When Should LLMs Reason? Adaptive Inference Routing via Entropy Dynamics
cs.LG, cs.AI, cs.CL
Submitted: 2026-05-20
Updated: 2026-10-22
Terminology
Sources
- Think Deep, Not Just Long: Measuring LLM Reasoning Effort via Deep-Thinking Tokens
- Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
- Training Verifiers to Solve Math Word Problems
- BEST-Route: Adaptive LLM Routing with Test-Time Optimal Compute
- Disentangling Recall and Reasoning in Transformer Models through Layer-wise Attention and Activation Analysis
- Mixture of Thoughts: Learning to Aggregate What Experts Think, Not Just What They Say
- RADAR: Reasoning-Ability and Difficulty-Aware Routing for Reasoning LLMs
- Rethinking External Slow-Thinking: From Snowball Errors to Probability of Correct Reasoning
- The Llama 3 Herd of Models
- Think Twice Before You Write -- an Entropy-based Decoding Strategy to Enhance LLM Reasoning
- EvolveRouter: Co-Evolving Routing and Prompt for Multi-Agent Question Answering
- Qwen2.5-Coder Technical Report
- Think Only When You Need with Large Hybrid-Reasoning Models
- Token Signature: Predicting Chain-of-Thought Gains with Token Decoding Feature in Large Language Models
- Mind Your Step (by Step): Chain-of-Thought can Reduce Performance on Tasks where Thinking Makes Humans Worse
- Plan before Solving: Problem-Aware Strategy Routing for Mathematical Reasoning with LLMs
- GPQA: A Graduate-Level Google-Proof Q&A Benchmark
- Solving General Arithmetic Word Problems
- Route-and-Reason: Scaling Large Language Model Reasoning with Reinforced Model Router
- Think Just Enough: Sequence-Level Entropy as a Confidence Signal for LLM Reasoning
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: The Predictive Cost of ELBO Binding
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks