Question Answering on Dynamic Event Benchmark Contemporary
96.3C1 AccuracyChronosGPT-4o
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| ChronosGPT-4oBackbone=GPT-4o, Method Paradigm=Chronos (Ours)2026.04 | 96.3 | 66.51 | 71.19 | |
| ChronosClaudeBackbone=Claude, Method Paradigm=Chronos (Ours)2026.04 | 93.76 | 66.97 | 74.58 | |
| ReAct RAGGPT-4oBackbone=GPT-4o, Method Paradigm=Retrieval-based Methods2026.04 | 93.37 | 56.42 | 45.76 | |
| ChronosLLaMA-3.1Backbone=LLaMA-3.1, Method Paradigm=Chronos (Ours)2026.04 | 92.4 | 63.76 | 61.02 | |
| ReAct RAGClaudeBackbone=Claude, Method Paradigm=Retrieval-based Methods2026.04 | 91.62 | 59.63 | 47.46 | |
| Vanilla RAGGPT-4oBackbone=GPT-4o, Method Paradigm=Retrieval-based Methods2026.04 | 86.74 | 50.46 | 25.42 | |
| Vanilla RAGClaudeBackbone=Claude, Method Paradigm=Retrieval-based Methods2026.04 | 84.02 | 44.5 | 23.73 | |
| Vanilla RAGLLaMA-3.1Backbone=LLaMA-3.1, Method Paradigm=Retrieval-based Methods2026.04 | 80.31 | 39.91 | 16.95 | |
| LoRA FTLLaMA-3.1Backbone=LLaMA-3.1, Method Paradigm=Parametric Updating Methods2026.04 | 71.15 | 4.13 | 10.17 | |
| ReAct RAGLLaMA-3.1Backbone=LLaMA-3.1, Method Paradigm=Retrieval-based Methods2026.04 | 64.52 | 47.25 | 32.2 | |
| GPT-4oBackbone=GPT-4o, Method Paradigm=Direct Generation2026.04 | 14.81 | 3.21 | 5.08 | |
| LLaMA-3.1Backbone=LLaMA-3.1, Method Paradigm=Direct Generation2026.04 | 11.31 | 1.38 | 1.69 | |
| WISELLaMA-3.1Backbone=LLaMA-3.1, Method Paradigm=Parametric Updating Methods2026.04 | 11.11 | 1.83 | 1.69 | |
| ClaudeBackbone=Claude, Method Paradigm=Direct Generation2026.04 | 10.53 | 2.75 | 3.39 | |
| MEMITLLaMA-3.1Backbone=LLaMA-3.1, Method Paradigm=Parametric Updating Methods2026.04 | 7.99 | 1.38 | 0 | |
| ROMELLaMA-3.1Backbone=LLaMA-3.1, Method Paradigm=Parametric Updating Methods2026.04 | 2.92 | 0.92 | 0 |