BDH-CQ: In-Context Learning with Recurrent Latent Reasoning
BDH-CQ combines in-context learning with recurrent latent reasoning, achieving 29.5% ARC-AGI-1 pass@2 at $0.0007 per task to set a new cost-efficiency frontier.
Published Aug 10, 2026▲ 797 on Hugging FaceCode ★ 11,072arXiv ↗
Only vote on papers you've read. Sign in with GitHub to vote.
BDH-CQ establishes a striking cost-accuracy frontier with 29.5% pass@2 at $0.0007/task via recurrent latent reasoning, yet its single-point evaluation lacks variance reporting, ablations isolating recurrence from scale, and rigorous validation of its nonverbal inference mechanism.
Abstract
We introduce BDH-CQ, a reasoning model that combines in-context learning with recurrent latent reasoning. Inputs presented at inference time continuously update the model's recurrent memory; the model then solves a query through iterative computation in a high-dimensional latent space, without verbalizing its intermediate reasoning. We evaluate the model on the public ARC-AGI-1 evaluation set and use controlled ARC-like interventions to study what it learns from demonstrations, how consistently it applies an inferred transformation, and which concepts remain difficult. A 150M-parameter configuration reaches 29.5% pass@2 at a computed inference cost of \$0.0007 per task. This operating point breaks through the previously reported ARC-AGI-1 cost-accuracy Pareto frontier, establishing a new state of the art in benchmark cost efficiency.