List your company

Best-action selection on Target n=30 (test)

65.9Macro F1Gemini 2.5 Flash
4.33220.31636.352.284Jun 2, 2026

Evaluation results

MethodLinks
Gemini 2.5 FlashState Conditioning=Oracle2026.0665.9
PIWM (full)State Conditioning=Oracle2026.0664.1
GPT-5.5Interface=Codex CLI, Protocol=diagnostic, State Conditioning=Oracle2026.0662.2
Claude Sonnet 4.6State Conditioning=Oracle2026.0657.9
GPT-4oState Conditioning=Oracle2026.0644.5
Random BaselineState Conditioning=Oracle2026.0641.4
Random BaselineState Conditioning=E2E2026.0641.4
Small-Scale Action TrainingState Conditioning=Oracle2026.0639
GPT-4oState Conditioning=E2E2026.0635.4
Qwen2.5-VL-7BProtocol=zero-shot, State Conditioning=Oracle2026.0631.3
PIWM (full)State Conditioning=E2E2026.0629.5
Claude Sonnet 4.6State Conditioning=E2E2026.0629.4
State-Outcome ModelState Conditioning=Oracle2026.0624
Gemini 2.5 FlashState Conditioning=E2E2026.0621.6
Constant-Action BaselinesState Conditioning=Oracle2026.066.7