# Researchers test how language models handle numerical formats in word problems

Digest AI · Research · published 2026-09-23T04:00:00Z

Canonical: https://digestai.news/story/researchers-test-how-language-models-handle-numerical-formats-in-word

## Summary

A new paper on arXiv examines whether language models consistently answer numerical word problems regardless of how quantities are expressed. The authors created 3,600 exact-rational problems and 8,600 prompts across five transformation types, then tested five open-weight models. After normalizing answers, the models scored between 0.969 and 0.996 on canonical accuracy but dropped to 0.848–0.981 on orbit correctness and invariance. Mistral Small 4 struggled with unit-converted inputs, scoring 0.699 and producing 265 errors off by exact powers of ten.

The study also found that representation consensus did not outperform paraphrase consensus in a 9,000-call experiment. The paper includes a benchmark, evaluation records, and raw responses, all available in an ancillary archive.

## Key points

- Researchers generated 3,600 exact-rational and 8,600 prompts testing numerical format invariance in language models
- Five open-weight models scored 0.969–0.996 on canonical accuracy but dropped to 0.848–0.981 on orbit correctness
- Mistral Small 4 scored 0.699 on unit-converted inputs, with 265 errors differing by exact powers of ten

## Why it matters

The findings highlight persistent flaws in how models handle numerical reasoning, especially unit conversions, which could affect real-world applications like finance or science where precision matters.

## Sources

1. [Same Quantity, Different Answer: Numerical Representation Invariance in Language Models](https://arxiv.org/abs/2609.25009) (arXiv cs.CL, 2026-09-23, primary source)

## Cite

Digest AI, "Researchers test how language models handle numerical formats in word problems", 23 September 2026, https://digestai.news/story/researchers-test-how-language-models-handle-numerical-formats-in-word

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/researchers-test-how-language-models-handle-numerical-formats-in-word.json