Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approach
Open accessThe DOI matches its Crossref record; 134 of the 134 references it cites were found in a registry or an index.
Abstract
Depth-recurrent transformers scale test-time compute by iterating a core block in latent space before emitting tokens. The geometry of these latent trajectories is a named open problem in the latent reasoning literature (survey arXiv:2604.02029, §6.3). We report a pre-registered experiment on Huginn-0125 (arXiv:2502.05171), a 3.5B-parameter depth-recurrent transformer, measuring the per-head attention power-law exponent Δ of the recurrent core at recurrence counts r ∈ {1, 2, 4, 8, 16, 32}. For natural-text inputs the median exponent decreases monotonically with recurrence (Spearman ρ = −0.94), from 0.29 at r=1 to 0.239 at r=32 — converging onto the Sachdev–Ye–Kitaev (SYK) q=4 conformal value Δ = 1/4 — while the count of SYK-near heads grows monotonically (ρ = +0.77). Random-token inputs show an architecture-driven convergence with a transient at Δ ≈ 1/2 (the q=2 prethermal plateau previously observed in training time) but no monotone SYK-near growth. A randomized-weights control freezes completely: Δ_med constant to the sixth decimal across all r, zero SYK-near heads. Inference-time recurrence therefore acts as renormalization-group flow toward the conformal fixed point, and the flow is a property of the trained model, not the iteration procedure. Together with prior results on architectural depth (10.5281/zenodo.19225996) and training time, this completes a three-axis triangulation: the conformal fixed point is an attractor of iterative attending. The published orbit/spiral phenomenology of recurrent-depth latent trajectories acquires a candidate theory — the spiral is flow near an infrared fixed point — and the geometric benefit of additional recurrence saturates when the flow arrives, consistent with convergence-based early-exit criteria. All code, pre-registration (commit 702efd95), per-head data, and experiment notes with recorded protocol deviations are in the public repository 3ld0n/attention-geometry.
Overview
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“latent CoT—recurrent reasoning in the hidden state, without token traces [64,65,66]—may soften the feedforward bottleneck, though it still iterates over activations rather than weight-level self-models.”
From Self-Referential Introspection in Large Language Models: The Critical Threshold for Recursive Self-Improvement (2026) · cites this paper as [65]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
The publisher hasn’t deposited this paper’s reference list.
Research works printed but not linked: 36
- Brandon Amos and J. Zico Kolter. 2017. OptNet: Differentiable Optimization as a Layer in Neural Networks. In International Conference on Machine Learning , pages 136-145.
- Zhangir Azerbayev, Hailey Schoelkopf, Keiran Paster, Marco Dos Santos, Stephen Marcus McAleer, Albert Q. Jiang, Jia Deng, Stella Biderman, and Sean Welleck. 2023. Llemma: An Open Language Model for Mathematics. In The Twelfth International Conference on Learning Representations .
- Shaojie Bai, J. Zico Kolter, and Vladlen Koltun. 2019. Deep Equilibrium Models. In Advances in Neural Information Processing Systems , volume 32. Curran Associates, Inc.
- Shaojie Bai, Vladlen Koltun, and J. Zico Kolter. 2022. Neural Deep Equilibrium Solvers. In International Conference on Learning Representations .
- Andrea Banino, Jan Balaguer, and Charles Blundell. 2021. PonderNet: Learning to Ponder. In 8th ICML Workshop on Automated Machine Learning (AutoML) .
- Valentino Braitenberg. 1986. Vehicles: Experiments in Synthetic Psychology . MIT press.
- Tianle Cai, Yuhong Li, Zhengyang Geng, Hongwu Peng, Jason D. Lee, Deming Chen, and Tri Dao. 2024. Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads. In Forty-First International Conference on Machine Learning . character.ai. 2024. Optimizing AI Inference at Character.AI. https://research.character.ai/optimizing-inference/
- Hantian Ding, Zijian Wang, Giovanni Paolini, Varun Kumar, Anoop Deoras, Dan Roth, and Stefano Soatto. 2024. Fewer Truncations Improve Language Modeling. In Forty-First International Conference on Machine Learning .
