LaPha: A New Approach to Math Reasoning in Hyperbolic Space
jiqizhixin · x · 2026-07-13
A Shanghai AI for Science team and collaborators proposed **LaPha**: a method for training mathematical reasoning agents in the **Poincaré hyperbolic latent space**. ### Core Ideas - Models the search process as a tree "growing outward" in hyperbolic geometry, leveraging the exponential growth of space with radius for exploration. - Provides **dense process rewards** based on the geodesic distance to "rule-verified correctness," helping the agent learn reasoning trajectories more stably. - Integrates a lightweight **value head** for self-guided search expansion during testing. ### Performance - Boosted **Qwen2.5-Math-1.5B** on **MATH500** from **66.0%** to **88.2%**. - Using value-head-guided search, **LaPha-1.5B** achieved **56.7%** on **AIME'24**. - **LaPha-7B** reached **60.0%** on **AIME'24** and **53.3%** on **AIME'25**. The post also mentions another paper, **T\\*: Progressive Block Scaling for Masked Diffusion Language Models Through Trajectory Aware Reinforcement Learning**, along with links to the paper and an explanation.
More from Models
- Claude 20x users report sharply tighter limits and faster quota burn — MarcJSchmidt · 2026-07-21
- Cola launches July, the latest model jokingly billed as “second only to Fable” — oran_ge · 2026-07-21
- Kimi K3 looks stronger and about 5× cheaper on a frontend dashboard task — OwariDa · 2026-07-21
- Last Week in AI recap: Anthropic’s $65B round, IPO filing, and Microsoft’s MAI push — Last Week in AI · 2026-07-21
- A user says Claude 4.6 felt worse yesterday and asks whether model quality can drift over time — Rahios · 2026-07-21
- Kimi K3 hits 89.4% peak on software tasks while Fable 5 is slightly steadier — FinanceYF5 · 2026-07-21