Huginn

Summary

Huginn tracks the recurrent-depth language model using latent-space loops as test-time compute.

Role In The Wiki

Use this page as the object card for the model or method. The source page carries the evidence details, limitations, and agenda mapping.

Relation To Foundation TSFM Agenda

Use the source-level agenda mapping in huginn-2025 rather than duplicating verdict rows here. At the entity level, this page should stay as the object card; source pages carry slot-level evidence, limitations, and missing pieces.

Evidence

• Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach

Official Artifacts

• Official model: tomg-group-umd/huginn-0125
• Official code and data: seal-rg/recurrent-pretraining

Related Pages

• Looped Transformers And Test-Time Memory
• Efficient Recurrent Sequence Models
• Time-Series Scaling And Efficiency
• LLM Sleep