- Maha Elbayad, Jiatao Gu, Edouard Grave, and Michael Auli. 2019. Depth-Adaptive Transformer. In International Conference on Learning Representations .
- Ying Fan, Yilun Du, Kannan Ramchandran, and Kangwook Lee. 2025. Looped Transformers for Length Generalization. In The Thirteenth International Conference on Learning Representations .
- Jonas Geiping and Tom Goldstein. 2023. Cramming: Training a Language Model on a single GPU in one day. In Proceedings of the 40th International Conference on Machine Learning , pages 11117-11143. PMLR.
- Angeliki Giannou, Shashank Rajput, Jy-Yong Sohn, Kangwook Lee, Jason D. Lee, and Dimitris Papailiopoulos. 2023. Looped Transformers as Programmable Computers. In Proceedings of the 40th International Conference on Machine Learning , pages 11398-11442. PMLR.
- Tamir David Hay and Lior Wolf. 2023. Dynamic Layer Tying for Parameter-Efficient Transformers. In The Twelfth International Conference on Learning Representations .
- Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2021a. Measuring massive multitask language understanding. Proceedings of the International Conference on Learning Representations (ICLR) .
- Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. 2021b. Measuring Massive Multitask Language Understanding. In International Conference on Learning Representations .
- Pavel Izmailov, Dmitrii Podoprikhin, Timur Garipov, Dmitry Vetrov, and Andrew Gordon Wilson. 2018. Averaging weights leads to wider optima and better generalization: 34th Conference on Uncertainty in Artificial Intelligence 2018, UAI 2018. 34th Conference on Uncertainty in Artificial Intelligence 2018, UAI 2018 , pages 876-885.
- Angelos Katharopoulos, Apoorv Vyas, Nikolaos Pappas, and François Fleuret. 2020. Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention. In Proceedings of the 37th International Conference on Machine Learning , pages 5156-5165. PMLR.
- Diederik P. Kingma and Jimmy Ba. 2015. Adam: A Method for Stochastic Optimization. In International Conference on Learning Representations (ICLR) , San Diego.
- Zhenzhong Lan, Mingda Chen, Sebastian Goodman, Kevin Gimpel, Piyush Sharma, and Radu Soricut. 2019. ALBERT: A Lite BERT for Self-supervised Learning of Language Representations. In International Conference on Learning Representations .
- Yann LeCun. 2022. A Path Towards Autonomous Machine Intelligence. Preprint , Version 0.9.2:62.
- Yann LeCun and Fu Jie Huang. 2005. Loss functions for dis- criminative training of energy-based models. In AISTATS 2005 https://nyuscholars.nyu.edu/en/publications/loss-functions-for-discriminative-training-of-energy-based-models
- Yaniv Leviathan, Matan Kalman, and Yossi Matias. 2023. Fast Inference from Transformers via Speculative Decoding. In Proceedings of the 40th International Conference on Machine Learning , pages 19274-19286. PMLR.
- Mrinal Mathur, Barak A. Pearlmutter, and Sergey M. Plis. 2024. MIND over Body: Adaptive Thinking using Dynamic Computation. In The Thirteenth International Conference on Learning Representations .
- Neel Nanda, Lawrence Chan, Tom Lieberum, Jess Smith, and Jacob Steinhardt. 2022. Progress measures for grokking via mechanistic interpretability. In The Eleventh International Conference on Learning Representations .
- Keiran Paster, Marco Dos Santos, Zhangir Azerbayev, and Jimmy Ba. 2023. OpenWebMath: An Open Dataset of High-Quality Mathematical Web Text. In The Twelfth International Conference on Learning Representations .
- Victor Sanh, Albert Webson, Colin Raffel, Stephen Bach, Lintang Sutawika, Zaid Alyafeai, Antoine Chaffin, Arnaud Stiegler, Arun Raja, Manan Dey, M. Saiful Bari, Canwen Xu, Urmish Thakker, Shanya Sharma Sharma, Eliza Szczechla, Taewoon Kim, Gunjan Chhablani, Nihal Nayak, Debajyoti Datta, and 21 others. 2021. Multitask Prompted Training Enables Zero-Shot Task Generalization. In International Conference on Learning Representations .
