A 1.4B and a 2.6B looped (weight-tied, recurrent-depth) language model trained on 7.7T tokens match or exceed several 4B–8B transformer baselines on selected reasoning benchmarks.
Ultra-fineweb: Efficient data filtering and verification for high-quality llm training data
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2025 2representative citing papers
citing papers explorer
-
Scaling Latent Reasoning via Looped Language Models
A 1.4B and a 2.6B looped (weight-tied, recurrent-depth) language model trained on 7.7T tokens match or exceed several 4B–8B transformer baselines on selected reasoning benchmarks.
- The Ratchet Effect in Silico: How Interaction Drives Cumulative Intelligence in Large Language Models