# HALO: Hybrid Adaptive Latent Reasoning for Language Models

arXiv: https://arxiv.org/abs/2607.08775 · HTML: https://mlsanity.com/paper/2607.08775/

## 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 29 of its 30 references that are in the map.
- First posted on arXiv 3 May 2026.

## 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, 29 of 29)

- [Chain-of-Thought Prompting Elicits Reasoning in Large Language Models](https://mlsanity.com/paper/2201.11903/) (2022)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [Self-Refine: Iterative Refinement with Self-Feedback](https://mlsanity.com/paper/2303.17651/) (2023)
- [GPQA: A Graduate-Level Google-Proof Q&A Benchmark](https://mlsanity.com/paper/2311.12022/) (2023)
- [Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters](https://mlsanity.com/paper/2408.03314/) (2024)
- [MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark](https://mlsanity.com/paper/2406.01574/) (2024)
- [Enhancing Chat Language Models by Scaling High-quality Instructional Conversations](https://mlsanity.com/paper/2305.14233/) (2023)
- [Training Large Language Models to Reason in a Continuous Latent Space](https://mlsanity.com/paper/2412.06769/) (2024)
- [Phi-4 Technical Report](https://mlsanity.com/paper/2412.08905/) (2024)
- [Universal Transformers](https://mlsanity.com/paper/1807.03819/) (2018)
- [Phi-4-Mini Technical Report: Compact yet Powerful Multimodal Language Models via Mixture-of-LoRAs](https://mlsanity.com/paper/2503.01743/) (2025)
- [Adaptive Computation Time for Recurrent Neural Networks](https://mlsanity.com/paper/1603.08983/) (2016)
- [Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach](https://mlsanity.com/paper/2502.05171/) (2025)
- [LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding](https://mlsanity.com/paper/2404.16710/) (2024)
- [Think before you speak: Training Language Models With Pause Tokens](https://mlsanity.com/paper/2310.02226/) (2023)
- [Mixture-of-Depths: Dynamically allocating compute in transformer-based language models](https://mlsanity.com/paper/2404.02258/) (2024)
- [Reasoning with Latent Thoughts: On the Power of Looped Transformers](https://mlsanity.com/paper/2502.17416/) (2025)
- [Lessons from the Trenches on Reproducible Evaluation of Language Models](https://mlsanity.com/paper/2405.14782/) (2024)
- [PonderNet: Learning to Ponder](https://mlsanity.com/paper/2107.05407/) (2021)
- [Mixture-of-Recursions: Learning Dynamic Recursive Depths for Adaptive Token-Level Computation](https://mlsanity.com/paper/2507.10524/) (2025)
- [LoopFormer: Elastic-Depth Looped Transformers for Latent Reasoning via Shortcut Modulation](https://mlsanity.com/paper/2602.11451/) (2026)
- [Inner Thinking Transformer: Leveraging Dynamic Depth Scaling to Foster Adaptive Internal Thinking](https://mlsanity.com/paper/2502.13842/) (2025)
- [PonderLM: Pretraining Language Models to Ponder in Continuous Space](https://mlsanity.com/paper/2505.20674/) (2025)
- [AdaPonderLM: Gated Pondering Language Models with Token-Wise Adaptive Depth](https://mlsanity.com/paper/2603.01914/) (2026)
- [Early Exit Is a Natural Capability in Transformer-based Models: An Empirical Study on Early Exit without Joint Optimization](https://mlsanity.com/paper/2412.01455/) (2024)
- [Understanding Dynamic Compute Allocation in Recurrent Transformers](https://mlsanity.com/paper/2602.08864/) (2026)
- [Tiny Autoregressive Recursive Models](https://mlsanity.com/paper/2603.08082/) (2026)
- [Thinking in Latents: Adaptive Anchor Refinement for Implicit Reasoning in LLMs](https://mlsanity.com/paper/2603.15051/) (2026)
- [Pretraining with Token-Level Adaptive Latent Chain-of-Thought](https://mlsanity.com/paper/2602.08220/) (2026)

## Abstract (arXiv)

We study how to improve a frozen pretrained language model with a small amount of adaptive extra computation. A simple approach is to add additional refinement steps on top of the backbone hidden states, but fixed extra refinement can be wasteful: a one-step refinement head may be too weak, while forcing a second full-sequence refinement step everywhere can increase compute without improving transfer. We introduce HALO, a hybrid adaptive latent-refinement method that combines a coarse refinement stage with selective second-stage latent refinement on a subset of tokens chosen by token scoring and monotonic token halting. On the main public benchmark comparison built from MMLU-Pro and GPQA-Diamond, HALO achieves the best overall average among the paper-facing methods, outperforming the frozen backbone, fixed-1, and fixed-2. Internal analysis further shows that HALO reaches nearly the same token-accuracy level as fixed-2 while using fewer average applied refine steps than fixed-1 and far fewer than fixed-2. These results suggest that the key advantage is not simply more refinement, but a better allocation of refinement: HALO achieves the strongest paper-facing result while also using less measured controller compute than either fixed baseline.