- Sunny Sanyal, Atula Tejaswi Neerkaje, Jean Kaddour, Abhishek Kumar, and sujay sanghavi. 2024. Early weight averaging meets high learning rates for LLM pre-training. In First Conference on Language Modeling .
- Avi Schwarzschild. 2023. Deep Thinking Systems: Logical Extrapolation with Recurrent Neural Networks . Ph.D. thesis, University of Maryland, College Park, College Park.
- Avi Schwarzschild, Eitan Borgnia, Arjun Gupta, Furong Huang, Uzi Vishkin, Micah Goldblum, and Tom Goldstein. 2021b. Can You Learn an Algorithm? Generalizing from Easy to Hard Problems with Recurrent Networks. In Advances in Neural Information Processing Systems , volume 34, pages 6695-6706. Curran Associates, Inc.
- Yu Sun, Xiaolong Wang, Zhuang Liu, John Miller, Alexei Efros, and Moritz Hardt. 2020. Test-Time Training with SelfSupervision for Generalization under Distribution Shifts. In Proceedings of the 37th International Conference on Machine Learning , pages 9229-9248. PMLR.
- Ilya Sutskever, Geoffrey E Hinton, and Graham W Taylor. 2008. The Recurrent Temporal Restricted Boltzmann Machine. In Advances in Neural Information Processing Systems , volume 21. Curran Associates, Inc.
- Mitchell Wortsman, Tim Dettmers, Luke Zettlemoyer, Ari S. Morcos, Ali Farhadi, and Ludwig Schmidt. 2023b. Stable and low-precision training for large-scale vision-language models. In Thirty-Seventh Conference on Neural Information Processing Systems .
- Yuhuai Wu, Markus Norman Rabe, DeLesley Hutchins, and Christian Szegedy. 2022. Memorizing Transformers. In International Conference on Learning Representations .
- Liu Yang, Kangwook Lee, Robert D. Nowak, and Dimitris Papailiopoulos. 2024a. Looped Transformers are Better at Learning Learning Algorithms. In The Twelfth International Conference on Learning Representations .
- Longhui Yu, Weisen Jiang, Han Shi, Jincheng Yu, Zhengying Liu, Yu Zhang, James Kwok, Zhenguo Li, Adrian Weller, and Weiyang Liu. 2023. MetaMath: Bootstrap Your Own Mathematical Questions for Large Language Models. In The Twelfth International Conference on Learning Representations .
- Yifan Zhang, Yifan Luo, Yang Yuan, and Andrew C. Yao. 2024c. Autonomous Data Selection with Language Models for Mathematical Texts. In ICLR 2024 Workshop on Navigating and Addressing Data Problems for Foundation Models .
Cited by
Items 201–300 of 361
Training Continuous Chain of Thought Models: A Tale of Two RegimesPreprint
Varun, Y., He, H., & Choi, E. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.16972 · Cited by 2
Latent Thought Credit: Multi-Answer Credit Assignment for Latent ReasoningPreprint
Zhao, X., Zhang, L., Xu, Z., Chen, Y., Zeng, W., Zhao, S., & Li, Q. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.01593 · Cited by 2
Lam-Muir, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.03263 · Cited by 2
LoopVL: Recurrent Visual IntelligencePreprint
Qian, Z., Gong, Z., Xu, Z., Li, H., Wang, Z., Liu, H., Qian, L., Pei, X., Xu, P., Luo, F., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.38426 · Cited by 2
Scaling Laws for Looped Mixture of ExpertsPreprint
Chen, Y., Goyal, A., & Krishnamoorthi, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.40316 · Cited by 2
Decoding Looped Transformers Better for (Almost) FreePreprint
Liu, W., Zheng, H., Chen, T., Dilip, R., Bai, R. H., Jiao, Y., Wang, Y., & Zhang, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.02185 · Cited by 3
Hybrid Latent Attention for Looped Language ModelsPreprint
Chen, Y., Zhang, S., Wang, N., Kang, F., & Jia, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.07940 · Cited by 2
Asymmetric Looped Transformers: Fixed Global Representations for an Iterative DecoderPreprint
