# THREAD: Thinking Deeper with Recursive Spawning

arXiv: https://arxiv.org/abs/2405.17402 · HTML: https://mlsanity.com/paper/2405.17402/

## Key facts

- Cited by 25 papers in the map (data through 14 Oct 2026).
- Placed in Test-time compute and efficient thinking, one of 17 directions in LLM reasoning.
- Builds on 9 of its 39 references that are in the map.
- First posted on arXiv 27 May 2024.

## Placed in

- LLM reasoning / [Test-time compute and efficient thinking](https://mlsanity.com/lr/test-time-compute-and-efficient-thinking/)

## Builds on (most cited first, 9 of 9)

- [HuggingFace's Transformers: State-of-the-art Natural Language Processing](https://mlsanity.com/paper/1910.03771/) (2019)
- [Language Agent Tree Search Unifies Reasoning Acting and Planning in Language Models](https://mlsanity.com/paper/2310.04406/) (2023)
- [Adaptive Computation Time for Recurrent Neural Networks](https://mlsanity.com/paper/1603.08983/) (2016)
- [ADaPT: As-Needed Decomposition and Planning with Language Models](https://mlsanity.com/paper/2311.05772/) (2023)
- [RAP: Retrieval-Augmented Planning with Contextual Memory for Multimodal LLM Agents](https://mlsanity.com/paper/2402.03610/) (2024)
- [AutoGuide: Automated Generation and Selection of Context-Aware Guidelines for Large Language Model Agents](https://mlsanity.com/paper/2403.08978/) (2024)
- [PEARL: Prompting Large Language Models to Plan and Execute Actions Over Long Documents](https://mlsanity.com/paper/2305.14564/) (2023)
- [TDAG: A Multi-Agent Framework based on Dynamic Task Decomposition and Agent Generation](https://mlsanity.com/paper/2402.10178/) (2024)
- [Natural Language Embedded Programs for Hybrid Language Symbolic Reasoning](https://mlsanity.com/paper/2309.10814/) (2023)

## Built on by (most cited first, 25 of 25)

- [Advances and Challenges in Foundation Agents: From Brain-Inspired Intelligence to Evolutionary, Collaborative, and Safe Systems](https://mlsanity.com/paper/2504.01990/) (2025)
- [Recursive Language Models](https://mlsanity.com/paper/2512.24601/) (2025)
- [Learning Adaptive Parallel Reasoning with Language Models](https://mlsanity.com/paper/2504.15466/) (2025)
- [AOrchestra: Automating Sub-Agent Creation for Agentic Orchestration](https://mlsanity.com/paper/2602.03786/) (2026)
- [StateAct: Enhancing LLM Base Agents via Self-prompting and State-tracking](https://mlsanity.com/paper/2410.02810/) (2024)
- [ReCode: Unify Plan and Action for Universal Granularity Control](https://mlsanity.com/paper/2510.23564/) (2025)
- [AgentProg: Empowering Long-Horizon GUI Agents with Program-Guided Context Management](https://mlsanity.com/paper/2512.10371/) (2025)
- [Recursive Models for Long-Horizon Reasoning](https://mlsanity.com/paper/2603.02112/) (2026)
- [ThreadWeaver: Adaptive Threading for Efficient Parallel Reasoning in Language Models](https://mlsanity.com/paper/2512.07843/) (2025)
- [ROVER: Recursive Reasoning Over Videos with Vision-Language Models for Embodied Tasks](https://mlsanity.com/paper/2508.01943/) (2025)
- [Uno-Orchestra: Parsimonious Agent Routing via Selective Delegation](https://mlsanity.com/paper/2605.05007/) (2026)
- [Fine-tuning with RAG for Improving LLM Learning of New Skills](https://mlsanity.com/paper/2510.01375/) (2025)
- [Recursive Agent Optimization](https://mlsanity.com/paper/2605.06639/) (2026)
- [Hindsight Hint Distillation: Scaffolded Reasoning for SWE Agents from CoT-free Answers](https://mlsanity.com/paper/2605.11556/) (2026)
- [Reflect before Act: Proactive Error Correction in Language Models](https://mlsanity.com/paper/2509.18607/) (2025)
- [HyMem: Hierarchical Context Management for Long-Horizon Agents via Information Isolation](https://mlsanity.com/paper/2608.15703/) (2026)
- [Self-Manager: Parallel Agent Loop for Long-form Deep Research](https://mlsanity.com/paper/2601.17879/) (2026)
- [Spawn Freely, Act Sparingly: Progressive Risk Vesting for Recursive LLM-Agent Trees](https://mlsanity.com/paper/2609.01035/) (2026)
- [A Training-free LLM Framework with Interaction between Contextually Related Subtasks in Solving Complex Tasks](https://mlsanity.com/paper/2503.23053/) (2025)
- [RVLM: Recursive Vision-Language Models with Adaptive Depth](https://mlsanity.com/paper/2603.24224/) (2026)
- [H-Probes: Extracting Hierarchical Structures From Latent Representations of Language Models](https://mlsanity.com/paper/2605.00847/) (2026)
- [Second Thought: Reasoning in Parallel as LLM Agents Act and Observe](https://mlsanity.com/paper/2608.13667/) (2026)
- [Recursive Language Models Generalize Out of Domain](https://mlsanity.com/paper/2609.20831/) (2026)
- [Hermes: Learning Contextual Reasoning Unlocks Test-Time Scaling](https://mlsanity.com/paper/2609.38332/) (2026)
- [Self-Evaluating Recursive Agents](https://mlsanity.com/paper/2610.04902/) (2026)

## Abstract (arXiv)

Large language models (LLMs) have shown impressive capabilities across diverse settings, but still struggle as the length and complexity of the context increases. To address this challenge, we propose Thinking Recursively and Dynamically (ThReaD). THREAD frames model generation as a thread of execution that, based on the context, can run to completion or dynamically spawn new threads. By spawning, threads can offload work (e.g., thinking, retrieving information) to child threads, which only return tokens needed for the parent thread to do its work. In effect, this enables the model to adapt, as needed, the amount of intermediate work used to produce tokens. We apply THREAD in the settings of LLM task solving and question answering, where the dynamic threading allows the model to recursively decompose the given task or question into progressively simpler sub-problems that can be solved by separate child threads. We test THREAD, implemented using a few-shot learning approach, on diverse benchmarks for agent tasks and data-grounded question answering. THREAD achieves state-of-the-art performance with GPT-4 and GPT-3.5 on these benchmarks, including ALFWorld, TextCraft, and WebShop, along with two new benchmarks, DataCommons QA and MIMIC-III ICU QA. In addition, THREAD outperforms existing frameworks by 10% to 50% absolute points with smaller models, including Llama-3-8b and CodeLlama-7b.