BDH-CQ：以遞迴潛在推理進行上下文學習

摘要

我們提出 BDH-CQ，一種結合上下文學習與遞迴潛在推理的推理模型。推論時呈現的輸入會持續更新模型的遞迴記憶；模型接著在高維潛在空間中透過迭代計算來解出查詢，而不以語言表述其中間推理過程。我們在公開的 ARC-AGI-1 評估集上評估此模型，並使用受控的類 ARC 干預來研究模型從示範中學到了什麼、如何一致地應用推斷出的變換，以及哪些概念仍然難以處理。一個具有 150M 參數的設定，在每任務計算推理成本為 0.0007 美元的情況下，達到了 29.5% 的 pass@2。此操作點突破了先前報告的 ARC-AGI-1 成本-準確率帕累托前沿，在基準測試成本效率上樹立了新的最先進水準。

English

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.