Knowledge Base Question Answering on GrailQA Compositional (dev)
87F1 ScoreGAPD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GAPDLLM=Llama-3.1-8B, Method Category=Fine-tune-based Methods2026.05 | 87 | 82.4 | |
| KBQA-R1LLM=Llama-3.1-8B, Method Category=Fine-tune-based Methods2026.05 | 82.5 | 78 | |
| DecAFLLM=T5-large, Method Category=Fine-tune-based Methods2026.05 | 79.8 | 71.5 | |
| KBQA-o1LLM=Llama-3.1-8B, Method Category=Fine-tune-based Methods2026.05 | 77.6 | 76.3 | |
| TIARALLM=T5-large, Method Category=Fine-tune-based Methods2026.05 | 74.8 | 66.4 | |
| KnowCoder-A1LLM=Qwen2.5-Coder-7B, Method Category=Fine-tune-based Methods2026.05 | 70.6 | — | |
| RnG-KBQALLM=T5-large, Method Category=Fine-tune-based Methods2026.05 | 68.9 | 61.7 | |
| KBQA-R1 HarnessLLM=Claude 4.6 Sonnet, Method Category=Prompting Methods2026.05 | 68.8 | 62.9 | |
| KBQA-R1 HarnessLLM=Kimi-K2.5, Method Category=Prompting Methods2026.05 | 65.7 | 57.1 | |
| KBQA-R1 HarnessLLM=gpt-5.3-codex, Method Category=Prompting Methods2026.05 | 65.6 | 56.2 | |
| KBQA-R1 HarnessLLM=GLM-5, Method Category=Prompting Methods2026.05 | 64.4 | 51.4 | |
| ARG-KBQALLM=GPT-3.5-turbo, Method Category=Prompting Methods2026.05 | 41.8 | 36.4 | |
| KB-CoderLLM=GPT-3.5-turbo, Method Category=Prompting Methods2026.05 | 38.6 | 34.5 | |
| KB-BINDERLLM=Codex-davinci-002, Method Category=Prompting Methods2026.05 | 36.6 | 33.9 |