paper-with-me

홈 › Papers

TEMPO: Scaling Test-time Training for Large Reasoning Models

2026-04-21 · Qingyang Zhang, Xinke Kong, Haitao Wu, Qinghua Hu, Minghao Wu, Baosong Yang, Yu Cheng, Yun Luo, Ganqu Cui, Changqing Zhang arxiv

Test-time training (TTT) adapts model parameters on unlabeled test instances during inference time, which continuously extends capabilities beyond the reach of offline training. Despite initial gains, existing TTT methods for LRMs plateau quickly and do not benefit from additional test-time compute. Without external calibration, the self-generated reward signal increasingly drifts as the policy model evolves, leading to both performance plateaus and diversity collapse. We propose TEMPO, a TTT framework that interleaves policy refinement on unlabeled questions with periodic critic recalibration on a labeled dataset. By formalizing this alternating procedure through the Expectation-Maximization (EM) algorithm, we reveal that prior methods can be interpreted as incomplete variants that omit the crucial recalibration step. Reintroducing this step tightens the evidence lower bound (ELBO) and enables sustained improvement. Across diverse model families (Qwen3 and OLMO3) and reasoning tasks, TEMPO improves OLMO3-7B on AIME 2024 from 33.0% to 51.1% and Qwen3-14B from 42.3% to 65.8%, while maintaining high diversity.

📄 PDF Abstract BibTeX arXiv:2604.19295

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Temporal Scaling Law for Large Language Models

2024-04-27 · Yizhe Xiong, Xiansheng Chen, Xin Ye, Hui Chen 외

Recently, Large Language Models (LLMs) have been widely adopted in a wide range of tasks, leading to increasing attention towards the research on how scaling LLMs affects their performance. Existing works, termed Scaling…

Position

Efficient Test-Time Scaling via Temporal Reasoning Aggregation

2026-04-19 · Jiakun Li, Xingwei He, Kefan Li, Hongzheng Chai 외 arxiv

Test-time scaling improves the reasoning performance of large language models but often results in token-inefficient overthinking, where models continue reasoning beyond what is necessary for a correct answer. Existing d…

ARISE: An Adaptive Resolution-Aware Metric for Test-Time Scaling Evaluation in Large Reasoning Models

2025-10-07 · Zhangyue Yin, Qiushi Sun, Zhiyuan Zeng, Zhiyuan Yu 외 arxiv

Test-time scaling has emerged as a transformative paradigm for enhancing the performance of large reasoning models, enabling dynamic allocation of computational resources during inference. However, as the landscape of re…

Mathematical ReasoningCode Generation

Towards Neural Scaling Laws for Foundation Models on Temporal Graphs

2024-06-14 · Razieh Shirzadkhani, Tran Gia Bao Ngo, Kiarash Shamsi, Shenyang Huang 외

The field of temporal graph learning aims to learn from evolving network data to forecast future interactions. Given a collection of observed temporal graphs, is it possible to predict the evolution of an unseen network …

Graph LearningGraph Property PredictionProperty Prediction

Timely Machine: Awareness of Time Makes Test-Time Scaling Agentic

2026-01-23 · Yichuan Ma, Linyang Li, Yongkang chen, Peiji Li 외 arxiv

As large language models (LLMs) increasingly tackle complex reasoning tasks, test-time scaling has become critical for enhancing capabilities. However, in agentic scenarios with frequent tool calls, the traditional gener…

Reinforcement Learning