BDH-CQ: In-Context Learning with Recurrent Latent Reasoning
On the map: LLM reasoning → Test-time compute and efficient thinking
Abstract (arXiv)
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.
Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning. Builds on 23 of its 33 references that are in the map. First posted on arXiv 10 Aug 2026.
Builds on
23 of 33 references.
Closest ancestors · 3
Same direction, newest first
- 2025Hierarchical Reasoning Model26 Jun
- 2025Reasoning with Latent Thoughts: On the Power of Looped Transformers24 Feb
- 2025Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach7 Feb
Other references · 20
- 2017Attention Is All You Needcited by 60,361
- 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
- 2020Transformers are RNNs: Fast Autoregressive Transformers with Linear Attentioncited by 2,022
- 2020Transformer Feed-Forward Layers Are Key-Value Memoriescited by 1,336
- 2021Show Your Work: Scratchpads for Intermediate Computation with Language Modelscited by 886
- 2024Training Large Language Models to Reason in a Continuous Latent Spacecited by 797
- 2025ARC-AGI-2: A New Challenge for Frontier AI Reasoning Systemscited by 190
- 2016Using Fast Weights to Attend to the Recent Pastcited by 170
- 2025Soft Thinking: Unlocking the Reasoning Potential of LLMs in Continuous Concept Spacecited by 117
- 2024ARC Prize 2024: Technical Reportcited by 104
- 2023The ConceptARC Benchmark: Evaluating Understanding and Generalization in the ARC Domaincited by 101
- 2025Reasoning by Superposition: A Theoretical Perspective on Chain of Continuous Thoughtcited by 77
8 more
- 2024Combining Induction and Transduction for Abstract Reasoningcited by 53
- 2024Addressing the Abstraction and Reasoning Corpus via Procedural Example Generationcited by 33
- 2025A Formal Comparison Between Chain of Thought and Latent Thoughtcited by 15
- 2025ARC-GEN: A Mimetic Procedural Benchmark Generator for the Abstraction and Reasoning Corpuscited by 10
- 2026HRM-Text: Efficient Pretraining Beyond Scalingcited by 9
- 2026Thinking Without Words: Efficient Latent Reasoning with Abstract Chain-of-Thoughtcited by 8
- 2026Beyond tokens: a unified framework for latent communication in LLM-based multi-agent systemscited by 7
- 2025The Dragon Hatchling: The Missing Link between the Transformer and Models of the Braincited by 6
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026Penelope: Localized Latent Recurrence for Efficient Structured Reasoning
- 2025PERK: Long-Context Reasoning as Test-Time Learning
- 2026Closing the Loop on Latent Reasoning via Test-Time Reconstruction
- 2025Encode, Think, Decode: Scaling test-time reasoning with recursive latent thoughts
- 2026Batched Contextual Reinforcement: A Task-Scaling Law for Efficient Reasoning
- 2026RecurTrace: Adaptive Latent Reasoning with Loop-Time Memory
- 2026Unlocking the Working Memory of Large Language Models for Latent Reasoning
- 2025Tiny Recursive Models on ARC-AGI-1: Inductive Biases, Identity Conditioning, and Test-Time Compute