Gemini API gemini-3.5-flash-lite; high thinking; default sampling; pass@1 except where noted.
未知项
Exact thinking budget and sampling parameters are not published.
工具
允许工具
推理预算
high
框架
未记录
协议
Self-computed on canonical Partial 30 subset; average position score across independent k=2 runs; bash terminal, 1M context, H100 isolated sandbox, up to 256 tool calls and 20-minute command timeout.
条件
Each problem rollout typically completed within 12 hours.
Gemini 3.5 Flash-Lite record for MLE-Bench · Canonical Partial 30 subset. 1 score record for this release, including 1 exact value (may include historical records). Includes 1 source location.
Gemini 3.5 Flash-Lite · MLE-Bench scores and configurations
Metric and score
Configuration and evaluation conditions
Sources
Average position score39.2%Reported 39.2% · Unit %
Published Unconfirmed
Gemini 3.5 Flash-Lite high thinking
Settings
Gemini API gemini-3.5-flash-lite; high thinking; default sampling; pass@1 except where noted.
Unknowns
Exact thinking budget and sampling parameters are not published.
Tools
Tools allowed
Reasoning budget
high
Framework
not recorded
Protocol
Self-computed on canonical Partial 30 subset; average position score across independent k=2 runs; bash terminal, 1M context, H100 isolated sandbox, up to 256 tool calls and 20-minute command timeout.
Conditions
Each problem rollout typically completed within 12 hours.