Saved in:
Bibliographic Details
Main Authors: Liang, Kun, Bai, Clive, Xu, Xin, Tang, Chenming, Lee, Sanwoo, Liu, Weijie, Yang, Saiyong, Wu, Yunfang
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.08310
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911597039976448
author Liang, Kun
Bai, Clive
Xu, Xin
Tang, Chenming
Lee, Sanwoo
Liu, Weijie
Yang, Saiyong
Wu, Yunfang
author_facet Liang, Kun
Bai, Clive
Xu, Xin
Tang, Chenming
Lee, Sanwoo
Liu, Weijie
Yang, Saiyong
Wu, Yunfang
contents Recent Large Reasoning Models (LRMs) achieve strong performance by leveraging long-form Chain-of-Thought (CoT) reasoning, but uniformly applying overlong reasoning at inference time incurs substantial and often unnecessary computational cost. To address this, prior work explores various strategies to infer an appropriate reasoning budget from the input. However, such approaches are unreliable in the worst case, as estimating the minimal required reasoning effort is fundamentally difficult, and they implicitly fix the trade-off between reasoning cost and accuracy during training, limiting flexibility under varying deployment scenarios. Motivated by these limitations, we propose ORBIT, a controllable multi-budget reasoning framework with well-separated reasoning modes triggered by input. ORBIT employs multi-stage reinforcement learning to discover Pareto-optimal reasoning behaviors at each effort, followed by on-policy distillation to fuse these behaviors into a single unified model. Experiments show that ORBIT achieves (1) controllable reasoning behavior over multiple modes, (2) competitive reasoning density within each mode, and (3) integration of these frontier policies into a single unified student model while preserving clear mode separation and high per-mode performance.
format Preprint
id arxiv_https___arxiv_org_abs_2601_08310
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle ORBIT: On-policy Exploration-Exploitation for Controllable Multi-Budget Reasoning
Liang, Kun
Bai, Clive
Xu, Xin
Tang, Chenming
Lee, Sanwoo
Liu, Weijie
Yang, Saiyong
Wu, Yunfang
Machine Learning
Artificial Intelligence
Recent Large Reasoning Models (LRMs) achieve strong performance by leveraging long-form Chain-of-Thought (CoT) reasoning, but uniformly applying overlong reasoning at inference time incurs substantial and often unnecessary computational cost. To address this, prior work explores various strategies to infer an appropriate reasoning budget from the input. However, such approaches are unreliable in the worst case, as estimating the minimal required reasoning effort is fundamentally difficult, and they implicitly fix the trade-off between reasoning cost and accuracy during training, limiting flexibility under varying deployment scenarios. Motivated by these limitations, we propose ORBIT, a controllable multi-budget reasoning framework with well-separated reasoning modes triggered by input. ORBIT employs multi-stage reinforcement learning to discover Pareto-optimal reasoning behaviors at each effort, followed by on-policy distillation to fuse these behaviors into a single unified model. Experiments show that ORBIT achieves (1) controllable reasoning behavior over multiple modes, (2) competitive reasoning density within each mode, and (3) integration of these frontier policies into a single unified student model while preserving clear mode separation and high per-mode performance.
title ORBIT: On-policy Exploration-Exploitation for Controllable Multi-Budget Reasoning
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2601.08310