Liang, C., Zhang, T., Wenjie, D. U., Wu, J., Ma, C., Pu, S., Zhang, Y., Wang, Y., & Lin, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.28489 · Cited by 2
Energy-Based Transformers are Scalable Learners and Thinkers
Gladstone, A., Nanduru, G., Islam, M. M., Han, P., Ha, H., Chadha, A., Du, Y., Ji, H., Li, J., & Iqbal, T. (2025) · arXiv · doi:10.48550/arxiv.2507.02092 · Cited by 1
Zhao, W., Zhai, H., Qiu, X., Qi, Z., Li, S., & Zhu, L. (2025) · Conference on Empirical Methods in Natural Language Processing (EMNLP) · doi:10.18653/v1/2025.emnlp-main.690 · Cited by 3
Inference-Time Scaling for Visual AutoRegressive Modeling by Searching Representative Samples
Tang, W., Wan, X., Li, S., & Wang, X. (2026) · Lecture notes in computer science · doi:10.1007/978-981-95-5699-1_28 · Cited by 2
HAPO: Training Language Models to Reason Concisely via History-Aware Policy Optimization
Huang, C., Zhang, Z., & Cardie, C. (2026) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v40i37.40373 · Cited by 3
Depth-selective LoRA: Augmenting mixture-of-recursions with sparse low-rank adaptation
Zhu, Z., Feng, L., & Zhao, Q. (2026) · Expert Systems with Applications · doi:10.1016/j.eswa.2026.132390 · Cited by 1
LoopCoder: Scaling Code Intelligence via Looped Language Models
Yang, J., Zhang, W., Guo, S., Li, Y., Chai, L., Ye, Z., Liu, S., Song, Y., Wu, J., Liu, C., et al. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.796 · Cited by 4
Polymorphic Universal Transformer
Chen, Y., Gao, Z., Xiao, Y., Liu, J. K., Xinyu, Y., Luo, Y., Luo, H., Ye, Z., Liu, T., Tao, R., et al. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.1809 · Cited by 1
Sparse Growing Transformer: Training-Time Sparse Depth Allocation via Progressive Attention Looping
Chen, Y., Chen, Y., Yang, Y., Shang, J., Zhang, Z., Zhang, Z., Nie, S., Wang, S., Sun, Y., Wu, H., et al. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.307 · Cited by 1
Reich, A., Thoms, C., & Schrimpf, T. (2026) · Communication Methods and Measures · doi:10.1080/19312458.2026.2693637 · Cited by 1
TextNCA: Neural Cellular Automata for Language Modeling via Hierarchical Local Attention
Mittal, A., Anand, A., Kumar, A., Kukreja, D., Prasad, K., Dulloo, S., Cambria, E., Liu, T., Wang, Z., & Shah, R. R. (2026) · Findings of the Association for Computational Linguistics: EMNLP (Findings of EMNLP) · doi:10.18653/2026.findings-emnlp.1413 · Cited by 1
Agentic AI Systems in Radiology: Design, Applications, Evaluation, and Challenges
Blüthgen, C., Veen, D. V., Truhn, D., Kather, J. N., Moor, M., Polacin, M., Chaudhari, A. S., Frauenfelder, T., Langlotz, C. P., Krauthammer, M. O., et al. (2026) · Radiology · doi:10.1148/radiol.253226 · Cited by 1
Toward Efficient and Faithful Reasoning in Large Language ModelsPreprint
Schneider, L., Muller, A., & Gerhardt, M. (2025) · Preprints.org · doi:10.20944/preprints202507.1531.v1 · Cited by 1
Accelerating Large Language Model Inference via Early-Exiting AlgorithmsPreprint
Bae, S. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2509.05915 · Cited by 1
Selection, Reflection and Self-Refinement: Revisit Reasoning Tasks via a Causal Lens
Yun-long, D., Sun, B., Li, Y., Kong, L., Tang, Z., Zhang, K., & Chen, G. (2025) · International Conference on Learning Representations · doi:10.48550/arxiv.2510.08222 · Cited by 1
Catch Your Breath: Adaptive Computation for Self-Paced Sequence ProductionPreprint
Galashov, A., Jones, M., Ke, R., Cao, Y., Nagarajan, V., & Mozer, M. C. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2510.13879 · Cited by 1
Emergent Search and Backtracking in Latent Reasoning ModelsPreprint
