|
--- |
|
license: openrail |
|
--- |
|
|
|
MathT5-large is a version of FLAN-T5-large fine-tuned for 25 epochs on 15K (LaTeX) synthetic mathematical derivations (containing 5 - 9 equations), generated using a symbolic solver. |
|
|
|
It outperforms GPT-4 and ChatGPT (paper link soon) on a derivation generation task in ROUGE, BLEU, BLEURT, and GLEU, and can somewhat generalise to out-of-distribution examples. |
|
|
|
It struggles with out-of-vocabulary symbols, but was trained on 155 physics symbols. |
|
|
|
|
|
An example prompt: |
|
|
|
Given \cos{(q)} = \theta{(q)}, |
|
then derive - \sin{(q)} = \frac{d}{d q} \theta{(q)}, |
|
then obtain (- \sin{(q)})^{q} (\frac{d}{d q} \cos{(q)})^{q} = (- \sin{(q)})^{2 q} |
