Training Verifiers to Solve Math Word Problems
State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. 5K high quality linguistically diverse grade school math word problems.
Also cited · not yet reviewed (1)
- GPT-32020 · cited 2×, 1 in Method“Finetuning, our baseline method, uses the same language modeling objective as the generative pretraining in GPT-3 (Brown et al. 2020).”From this paper · §Methods
Led to
- PaLM2022 · cited 7דSeveral recent papers have shown that large language models can achieve significant accuracy improvements by generating intermediate reasoning steps before generating the final answer (Nye et al. 2021; Cobbe et al. 2021;…”From PaLM · §Evaluation
- U-PaLM2022 · cited 2דWe use the GSM8K (Cobbe et al. 2021), BBH (Suzgun et al. 2022), StrategyQA (Geva et al. 2021) and CommonsenseQA (Talmor et al. 2019) benchmarks.”From U-PaLM · §Experiments
- Flan-T5 / Flan-PaLM2022 · cited 6דThese nine datasets include tasks such as arithmetic reasoning (Cobbe et al. 2021), multi-hop reasoning (Geva et al. 2021), and natural language inference (Camburu et al. 2020).”From Flan-T5 / Flan-PaLM · §unknown section
Abstract
State-of-the-art language models can match human performance on many tasks, but they still struggle to robustly perform multi-step mathematical reasoning. To diagnose the failures of current models and support research, we introduce GSM8K, a dataset of 8.5K high quality linguistically diverse grade school math word problems. We find that even the largest transformer models fail to achieve high test performance, despite the conceptual simplicity of this problem distribution. To increase performance, we propose training verifiers to judge the correctness of model completions. At test time, we generate many candidate solutions and select the one ranked highest by the verifier. We demonstrate that verification significantly improves performance on GSM8K, and we provide strong empirical evidence that verification scales more effectively with increased data than a finetuning baseline.