Latest Test-Time Compute Research Papers

The newest Test-Time Compute papers from across the field — arXiv, NeurIPS, CVPR, Nature, and more — refreshed daily and ranked by relevance. Distill AI tracks Test-Time Compute so you don’t have to: get the standout work delivered to your inbox every morning, with 2-sentence summaries and the option to chat with any paper.

Recent papers

• Legibility is Not Interpretability: Comparing Judged and Actual Importance in Chain-Of-Thought Reasoning
  Kevin Du, Alexander Hoyle, Laura Ruis, Acyr Locatelli · arXiv · Sep 3, 2026

  Reasoning traces from chain-of-thought models appear to offer a legible window into how a model arrives at its answer. A growing body of work treats them as such, using LLM judges to diagnose errors, evaluate faithfulness, and provide step-…

• Discriminative World Models for Web Agents
  Kelvin Li, Dhruv Pendharkar, Anish Pahilajani, Chuyi Shang et al. · arXiv · Sep 2, 2026

  Recent web agents use world models for test-time action selection by sampling candidate actions, predicting the resulting web states, and ranking them with a ranker model or a Process Reward Model (PRM). These world models are typically tra…

• Cliff: Learning Process Rewards from the First Mistake
  Peixuan Han, Runhui Wang, Ketan Ramaneti, Jie Hao et al. · arXiv · Sep 2, 2026

  Reinforcement learning with verifiable rewards (RLVR) has emerged as a powerful paradigm for large language model (LLM) post-training, but its reliance on coarse outcome rewards leads to limited guidance on intermediate reasoning processes.…

• LoRA-TSD: Tangent-Space Spectral Descent for LoRA via Muon-Style Updates
  Dmitrii Andriianov, Andrey Veprikov, Aleksandr Beznosikov · arXiv · Sep 2, 2026

  Low-rank adaptation (LoRA) is the standard way to fine-tune large models, yet when its two factors are trained independently, the update ignores the geometry of the low-rank weight change it induces. We introduce LoRA-TSD, an optimizer that…

• DIRECT: When and Where Should You Allocate Test-Time Compute in Embodied Planners?
  Jadelynn Dao, Milan Ganai, Yasmina Abukhadra, Ajay Sridhar et al. · X-Reason Poster · Aug 23, 2026

  Vision-Language Models (VLMs) are increasingly deployed as high-level planners for embodied agents, with an emerging strategy of scaling test-time compute to improve capability. However, we observe that doing so increases latency, token usa…

• On the Fragility of Self-Improving Agents: Variance, Task Order, and Underspecification
  Qinyuan Ye, Yu Li, Yada Pruksachatkun, Jiaxin Zhang et al. · arXiv · Aug 18, 2026

  Memory-based self-improving agents--those that learn from an online stream of tasks and improve over time by maintaining a textual memory bank--have shown great promise in recent literature. However, the reliability aspects of these methods…

• More Correct Mass, Worse Answers: Why Power Sampling Can Fail and How to Fix It
  Haohui Yang, Jiaxing Sun, Xiujun Ma · arXiv · Aug 14, 2026

  Power Sampling sharpens a language model's distribution over complete generation trajectories, offering a verifier-free way to improve reasoning at inference time. It also has the potential to serve as a general-purpose front end for a broa…

• Test-Time Scaling in Reasoning LLMs: Inference Regimes, Evaluation, and Reproducibility
  Mohsen Hariri, Weicong Chen, Nahal Shahini, Vikash Singh et al. · arXiv · Aug 4, 2026

  Large language models can solve substantially harder reasoning problems with more inference-time compute. The term "test-time scaling," however, now covers diverse inference algorithms that extend deliberation along a single trajectory, sam…

• Test-Time Scaling via Error Localization
  Rajiv Shailesh Chitale, Rahul Madhavan, Taneesh Gupta, Deepanway Ghosal et al. · arXiv · Jul 23, 2026

  Scaling inference-time computation has emerged as a reliable method to improve the performance of large language models on complex reasoning and programming tasks. However, standard approaches such as independent sampling and sequential mul…

• DIRECT: When and Where Should You Allocate Test-Time Compute in Embodied Planners?
  RSS26-W: FM4RoboPlan Oral · Jul 8, 2026

  Vision-Language Models (VLMs) are increasingly deployed as high-level planners for embodied agents, with an emerging strategy of scaling test-time compute to improve capability. However, we observe that doing so increases latency, token usa…