Cui, J. J., & Ye, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.08100 · Cited by 1
COMiT: Learning Structured Visual Tokens through Sequential CommunicationPreprint
Davtyan, A., Şahin, Y. H., Haghighi, Y., Stapf, S., Acuaviva, P., Alahi, A., & Favaro, P. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2602.20731 · Cited by 1
Beyond Test-Time Memory: State-Space Optimal Control for LLM ReasoningPreprint
Wang, P., Yang, S., Wang, X., Xiao, T., Liu, X., Yu, C., Lou, Y., Li, P., Wang, Z., Lin, M. C., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.09221 · Cited by 1
Jaber, J., & Jaber, O. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.02051 · Cited by 1
Rawat, S., Flek, L., Mai, F., & Corrêa, N. K. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.21454 · Cited by 1
Ulterior Motives: Detecting Misaligned Reasoning in Continuous Thought ModelsPreprint
Ramjee, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.23460 · Cited by 1
Bifurcation Models: Learning Set-Valued Solution Maps with Weight-Tied DynamicsPreprint
Jore, C., & Liu, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.07277 · Cited by 1
Unified Synthesis of Compositional Speech and Sound from Free-Form Text PromptsPreprint
Wang, Y., Wang, X., Cheng, X., Chen, Y., & Song, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.28063 · Cited by 1
LatentMT: Machine Translation with Latent ReasoningPreprint
Chen, W.-R., Magdy, S. M., Zhang, C., Zhu, W., Wang, Z., & Abdul-Mageed, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.18618 · Cited by 1
Kirin, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.18553 · Cited by 1
The Ignition Index: Measuring Global Workspace Dynamics in Language ModelsPreprint
Rahbar, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.05160 · Cited by 1
Think Deep, Speak Once: Relit, A Recursive Latent Implicit Transformer FrameworkPreprint
Panwar, A., Singh, M., & Bansal, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.08113 · Cited by 1
Elastic Token Compression for Pixel-Space Diffusion TransformersPreprint
Zamfir, E., Reisswig, C., Wu, Z., Xian, Y., & Timofte, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.29281 · Cited by 1
Diffusion as a Training Curriculum for Timestep-Free Iterative ReasoningPreprint
Drozdova, M., Sirbu, A., Miotti, P., Obryk, R., Etcheverry, M., Niklasson, E., & Richards, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.01449 · Cited by 1
Prediction Dynamics in Depth-Recurrent Language ModelsPreprint
Luo, X., & Yu, F. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.21383 · Cited by 1
When Recursive Models Finish ComputingPreprint
Krishna, H., Singh, S., Ebert, S., & Sun, H.-Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.26487 · Cited by 1
Self-citationGao, Q., Yang, Z., & Chen, X. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.28448 · Cited by 1
Beyond the Training Horizon: Mechanisms and Limits of Length Generalization in Looped TransformersPreprint
Liang, J., Jin, X., & Pan, L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.33144 · Cited by 1
Shared Weights, Selected Computations: How Looped Transformers Route What Each Loop DoesPreprint
Wu, J., Hu, Y., & Zhang, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.39892 · Cited by 1
What Limits Recursive Reasoning Models: Optimization, Architecture and Test-Time ScalingPreprint
Shakhvalieva, Y., Kharchev, D., Bezrukov, V., Fedorova, I., Bocharov, D., Oseledets, I., & Ternovskii, V. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.39967 · Cited by 1
Recurrent Looped TransformerPreprint
Zhang, Y., Feng, J., Qin, S., Kung, S.-Y., & Wang, M. (2026) · arXiv (Cornell University) · doi:10.13140/rg.2.2.23966.50249 · Cited by 1
