Long-horizon Question Answering on LoCoMo Temporal 1.0
1,121EMMemWeaver
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MemWeaverBackbone=GPT-4o-mini2026.01 | 1,121 | 25.25 | |
| MemWeaverBackbone=Qwen2.5-1.5B2026.01 | 810 | 20.84 | |
| A-MemBackbone=GPT-4o-mini2026.01 | 374 | 19.35 | |
| MemWeaverBackbone=Llama3.2-3B2026.01 | 249 | 5.48 | |
| A-MemBackbone=Qwen2.5-1.5B2026.01 | 125 | 7.01 | |
| ReadAgentBackbone=Llama3.2-3B2026.01 | 62 | 2.33 | |
| A-MemBackbone=Llama3.2-3B2026.01 | 62 | 4.42 | |
| A-MemBackbone=Llama3.2-1B2026.01 | 62 | 7.5 | |
| MemWeaverBackbone=Llama3.2-1B2026.01 | 62 | 8.23 | |
| LoCoMoBackbone=GPT-4o-mini2026.01 | 0 | 9.87 | |
| MemoryBankBackbone=GPT-4o-mini2026.01 | 0 | 3.95 | |
| ReadAgentBackbone=GPT-4o-mini2026.01 | 0 | 4.76 | |
| LoCoMoBackbone=Llama3.2-3B2026.01 | 0 | 3.17 | |
| MemoryBankBackbone=Llama3.2-3B2026.01 | 0 | 2.3 | |
| LoCoMoBackbone=Llama3.2-1B2026.01 | 0 | 5.81 | |
| MemoryBankBackbone=Llama3.2-1B2026.01 | 0 | 1.76 | |
| ReadAgentBackbone=Llama3.2-1B2026.01 | 0 | 1.31 | |
| LoCoMoBackbone=Qwen2.5-1.5B2026.01 | 0 | 3.74 | |
| MemoryBankBackbone=Qwen2.5-1.5B2026.01 | 0 | 3.49 | |
| ReadAgentBackbone=Qwen2.5-1.5B2026.01 | 0 | 1.88 |