• DIRECT: When and Where Should You Allocate Test-Time Compute in Embodied Planners?
  Jadelynn Dao, Milan Ganai, Yasmina Abukhadra, Ajay Sridhar et al. · RSS SemRob 2026 Poster · Jun 30, 2026

  Vision-Language Models (VLMs) are increasingly deployed as high-level planners for embodied agents, with an emerging strategy of scaling test-time compute to improve capability. However, we observe that doing so increases latency, token usa…

• Allocation, Not Volume: Test-Time Compute for Agentic Forecasting
  Atin Aboutorabi, Gaetan de Rassenfosse, Nicolas Flammarion, Maksym Andriushchenko · Forecast@ICML26 Oral · Jun 11, 2026

  Test-time compute scaling has been studied extensively in verifiable domains such as math and code; how to spend an inference budget for forecasting future events, where no test-time verifier exists, is far less studied. We compare three mu…

• Test-Time Compute Games
  Ander Artola Velasco, Dimitrios Rontogiannis, Stratis Tsirtsis, Manuel Gomez Rodriguez · ICLR 2026 Workshop AIMS · Mar 2, 2026

  Test-time compute has emerged as a promising strategy to enhance the reasoning abilities of large language models (LLMs). However, this strategy has in turn increased how much users pay cloud-based providers offering LLM-as-a-service, since…

• Mode-conditioning unlocks superior test-time compute scaling
  Chen Henry Wu, Sachin Goyal, Aditi Raghunathan · ICLR 2026 Poster · Jan 26, 2026

  Parallel sampling is essential to test-time scaling and reinforcement learning (RL), but its effectiveness is sharply limited by diversity collapse, where models concentrate on a few modes and repeated samples produce the same mistakes. We …

• ParaThinker: Native Parallel Thinking as a New Paradigm to Scale LLM Test-time Compute
  Hao Wen, Yifan Su, Feifei Zhang, Yunxin Liu et al. · arXiv.org · Aug 30, 2025

  Recent advances in Large Language Models (LLMs) have been driven by test-time compute scaling - a strategy that improves reasoning by generating longer, sequential thought processes. While effective, this approach encounters a significant b…

• Inverse Scaling in Test-Time Compute
  Aryo Pradipta Gema, Alexander Hägele, Runjin Chen, Andy Arditi et al. · Trans. Mach. Learn. Res. · Jul 19, 2025

  We construct evaluation tasks where extending the reasoning length of Large Reasoning Models (LRMs) deteriorates performance, exhibiting an inverse scaling relationship between test-time compute and accuracy. Our evaluation tasks span four …

• Multi-Agent Verification: Scaling Test-Time Compute with Multiple Verifiers
  Shalev Lifshitz, Sheila A. McIlraith, Yilun Du · COLM 2025 · Jul 8, 2025

  By utilizing more computational resources at test-time, large language models (LLMs) can improve without additional training. One common strategy uses *verifiers* to evaluate candidate outputs. In this work, we propose a novel scaling dimen…

• Rank1: Test-Time Compute for Reranking in Information Retrieval
  Orion Weller, Kathryn Ricci, Eugene Yang, Andrew Yates et al. · COLM 2025 · Jul 8, 2025

  We introduce Rank1, the first reranking model trained to take advantage of test-time compute. Rank1 demonstrates the applicability within retrieval of using a reasoning language model (i.e. OpenAI's o1, Deepseek's R1, etc.) for distillation…

• Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMs
  Mohammad Ali Alomrani, Yingxue Zhang, Derek Li, Qianyi Sun et al. · arXiv.org · Jul 2, 2025

  Large language models (LLMs) have rapidly progressed into general-purpose agents capable of solving a broad spectrum of tasks. However, current models remain inefficient at reasoning: they apply fixed inference-time compute regardless of ta…

• MiniMax-M1: Scaling Test-Time Compute Efficiently with Lightning Attention
  MiniMax Aili Chen, Aonian Li, Bangwei Gong, Binyan Jiang et al. · arXiv.org · Jun 16, 2025

  We introduce MiniMax-M1, the world's first open-weight, large-scale hybrid-attention reasoning model. MiniMax-M1 is powered by a hybrid Mixture-of-Experts (MoE) architecture combined with a lightning attention mechanism. The model is develo…

