Recurrent-depth latent reasoning scales test-time compute without extra params
gordic_aleksa · x · 2026-09-17
Gordic Aleksa highlights the underappreciated paper "Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach" (Geiping, Goldstein et al.)
- Iterates a recurrent block to unroll arbitrary depth at test time, reasoning implicitly in latent space instead of emitting more CoT tokens
- Needs no specialized training data, works with small context windows, captures reasoning hard to verbalize
- Proof-of-concept at 3.5B params / 800B tokens shows dramatic benchmark gains, compute load equivalent to 50B params
- Positioned as a third scaling axis beyond parameter count and chain-of-thought length; rumors say frontier labs already use it
Related event: Looped Transformers Gain Attention as a Way to Scale Test-Time Compute(2 posts)→
More from Research
- Xiaomi MiMo's CodeSpecWeave trains coding agents on interdependent specs for long-horizon coding — XiaomiMiMo · 2026-11-02
- Mistral Large 4 writes a quantum chemistry SCF solver from scratch with one short prompt — MistralAI · 2026-11-02
- Hunting for the most concrete "curing cancer with AI" roadmap of 2025 — iskander · 2026-11-02
- New theory paper proves linear decay is the unique minimax-optimal stepsize schedule for SGD — konstmish · 2026-11-02
- Karpathy's nanochat: the best ChatGPT $100 can buy, near 60k GitHub stars — karpathy · 2026-11-02
- After reading 46 resources on agent evals, one author explains why her 95% score was misleading — blaizedsouza · 2026-11-02