Wang, X., Wang, S., & Liu, B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.39740 · Cited by 1
ALoDLM: Adaptively Looped Diffusion Language ModelsPreprint
Fang, L., Li, Z., Kim, Y., Zhao, T., Koner, R., Wu, J., Xu, L., Chen, X., Xu, X., Zhang, Z., et al. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.04198 · Cited by 1
Truncate or Not? A Scaling Study of Looped Transformers with Truncated BackpropagationPreprint
Xiong, X., Chen, S., Shao, Z., Su, Y., Jin, W., Luo, P., Yang, Z., & Liu, X. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.26030 · Cited by 1
LSTM-Style Looped Transformers with Shared KV CachesPreprint
Kirsch, A. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2610.33360 · Cited by 1
LatentEvolve: Self-Evolving Test-Time Scaling in Latent Space
Zhang, G., Meng, F., Wan, G., Li, Z., Wang, K., Yin, Z., Bai, L., & Yan, S. (2025) · arXiv · doi:10.48550/arxiv.2509.24771
Beyond Symmetric Agents: Cognitive Diversity and Multi-Agent Debate in Small Language Models
Ferreira, L., Liu, G., & Zheng, K. (2026) · arXiv · doi:10.48550/arxiv.2609.35875
Scaling Latent Reasoning via Looped Language Models
Zhu, R.-J., Wang, Z., Hua, K., Zhang, T., Li, Z., Que, H., Wei, B., Wen, Z., Yin, F., Xing, H., et al. (2025) · arXiv · doi:10.48550/arxiv.2510.25741
ThinkPrune: Pruning Long Chain-of-Thought of LLMs via Reinforcement Learning
Hou, B., Zhang, Y., Ji, J., Liu, Y., Qian, K., Andreas, J., & Chang, S. (2025) · arXiv · doi:10.48550/arxiv.2504.01296
SealQA: Raising the Bar for Reasoning in Search-Augmented Language Models
Pham, T., Nguyen, N., Zunjare, P., Chen, W., Tseng, Y.-M., & Vu, T. (2025) · arXiv · doi:10.48550/arxiv.2506.01062
Feedback Descent: Open-Ended Text Optimization via Pairwise Comparison
Lee, Y., Boen, J., & Finn, C. (2025) · arXiv · doi:10.48550/arxiv.2511.07919
Harnessing the Reasoning Economy: A Survey of Efficient Reasoning for Large Language Models
Wang, R., Wang, H., Xue, B., Li, Y., Pang, J., Liu, S., Chen, Y., Qiu, J., Wong, D. F., Chen, G., et al. (2025) · arXiv · doi:10.48550/arxiv.2503.24377
From Tokens to Thought: Efficient Architectures for Machine ReasoningPreprint
Chen, A., O’Connor, L., Al-Mansouri, F., Gerhardt, M., Iyer, R., Hernández, S., Weber, D., Lee, M.-J., & Dubois, C. (2025) · doi:10.36227/techrxiv.174803687.70425427/v1
A Theory of Inference Compute Scaling: Reasoning through Directed Stochastic Skill SearchPreprint
Ellis-Mohr, A. R., Nayak, A. K., & Varshney, L. R. (2025) · SSRN Electronic Journal · doi:10.2139/ssrn.5288682 · Cited by 2
From Long to Short: LLMs Excel at Trimming Own Reasoning ChainsPreprint
Han, W., Zhan, G., Yu, S., Wang, C., & Hooi, B. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2509.06174
In Dialogue with Intelligence: Toward Insightful Co-AugmentationPreprint
Vasilaki, E. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2505.22767
Multi-Path Collaborative Reasoning via Reinforcement LearningPreprint
Lv, J., Zhou, Y., Zhu, Z. D., Wang, X., Huang, G., & Lv, J. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2512.01485
Learning Temporal Abstractions via Variational Homomorphisms in Option-Induced Abstract MDPsPreprint
Li, C., Zhang, Y., Lv, H., Cao, Q., Xue, C., & He, X. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2507.16473
CHOIR: Collaborative Harmonization fOr Inference RobustnessPreprint
Dong, X., Wang, C., Teleki, M., Bismay, M., & Caverlee, J. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2510.22475
AdaPerceiver: Transformers with Adaptive Width, Depth, and TokensPreprint
Jajal, P., Eliopoulos, N. J., Chou, B. S.-H., Thiruvathukal, G. K., Lu, Y.-H., & Davis, J. C. (2025) · arXiv (Cornell University) · doi:10.48550/arxiv.2511.18105