• Scaling Test-time Compute for LLM Agents
  King Zhu, Hanhao Li, Siwei Wu, Tianshun Xing et al. · arXiv.org · Jun 15, 2025

  Scaling test time compute has shown remarkable success in improving the reasoning abilities of large language models (LLMs). In this work, we conduct the first systematic exploration of applying test-time scaling methods to language agents …

• e3: Learning to Explore Enables Extrapolation of Test-Time Compute for LLMs
  Amrith Rajagopal Setlur, Matthew Y. R. Yang, C. Snell, Jeremy Greer et al. · arXiv.org · Jun 10, 2025

  Test-time scaling offers a promising path to improve LLM reasoning by utilizing more compute at inference time; however, the true promise of this paradigm lies in extrapolation (i.e., improvement in performance on hard problems as LLMs keep…

• ScaleRTL: Scaling LLMs with Reasoning Data and Test-Time Compute for Accurate RTL Code Generation
  Chenhui Deng, Yun-Da Tsai, Guan-Ting Liu, Zhongzhi Yu et al. · Workshop on Machine Learning for CAD · Jun 5, 2025

  Recent advances in large language models (LLMs) have enabled near-human performance on software coding benchmarks, but their effectiveness in RTL code generation remains limited due to the scarcity of high-quality training data. While prior…

• Optimizing Test-Time Compute via Meta Reinforcement Finetuning
  Yuxiao Qu, Matthew Y. R. Yang, Amrith Setlur, Lewis Tunstall et al. · ICML 2025 poster · May 1, 2025

  Training models to efficiently use test-time compute is crucial for improving the reasoning performance of LLMs. While current methods mostly do so via fine-tuning on search traces or running RL against the 0/1 outcome reward, do these appr…

• GenPRM: Scaling Test-Time Compute of Process Reward Models via Generative Reasoning
  Jian Zhao, Runze Liu, Kaiyan Zhang, Zhimu Zhou et al. · AAAI Conference on Artificial Intelligence · Apr 1, 2025

  Recent advancements in Large Language Models (LLMs) have shown that it is promising to utilize Process Reward Models (PRMs) as verifiers to enhance the performance of LLMs. However, current PRMs face three key challenges: (1) limited proces…

• Optimizing Test-Time Compute via Meta Reinforcement Fine-Tuning
  Yuxiao Qu, Matthew Y. R. Yang, Amrith Rajagopal Setlur, Lewis Tunstall et al. · International Conference on Machine Learning · Mar 10, 2025

  Training models to effectively use test-time compute is crucial for improving the reasoning performance of LLMs. Current methods mostly do so via fine-tuning on search traces or running RL with 0/1 outcome reward, but do these approaches ef…

• Multi-Agent Verification: Scaling Test-Time Compute with Multiple Verifiers (Abridged)
  Shalev Lifshitz, Sheila A. McIlraith, Yilun Du · SSI-FM Poster · Mar 8, 2025

  By utilizing more computational resources at test-time, large language models (LLMs) can improve without additional training. One common strategy uses *verifiers* to evaluate candidate outputs. In this work, we propose a novel scaling dimen…

• Optimizing Test-Time Compute via Meta Reinforcement Finetuning
  Yuxiao Qu, Matthew Y. R. Yang, Lewis Tunstall, Edward Emanuel Beeching et al. · SSI-FM Poster · Mar 8, 2025

  Training models to efficiently use test-time compute is crucial for improving the reasoning performance of LLMs. While current methods mostly do so via fine-tuning on search traces or running RL against the 0/1 outcome reward, do these appr…

• Optimizing Test-Time Compute via Meta Reinforcement Finetuning
  Yuxiao Qu, Matthew Y. R. Yang, Amrith Setlur, Lewis Tunstall et al. · ICLR 2025 FM-Wild Workshop · Mar 6, 2025

  Training models to efficiently use test-time compute is crucial for improving the reasoning performance of LLMs. While current methods mostly do so via fine-tuning on search traces or running RL against the 0/1 outcome reward, do these appr…

• Scaling Test-Time Compute Without Verification or RL is Suboptimal
  Amrith Setlur, Nived Rajaraman, Sergey Levine, Aviral Kumar · ICLR 2025 Workshop VerifAI Oral · Mar 6, 2025

  Despite substantial improvements in LLM capabilities by scaling test-time compute, an ongoing debate in the community is how it should be scaled up so as to enable continued and efficient improvements with scaling. There are largely two app…