek-V4-Flash2026.0619.5——
Reactive UpdateBackbone=Qwen3.6-Flash2026.0618.3——
Trace2SkillBackbone=Qwen3.6-Flash2026.0617.5——
Single-shot (ExpeL-style)Backbone=Qwen3.6-Flash2026.0615.7——
Frozen (no memory)Backbone=DeepSeek-V4-Flash2026.0614.8——
DirectBase Model=GPT-4o, Knowledge Condition=Without Knowledge2026.0311.882.652.9
Frozen (no memory)Backbone=Qwen3.6-Flash2026.0611.3——
DirectBase Model=GPT-4o, Knowledge Condition=With Knowledge2026.0310.883.841.2