BDH-CQ: In-Context Learning with Recurrent Latent Reasoning

Hugging Face Daily Papers Papers

Summary

This paper introduces BDH-CQ, a 150M-parameter reasoning model that combines in-context learning with recurrent latent reasoning, achieving 29.5% pass@2 on ARC-AGI-1 at very low inference cost and establishing a new cost-accuracy frontier.

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.
Original Article
View Cached Full Text

Cached at: 08/11/26, 10:20 AM

Paper page - BDH-CQ: In-Context Learning with Recurrent Latent Reasoning

Source: https://huggingface.co/papers/2608.09888

Abstract

A 150M-parameter reasoning model using recurrent latent reasoning and in-context learning achieves a new cost-accuracy frontier on ARC-AGI-1.

We introduce BDH-CQ, a reasoning model that combinesin-context learningwithrecurrent latent reasoning. Inputs presented at inference time continuously update the model’s recurrent memory; the model then solves a query through iterative computation in ahigh-dimensional latent space, without verbalizing its intermediate reasoning. We evaluate the model on the publicARC-AGI-1evaluation 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 reportedARC-AGI-1cost-accuracy Pareto frontier, establishing a new state of the art in benchmark cost efficiency.

View arXiv pageView PDFGitHub2Add to collection

Get this paper in your agent:

hf papers read 2608\.09888

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2608.09888 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.09888 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.09888 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

Enhanced and Efficient Reasoning in Large Learning Models

arXiv cs.AI

This paper proposes a method for improving reasoning in large language models by recoding data to explicitly represent relationships, enabling efficient principled reasoning with polynomial-time learnability for relational rules, which addresses hallucinations and supports sound reasoning across multiple calls.