Complex retrieval and positional sorting on MRCR 128K~512K
53.98ScoreGemini-2.5-Flash-Thinking
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini-2.5-Flash-ThinkingEvaluation Framework=Full-context Inference2025.12 | 53.98 | |
| Gemini-2.5-ProEvaluation Framework=Full-context Inference2025.12 | 53.83 | |
| QwenLong-L1.5-30B-A3BEvaluation Framework=Memory Agent Framework2025.12 | 34.87 | |
| Qwen-Flash-Thinking-1MEvaluation Framework=Full-context Inference2025.12 | 22.1 | |
| Qwen3-30B-A3B-Thinking-2507Evaluation Framework=Memory Agent Framework2025.12 | 16.55 | |
| MemAgent-14BEvaluation Framework=Memory Agent Framework2025.12 | 6.78 |