Long-context Memory Question Answering on LoCoMo 1.0 (test)
58.11Single Hop F1AnchorMem
Evaluation Results
| Method | Links | |||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| AnchorMemLLM Backbone=GPT-4o-mini, Retrieval top-k=202026.04 | 58.11 | 46.26 | 82.16 | 36.79 | 24.57 | 63.83 | 26.34 | 16.72 | 51.04 | 56.6 | 45.34 | 67.6 | 51.91 | 40.25 | 73.83 | |
| AnchorMemLLM Backbone=Qwen2.5-32B-Instruct2026.04 | 57.82 | 49.76 | 78.83 | 35.53 | 25.71 | 58.51 | 25.62 | 19.95 | 44.79 | 53.12 | 45.26 | 62.31 | 50.57 | 42.54 | 69.54 | |
| AnchorMemLLM Backbone=GPT-4o-mini2026.04 | 55.84 | 44.38 | 78.6 | 34.45 | 23.87 | 56.03 | 25.38 | 17.91 | 47.92 | 55.09 | 43.79 | 68.85 | 49.87 | 38.85 | 70.52 | |
| AnchorMemLLM Backbone=Qwen2.5-7B-Instruct2026.04 | 54.32 | 45.46 | 72.89 | 34.5 | 24.45 | 55.67 | 18.4 | 15.37 | 38.54 | 43.85 | 31.65 | 40.81 | 46.27 | 36.86 | 60.91 | |
| MemoryOSLLM Backbone=GPT-4o-mini, Result Source=Original paper2026.04 | 48.62 | 42.99 | — | 35.27 | 25.22 | — | 20.02 | 16.52 | — | 41.15 | 30.76 | — | 42.84 | 35.53 | — | |
| Mem0LLM Backbone=GPT-4o-mini, Result Source=Original paper2026.04 | 47.65 | 38.72 | 72.93 | 38.72 | 27.13 | 67.13 | 28.64 | 21.58 | 51.15 | 48.93 | 40.51 | 55.51 | 45.1 | 35.9 | 66.88 | |
| HippoRAG 2LLM Backbone=GPT-4o-mini2026.04 | 46.08 | 32.66 | 66.11 | 30.65 | 21.58 | 55.67 | 24.87 | 17.11 | 50 | 51.73 | 36.75 | 59.5 | 43.11 | 30.52 | 61.82 | |
| LightMemLLM Backbone=GPT-4o-mini2026.04 | 44.48 | 35.49 | 67.18 | 29.04 | 20.17 | 50 | 23.81 | 18.09 | 41.67 | 50.92 | 38.06 | 64.8 | 41.71 | 32.14 | 61.95 | |
| NaiveRAGLLM Backbone=Qwen2.5-32B-Instruct2026.04 | 43.12 | 38.34 | 59.57 | 23.89 | 17.38 | 42.2 | 14.19 | 11.84 | 34.38 | 30.75 | 25.36 | 33.33 | 35.16 | 30.09 | 49.35 | |
| LightMemLLM Backbone=Qwen2.5-32B-Instruct2026.04 | 42.55 | 35.42 | 64.33 | 29.71 | 23.4 | 50.71 | 15.58 | 12.06 | 40.63 | 41.23 | 31.4 | 61.99 | 38.24 | 30.93 | 59.87 | |
| LightMemLLM Backbone=Qwen2.5-7B-Instruct2026.04 | 41.34 | 35.3 | 59.93 | 28.55 | 21.9 | 46.81 | 19.56 | 17.38 | 34.38 | 38.37 | 28.26 | 40.81 | 37.02 | 30.26 | 51.95 | |
| NaiveRAGLLM Backbone=Qwen2.5-7B-Instruct2026.04 | 40.43 | 35.27 | 59.45 | 23.76 | 18.15 | 45.04 | 15.67 | 13.17 | 42.71 | 27.41 | 23.83 | 25.55 | 33.12 | 28.37 | 48.7 | |
| NaiveRAGLLM Backbone=GPT-4o-mini2026.04 | 40.2 | 31.56 | 57.55 | 20.41 | 14.81 | 39.36 | 10.91 | 10.01 | 34.38 | 32.07 | 25.56 | 34.27 | 33.06 | 25.9 | 47.92 | |
| A-MemLLM Backbone=Qwen2.5-7B-Instruct2026.04 | 38.93 | 33.95 | 58.33 | 23.7 | 16.62 | 46.81 | 15.34 | 12.71 | 38.54 | 32.87 | 27.55 | 28.66 | 33.41 | 28.12 | 48.8 | |
| MemoryOSLLM Backbone=GPT-4o-mini2026.04 | 38.53 | 33.9 | 55.65 | 35.89 | 27.41 | 56.74 | 22.95 | 17.85 | 42.71 | 22.46 | 20.84 | 23.36 | 33.73 | 28.99 | 48.31 | |
| MemoryOSLLM Backbone=Qwen2.5-32B-Instruct2026.04 | 38.53 | 33.9 | 55.65 | 35.89 | 27.41 | 56.74 | 22.95 | 17.85 | 42.71 | 24.46 | 20.84 | 23.36 | 34.14 | 28.99 | 48.31 | |
| HippoRAG 2LLM Backbone=Qwen2.5-32B-Instruct2026.04 | 38.11 | 26.13 | 62.66 | 27.83 | 20.29 | 56.38 | 23.26 | 16.02 | 42.71 | 43.64 | 32.52 | 56.07 | 36.45 | 25.76 | 58.9 | |
| A-MemLLM Backbone=GPT-4o-mini2026.04 | 35.84 | 29.22 | 57.4 | 23.19 | 16.71 | 42.91 | 9.06 | 8.34 | 23.96 | 38.4 | 33.72 | 51.09 | 32.39 | 26.57 | 51.35 | |
| A-MemLLM Backbone=Qwen2.5-32B-Instruct2026.04 | 35.34 | 31.39 | 47.56 | 25.56 | 17.98 | 43.62 | 11.96 | 10.79 | 28.12 | 29.72 | 25.37 | 34.58 | 30.92 | 26.4 | 42.92 | |
| MemoryOSLLM Backbone=Qwen2.5-7B-Instruct2026.04 | 32.62 | 28.17 | 51.72 | 25.94 | 19.1 | 40.07 | 18.94 | 14.81 | 35.42 | 25.87 | 21.6 | 17.45 | 29.14 | 24.31 | 41.43 | |
| Mem0LLM Backbone=GPT-4o-mini2026.04 | 29.69 | 24.59 | 45.54 | 20.33 | 13.28 | 30.85 | 14.9 | 10.22 | 28.12 | 39.38 | 32.85 | 33.96 | 29.07 | 23.34 | 39.31 | |
| Mem0LLM Backbone=Qwen2.5-7B-Instruct2026.04 | 27.03 | 23.31 | 37.11 | 18.69 | 13.92 | 30.85 | 12.89 | 9.94 | 26.04 | 25.37 | 20.35 | 21.81 | 24.25 | 20.11 | 32.05 | |
| HippoRAG 2LLM Backbone=Qwen2.5-7B-Instruct2026.04 | 27 | 18.34 | 67.78 | 19.3 | 15.32 | 54.96 | 16.7 | 13.14 | 53.13 | 26.04 | 18.4 | 42.68 | 24.75 | 17.48 | 59.26 | |
| Mem0LLM Backbone=Qwen2.5-32B-Instruct2026.04 | 24.3 | 20.42 | 36.75 | 17.51 | 12.46 | 29.43 | 14.81 | 11.48 | 31.25 | 32.8 | 27.2 | 31.15 | 24.24 | 19.81 | 33.88 |