[PR #719] Keep the letters and inside a math answer word #293

New issue

Open

opened 2026-10-26 21:05:28 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-26 21:05:28 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/deepseek-ai/DeepSeek-Coder/pull/719
Author: @SashaMIT
Created: 9/28/2026
Status: 🔄 Open

Base: main ← Head: codered-and-word

📝 Commits (1)

bfd49fa Keep the letters and inside a math answer word.

📊 Changes

2 files changed (+30 additions, -2 deletions)

View changed files

📝 Evaluation/PAL-Math/utils/parser.py (+2 -2)
➕ Evaluation/PAL-Math/utils/test_extract_answer.py (+28 -0)

📄 Description

the answer is thousand came back as thous. sand came back as s. standard came back as stard. The normalizer deletes every occurrence of the letters and, including the ones inside a word.

Those letters are now removed only when they are their own word. thousand, sand, and standard stay intact. 1000 stays 1000. \boxed{1/2} is still \frac{1}{2}.

Test plan

python -m unittest Evaluation.PAL-Math.utils.test_extract_answer (red on tip: thous vs thousand; green after: 2 tests pass)

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/deepseek-ai/DeepSeek-Coder/pull/719 **Author:** [@SashaMIT](https://github.com/SashaMIT) **Created:** 9/28/2026 **Status:** 🔄 Open **Base:** `main` ← **Head:** `codered-and-word` --- ### 📝 Commits (1) - [`bfd49fa`](https://github.com/deepseek-ai/DeepSeek-Coder/commit/bfd49faf496de72f278a38b2d27c4a1687af19ce) Keep the letters and inside a math answer word. ### 📊 Changes **2 files changed** (+30 additions, -2 deletions) <details> <summary>View changed files</summary> 📝 `Evaluation/PAL-Math/utils/parser.py` (+2 -2) ➕ `Evaluation/PAL-Math/utils/test_extract_answer.py` (+28 -0) </details> ### 📄 Description `the answer is thousand` came back as `thous`. `sand` came back as `s`. `standard` came back as `stard`. The normalizer deletes every occurrence of the letters `and`, including the ones inside a word. Those letters are now removed only when they are their own word. `thousand`, `sand`, and `standard` stay intact. `1000` stays `1000`. `\boxed{1/2}` is still `\frac{1}{2}`. ## Test plan - [x] `python -m unittest Evaluation.PAL-Math.utils.test_extract_answer` (red on tip: `thous` vs `thousand`; green after: 2 tests pass) --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi added the

pull-request

label

2026-10-26 21:05:28 +00:00