Memory-Augmented Question Answering on AdaMem-Bench Implicit Feedback 1.0 (test)
97.2AccuracyFull Context
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Full ContextQA Evaluator=DeepSeek-V4-Flash, Extractor=Extractor-independent2026.06 | 97.2 | — | — | 0 | |
| Ideal MemoryQA Evaluator=DeepSeek-V4-Flash, Extractor=Extractor-independent2026.06 | 96.4 | 99.4 | 1 | 110 | |
| AdaMemExtraction model=Gemini-3.5-Flash, QA Evaluator=DeepSeek-V4-Flash2026.06 | 85.4 | 47 | 0.276 | 355 | |
| Mem0Extraction model=Gemini-3.5-Flash, QA Evaluator=DeepSeek-V4-Flash2026.06 | 83.2 | 45.1 | 0.244 | 391 | |
| AdaMemExtraction model=DeepSeek-V4-Flash, QA Evaluator=DeepSeek-V4-Flash2026.06 | 82 | 40.3 | 0.247 | 380 | |
| Mem0Extraction model=DeepSeek-V4-Flash, QA Evaluator=DeepSeek-V4-Flash2026.06 | 80.2 | 33.9 | 0.225 | 407 |