From Fuzzy to Exact: The Halo Architecture for Infinite-Depth Reasoning via Rational ArithmeticPreprint
Ren, H.-S. (2026) · arXiv.org · doi:10.48550/arxiv.2601.18702 · Cited by 1
Umar, A. (2026) · Engineering Archive (engrXiv) · doi:10.31224/6512
Cheng, J., Gao, M., Wang, Z., Zhu, X., Liu, Z., Li, W., (7081), W. L., & Wu, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.03782
Zhang, Y., Wang, X. E., Chen, H., Pan, J., Li, Q., Zhang, J., Jiang, J., Mei, H., & Zhu, W. (2026) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v40i19.38680
Pu, M., Lim, M. K., Chong, C. Y., & Loy, C. C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2603.29057 · Cited by 1
C-voting: Confidence-Based Test-Time Voting without Explicit Energy FunctionsPreprint
Kubo, K., Kamiya, S., Koyama, M., Hayashi, K., Iwasawa, Y., & Matsuo, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2604.13521 · Cited by 2
Olcott, J. P. (2026) · ScholarsArchive (Brigham Young University)
SMolLM: Small Language Models Learn Small Molecular GrammarPreprint
Jindal, A., & Ju, H. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.06322
Continuous Latent Contexts Enable Efficient Online Learning in TransformersPreprint
Anand, E., Ateyeh, A., Cao, X., & Dabagia, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.09867 · Cited by 1
Block-Based Double DecodersPreprint
Labovich, A., Bradley, B., Alexander, V., & Harsha, C. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.18807
Anti Mode-Collapse in Mean-Field Transformer via Auxiliary VariablesPreprint
Imaizumi, M., Koyama, M., Isobe, N., & Hayashi, K. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2605.30229
A First-Principles Theory of Slow Thinking and Active Perception
Yang, H., Xu, Z. S., Xiong, F., & Weinan, E. (2026) · Journal of Machine Learning · doi:10.4208/jml.260213
Repeated Shared Access Enables Grokking, but Edit Propagation Depends on an Addressable MemoryPreprint
Niu, Y. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.20737
The Context-Ready TransformerPreprint
Godavarti, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2606.27538
Learning Shortcut Models for Efficient Recursive Reasoning
Shankar, S. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-srw.120
MODE-RAG: Manifold Outlier Diagnosis and Energy-based Retrieval-Augmented Generation Evaluation
Zehang, W., Dai, J., Yan, J., & Xiang, X. (2026) · Workshop on Multimodal Augmented Generation via Multimodal Retrieval (MAGMaR) · doi:10.18653/v1/2026.magmar-main.6
LEPO: Latent Reasoning Policy Optimization for Large Language Models
Zhou, Y., Yu, J., Dong, H., Hao, Z., Wang, H., Zhang, J., & Lin, Q. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.707 · Cited by 5
CHOIR: Harmonizing Structured Persona Diversity for Robust Collaborative LLM Reasoning
Dong, X., Wang, C., Teleki, M., Bismay, M., Huang, R., & Caverlee, J. (2026) · Proceedings of the conference - Association for Computational Linguistics. Meeting · doi:10.18653/v1/2026.acl-long.2175
Efficient Transformer Parameter Reuse via Zero-Token Mechanism
Li, G., Jiang, W., Shen, L., Tang, M., & Yuan, C. (2026) · Findings of the Association for Computational Linguistics: ACL (Findings of ACL) · doi:10.18653/v1/2026.findings-acl.711 · Cited by 1
HALO: Hybrid Adaptive Latent Reasoning for Language ModelsPreprint
Zhang, M. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.08775
Singh, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.11116
Mining Informative Interests via Latent Cross Reasoning for Search Enhanced Recommendation
Shi, T., Qin, W., Yu, W., Zhang, X., He, M., Jianping Fan, B., & Xu, J. (2026) · International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) · doi:10.1145/3805712.3809678 · Cited by 2
