paper-with-me

홈 › Papers

Don't Overthink It: Inter-Rollout Action Agreement as a Free Adaptive-Compute Signal for LLM Agents

2026-04-09 · Khushal Sethi arxiv

Inference-time compute scaling has emerged as a powerful technique for improving the reliability of large language model (LLM) agents, but existing methods apply compute uniformly: every decision step receives the same budget regardless of its difficulty. We introduce TrACE (Trajectorical Adaptive Compute via agrEement), a training-free controller that allocates LLM calls adaptively across agent timesteps by measuring inter-rollout action agreement. At each step, TrACE samples a small set of candidate next actions and measures how consistently the model commits to the same action. High agreement signals an easy decision; the controller commits immediately. Low agreement signals uncertainty; the controller samples additional rollouts up to a configurable cap before committing to the plurality action. No learned components, no external verifier, and no human labels are required. We evaluate TrACE against greedy decoding and fixed-budget self-consistency (SC-4, SC-8) on two benchmarks spanning single-step reasoning (GSM8K, n=50) and multi-step household navigation (MiniHouse, n=30), using a Qwen 2.5 3B Instruct model running on CPU. TrACE-4 matches SC-4 accuracy while using 33% fewer LLM calls on GSM8K and 39% fewer on MiniHouse. TrACE-8 matches SC-8 accuracy with 55% fewer calls on GSM8K and 65% fewer on MiniHouse. We further show that inter-rollout agreement is a reliable signal of step-level success, validating the core hypothesis that the model's own output consistency encodes difficulty information that can be exploited without training. TrACE is the first training-free, per-timestep adaptive-compute controller for LLM agents to be evaluated on multi-step sequential decision tasks.

📄 PDF Abstract BibTeX arXiv:2604.08369

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Dynamic Rollout Editing for Reducing Overthinking in RL-Trained Reasoning Models

2026-06-16 · Zihao Wei, Wenjie Shi, Liang Pang, Jingcheng Deng 외 arxiv

Long-form chain-of-thought reasoning can improve LLM performance on complex tasks, but models often continue generating unnecessary reasoning after a correct answer has emerged. We refer to this behavior as overthinking.…

Reinforcement Learning

Shallow-Deep Networks: Understanding and Mitigating Network Overthinking

2018-10-16 · Yigitcan Kaya, Sanghyun Hong, Tudor Dumitras

We characterize a prevalent weakness of deep neural networks (DNNs)---overthinking---which occurs when a DNN can reach correct predictions before its final layer. Overthinking is computationally wasteful, and it can also…

image-classificationImage ClassificationPrediction

Implicit Compression Regularization: Concise Reasoning via Internal Shorter Distributions in RL Post-Training

2026-05-08 · Chen Wang, Hexuan Deng, Yining Zhang, Yuchen Zhang 외 arxiv

Reinforcement learning with verifiable rewards improves LLM reasoning but often induces overthinking, where models generate unnecessarily long reasoning traces. Existing methods mainly rely on length penalties or early-e…

Reinforcement Learning

NEAT: Neuron-Based Early Exit for Large Reasoning Models

2026-02-02 · Kang Liu, Yongkang Liu, Xiaocui Yang, Peidong Wang 외 arxiv

Large Reasoning Models (LRMs) often suffer from \emph{overthinking}, a phenomenon in which redundant reasoning steps are generated after a correct solution has already been reached. Existing early reasoning exit methods …

DR$^2$Seg: Decomposed Two-Stage Rollouts for Efficient Reasoning Segmentation in Multimodal Large Language Models

2026-01-15 · Yulin He, Wei Chen, Zhikang Jian, Tianhang Guo 외 arxiv

Reasoning segmentation is an emerging vision-language task that requires reasoning over intricate text queries to precisely segment objects. However, existing methods typically suffer from overthinking, generating verbos…

Multimodal ReasoningObject Localization