DART: Draft-Agreement Routing for Training-Free Adaptive Thinking Budgets in Hybrid Reasoning Models
cs.AI, cs.CL
Submitted: 2026-06-22
Updated: 2026-10-03
Comments: 16 pages, 4 figures, 17 tables. Accepted to EMNLP 2026 (findings). Code: https://github.com/js-lee-AI/DART
License: http://creativecommons.org/licenses/by/4.0/
The gist: Hybrid reasoning models can answer directly or spend extra tokens on extended thinking.
Terminology
Abstract
Hybrid reasoning models can answer directly or spend extra tokens on extended thinking. A practical router should choose between these modes for each query, so easy problems avoid unnecessary reasoning and hard problems receive enough budget to finish the answer. Existing routers move in this direction, but they typically require labeled training data or fix thinking budgets up front, ignoring answer-level evidence from the model itself. We introduce DART, a training-free routing framework that samples two cheap no-think drafts, accepts direct answering when the drafts agree, and predicts a thinking budget from draft entropy when they disagree. Across the main comparisons, DART preserves or improves always-thinking accuracy in most settings while reducing thinking-token use. Accuracy improves by up to +9.0 points on Olympiad-level math and by up to +22.5 points on code under execution-based equivalence, while thinking-token use drops by 32-73%. The Stage 1 signal extends across model scales (0.6B--32B), model families, and API-only hosted settings, with no labeled data and no gradient updates required. Our code is available at https://github.com/js-lee-AI/DART.
Sources
- Reasoning on a Budget: A Survey of Adaptive and Controllable Test-Time Compute in LLMs
- Program Synthesis with Large Language Models
- Language Models (Mostly) Know What They Know
- Evaluating Large Language Models Trained on Code
- Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
- SelfBudgeter: Adaptive Token Allocation for Efficient LLM Reasoning
- Do Thinking Tokens Help or Trap? Towards More Efficient Large Reasoning Model
- Adaptive Computation Time for Recurrent Neural Networks
- Hierarchical Budget Policy Optimization for Adaptive Reasoning
- Nemotron 3 Nano Omni: Efficient and Open Multimodal Intelligence
- Qwen2.5-Math Technical Report: Toward Mathematical Expert Model via Self-Improvement
- MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipe
- Route to Reason: Adaptive Routing for LLM and Reasoning Strategy Selection
- Qwen3 Technical Report
- Demystifying Hybrid Thinking: Can LLMs Truly Switch Between Think and No-Think?
Related papers
- MAVEN-T: Reinforced Heterogeneous Distillation for Real-Time Multi-Agent Trajectory Prediction
- Model Discovery Agent: LLM-assisted Bayesian experiment design for data-efficient discovery of mechanistic world models
- Agents' Last Exam
- The Clinician's Veto: Navigating Trust, Liability, and Uncertainty in Autonomous AI Prescribing
- MindHelper: Closed-Loop Embodied Mental-State Reasoning for Precision Intervention
- Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems