Mathematical Reasoning on GSM8K Zero (test) (ACC, Output Tokens)
78.43AccuracyTALE-PT-SFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TALE-PT-SFTLLM=Llama-3.1-8B-Instruct2024.12 | 78.43 | 77.85 | |
| TALE-PT-DPOLLM=Llama-3.1-8B-Instruct2024.12 | 78.41 | 113.41 | |
| Directly AnsweringLLM=Llama-3.1-8B-Instruct2024.12 | 70.32 | 13.49 | |
| Vanilla CoTLLM=Llama-3.1-8B-Instruct2024.12 | 65.04 | 251.08 |