# AdaPonderLM: Gated Pondering Language Models with Token-Wise Adaptive Depth

arXiv: https://arxiv.org/abs/2603.01914 · HTML: https://mlsanity.com/paper/2603.01914/

## Key facts

- Cited by 10 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 41 of its 48 references that are in the map.
- First posted on arXiv 2 Mar 2026.

## Placed in

- LLM reasoning / [Test-time compute and efficient thinking](https://mlsanity.com/lr/test-time-compute-and-efficient-thinking/)
- Efficient LLM inference and serving / [Early exit and edge inference](https://mlsanity.com/inference/early-exit-and-edge-inference/)

## Builds on (most cited first, 30 of 41)

- [Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge](https://mlsanity.com/paper/1803.05457/) (2018)
- [HellaSwag: Can a Machine Really Finish Your Sentence?](https://mlsanity.com/paper/1905.07830/) (2019)
- [FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness](https://mlsanity.com/paper/2205.14135/) (2022)
- [OPT: Open Pre-trained Transformer Language Models](https://mlsanity.com/paper/2205.01068/) (2022)
- [PIQA: Reasoning about Physical Commonsense in Natural Language](https://mlsanity.com/paper/1911.11641/) (2019)
- [The Pile: An 800GB Dataset of Diverse Text for Language Modeling](https://mlsanity.com/paper/2101.00027/) (2020)
- [The LAMBADA dataset: Word prediction requiring a broad discourse context](https://mlsanity.com/paper/1606.06031/) (2016)
- [STaR: Bootstrapping Reasoning With Reasoning](https://mlsanity.com/paper/2203.14465/) (2022)
- [RACE: Large-scale ReAding Comprehension Dataset From Examinations](https://mlsanity.com/paper/1704.04683/) (2017)
- [Crowdsourcing Multiple Choice Science Questions](https://mlsanity.com/paper/1707.06209/) (2017)
- [TinyLlama: An Open-Source Small Language Model](https://mlsanity.com/paper/2401.02385/) (2024)
- [Deep Equilibrium Models](https://mlsanity.com/paper/1909.01377/) (2019)
- [BranchyNet: Fast Inference via Early Exiting from Deep Neural Networks](https://mlsanity.com/paper/1709.01686/) (2017)
- [Adaptive Computation Time for Recurrent Neural Networks](https://mlsanity.com/paper/1603.08983/) (2016)
- [Reducing Transformer Depth on Demand with Structured Dropout](https://mlsanity.com/paper/1909.11556/) (2019)
- [Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach](https://mlsanity.com/paper/2502.05171/) (2025)
- [DeeBERT: Dynamic Early Exiting for Accelerating BERT Inference](https://mlsanity.com/paper/2004.12993/) (2020)
- [LayerSkip: Enabling Early Exit Inference and Self-Speculative Decoding](https://mlsanity.com/paper/2404.16710/) (2024)
- [Quiet-STaR: Language Models Can Teach Themselves to Think Before Speaking](https://mlsanity.com/paper/2403.09629/) (2024)
- [BERT Loses Patience: Fast and Robust Inference with Early Exit](https://mlsanity.com/paper/2006.04152/) (2020)
- [Think before you speak: Training Language Models With Pause Tokens](https://mlsanity.com/paper/2310.02226/) (2023)
- [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)
- [Conditional Computation in Neural Networks for faster models](https://mlsanity.com/paper/1511.06297/) (2015)
- [Scaling Latent Reasoning via Looped Language Models](https://mlsanity.com/paper/2510.25741/) (2025)
- [Adaptive Neural Networks for Efficient Inference](https://mlsanity.com/paper/1702.07811/) (2017)
- [Mixture-of-Recursions: Learning Dynamic Recursive Depths for Adaptive Token-Level Computation](https://mlsanity.com/paper/2507.10524/) (2025)
- [The Right Tool for the Job: Matching Model and Instance Complexities](https://mlsanity.com/paper/2004.07453/) (2020)
- [Soft Thinking: Unlocking the Reasoning Potential of LLMs in Continuous Concept Space](https://mlsanity.com/paper/2505.15778/) (2025)

## Built on by (most cited first, 10 of 10)

- [The Latent Space: Foundation, Evolution, Mechanism, Ability, and Outlook](https://mlsanity.com/paper/2604.02029/) (2026)
- [Fixed-Point Reasoners: Stable and Adaptive Deep Looped Transformers](https://mlsanity.com/paper/2606.18206/) (2026)
- [Solve the Loop: Attractor Models for Language and Reasoning](https://mlsanity.com/paper/2605.12466/) (2026)
- [Bridging the Gap Between Latent and Explicit Reasoning with Looped Transformers](https://mlsanity.com/paper/2606.31779/) (2026)
- [Adaptive Depth in Looped Transformers: Diagnosing Learned Halting Gates and Trajectory Readouts](https://mlsanity.com/paper/2607.20519/) (2026)
- [Depth-adaptive Inference of Looped Language Models via Continuous Depth Batching](https://mlsanity.com/paper/2608.09444/) (2026)
- [Improving Test-Time Scaling with Adaptive Looped Transformers](https://mlsanity.com/paper/2609.35748/) (2026)
- [HALO: Hybrid Adaptive Latent Reasoning for Language Models](https://mlsanity.com/paper/2607.08775/) (2026)
- [WhiteMatter: All-to-All Cross-Layer Connections via KV Source Mixing](https://mlsanity.com/paper/2608.18486/) (2026)
- [Enabling Dynamic Computation in Looped LMs](https://mlsanity.com/paper/2610.09013/) (2026)

## Abstract (arXiv)

Test-time scaling via recurrent/iterative Transformers enables large language models to spend more computation at inference, but most pretrained recurrent LMs run a fixed number of iterations, wasting compute on easy tokens and lacking token-wise adaptivity. Following the core idea of Adaptive Computation Time(ACT) and Early Exit(EE), we propose AdaPonderLM, a self-supervised recurrent language model that learns token-wise early exiting during pretraining without manually tuned per-token/per-layer pruning ratios. AdaPonderLM uses iteration-specific MLP gates with a monotonic halting mask to decide when each token stops recurring, and introduces a KV reuse mechanism that reuses cached key/value states for halted tokens, ensuring train--test consistency and practical acceleration. Across Pythia backbones from 70M to 410M (pretraining) and up to 2.8B (continued pretraining), AdaPonderLM reduces inference compute at about 10% while maintaining comparable language modeling perplexity and competitive downstream accuracy. Our analysis shows the learned gates allocate more computation to high-NLL (hard) tokens, exhibiting adaptive computation time behavior in a fully self-supervised setting. Meanwhile, under iso-FLOPs, the learned halting policy consistently outperforms fixed pruning, showing AdaPonderLM allocates compute to the right tokens rather than just reducing average depth.