Slot-Filling on T-REx (Accuracy)
66.95AccuracySearChain + INFO-RAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SearChain + INFO-RAGBackbone=LLaMA-2-13B-chat2024.02 | 66.95 | — | |
| SearChainBackbone=LLaMA-2-13B-chat2024.02 | 64.58 | — | |
| Previous SOTA (DSP, Self-Ask, React)Backbone=LLaMA-2-13B-chat2024.02 | 63.1 | — | |
| CDKCBackbone=Qwen2.5-7B-Instruct, Number of Rounds=22026.02 | 57.47 | 52.43 | |
| CDKCBackbone=Qwen2.5-7B-Instruct2026.02 | 51.93 | 47.52 | |
| GRPOBackbone=Qwen2.5-7B-Instruct2026.02 | 51.7 | 46.53 | |
| BARRELBackbone=Qwen2.5-7B-Instruct2026.02 | 44.57 | 38.45 | |
| CRew-DPOBackbone=Qwen2.5-7B-Instruct2026.02 | 36.57 | 35.12 | |
| CoTBackbone=Qwen2.5-7B-Instruct2026.02 | 31.3 | 48.83 | |
| Know WhatBackbone=Qwen2.5-7B-Instruct2026.02 | 28.77 | 32.08 | |
| LLKD-SFTBackbone=Qwen2.5-7B-Instruct2026.02 | 23.63 | 30.99 | |
| Vanilla SFTBackbone=Qwen2.5-7B-Instruct2026.02 | 23.4 | 27.1 | |
| CGKEBackbone=Qwen2.5-7B-Instruct2026.02 | 23.1 | 30.4 | |
| Vanilla LLMBackbone=Qwen2.5-7B-Instruct2026.02 | 18.5 | 24.22 |