# CALR method boosts math reasoning accuracy by 26 points

> The new Continuous Anchored Latent Reasoning (CALR) technique improves performance on five math benchmarks, according to a paper posted on Oct 7 2026.

Oossa · 2026-10-07 · https://oossa.com/en/calr-method-boosts-math-reasoning-accuracy-by-26-points

Researchers led by Zhaoyang Wei introduced CALR, a way to link hidden model states with answer supervision. The method compresses derivations into latent blocks that stay tied to the original problem. Tests on five mathematical reasoning benchmarks showed CALR outperformed a similar continuous latent approach by 26 percentage points when using the same compute budget. The paper was posted on arXiv on Oct 7 2026.

## The facts

- 26.0 percentage‑point accuracy gain over a comparable continuous latent reasoning method
- Paper posted on arXiv on Oct 7 2026

## Why it matters

The improvement suggests AI systems can solve math problems more reliably without extra compute, which could benefit educational tools.

## Sources & references

1. [CALR: Continuous Anchored Latent Reasoning via Render-of-Thought Compression](https://arxiv.org/abs/2610.07175) – arXiv cs.CV, 2026-10-07

Last updated: 2026-10-07