Mobius Learning: Cyclic Depth Folding in TransformersPreprint
Zhu, T. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.17843
Memoir: Should a Model Write to Its Memory While It Thinks?Preprint
Jaber, J., & Jaber, O. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2607.20792
Rajpoot, B. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7143598
A Recurrent‐Depth Multimodal Model for Skin Lesion Classification: A Feasibility Study Using Derm7pt
Yamamura, Y., Fujii, K., Kagawa, N., Nakajima, M., Morita, S., Iinuma, K., Yamamoto, K., Nakashima, C., Takada, A., & Otsuka, A. (2026) · The Journal of Dermatology · doi:10.1111/1346-8138.70442
Unifying Behavior Modeling and Semantic Generation for Generative Recommendation
Wu, B., Chen, X., Luo, Y., Ke, Y., Li, J., Zeng, K., & Ma, Q. (2026) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3770855.3818045
Hao, S., & Zhang, W. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.06014
BDH-CQ: In-Context Learning with Recurrent Latent ReasoningPreprint
Engdahl, B., Kosowski, A., Chorowski, J., Stamirowska, Z., Uznański, P., Jiang, J., Phadke, R., Kinas, R., & Zhong, R. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.09888
TANGO: Treating Tokens as OperatorsPreprint
Nunley, J. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.22117
Reward-Informed Sparse Autoencoders and the Solution-Completeness ConfoundPreprint
Nagilla, T., Jameson, A., Manta, D., & Uddin, S. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.26136
Evaluating Tiny Recursive Models Across Training for Code GenerationPreprint
Sirivella, A., Newaz, A., & Melo, G. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2608.29376
Evaluating Reasoning vs. No‑Reasoning Models on Insider‑Threat Datasets
Sun, L. (2025) · TRAP@NCI (National College of Ireland)
LoopVAE: Recurrent Depth Across Scales for Visual TokenizationPreprint
Lu, Z. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.11516
Storing Is Not Remembering: LSTM-UT and Bounded Gated Memory for Looped TransformersPreprint
Kavuncu, A., & Hafez, M. B. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.19521
The Dynamic Control PlanePreprint
Pierce, B. (2026) · Zenodo (CERN European Organization for Nuclear Research) · doi:10.5281/zenodo.22836571
Learn Your Own Thoughts: Abstract Token CurriculumPreprint
Gatmiry, K., Ghosh, A., Mirtaheri, P., Lee, J. D., Haghtalab, N., Abbe, E. A., & Bartlett, P. L. (2026) · arXiv (Cornell University) · doi:10.48550/arxiv.2609.19717
Metadata
- Authors
- Geiping, J., McLeish, S., Jain, N., Kirchenbauer, J., Singh, S., Bartoldson, B., Kailkhura, B., Bhatele, A., & Goldstein, T.
- Affiliations
- Tübingen AI Center (DE)
- University of Maryland, College Park (US)
- Lawrence Livermore National Laboratory (US)
- ORCID
- Venue
- Advances in Neural Information Processing Systems 38
- Published
- 2025
- Pages
- 46242-46293
- ISSN
- 1049-5258 (ISSN-L)
- OpenAlex
- W4407310640
- Cited by
- 10
- References
- 0
- Citations by year
- 2026: 10
- FWCI
- 9.53
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- neural information processing systems (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling
- Physical Sciences › Computer Science › Artificial Intelligence › Explainable Artificial Intelligence (XAI)
- Physical Sciences › Computer Science › Artificial Intelligence › Natural Language Processing Techniques
- Keywords
- latent reasoning, test-time compute, Sachdev-Ye-Kitaev model, conformal fixed points, prethermalization, renormalization group flow, latent trajectories
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.