Papers worth knowing this week

Recent work extends recurrent latent reasoning to large-scale language models, multimodal generation, and hybrid reasoning.

🔭 Paper scouting · Trending papers
on Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach
by golombi · November 6, 2026 · edit this page

Papers worth knowing

  1. Dory1.0-4B: A Looped Normalized-GPT (2026): Introduces a 4-billion-parameter recurrent language model that uses elastic depth, difficulty-aware loop allocation, and reinforcement learning to improve reasoning performance while reducing generated tokens.

  2. Latent Reasoning in Continuous Space for Unified Multimodal Models (2026): Introduces LARC, a framework that enables multimodal models to perform reasoning directly in continuous latent space, improving image generation quality and enabling test-time compute scaling.

  3. Learning When to Think Aloud in Latent Reasoning (2026): Introduces VITAL, a hybrid reasoning framework that selectively generates important intermediate tokens while performing the remaining reasoning in latent space, improving the balance between reasoning accuracy and token efficiency.



Enjoy Reading This Article?

Here are some more articles you might like to read next:

  • Critique: Parameter-efficient, but is it compute-efficient?
  • Reproducing the exit rule and the latent trajectories on Huginn-0125
  • Learning about Information Bottleneck
  • Key points regarding Latenct reasoning
  • A better graph instead of Table 4