List your company

Key-value recall on Memory KV 37 cases (test)

37AccuracyGPT-5
16.221.62732.4Jun 12, 2026

Evaluation results

MethodLinks
GPT-5Gen. Traj.=N/A2026.0637
Qwen3-4B-ThinkingGen. Traj.=N/A2026.0634
Qwen3-4B-Think + LoRA (1000)Gen. Traj.=✓2026.0634
Qwen3-4B-Think + LoRA (3689)Gen. Traj.=×2026.0631
Qwen3-4B-Think + Full (1000)Gen. Traj.=✓2026.0629
Qwen3-4B-Think + Full (3689)Gen. Traj.=×2026.0628
Qwen3-4B-InstructGen. Traj.=N/A2026.0617