The Thinking Leakage in Reinforcement Learning for LLMs
cs.CL
Submitted: 2026-10-09
Updated: 2026-10-09
Terminology
Sources
- L1: Controlling How Long A Reasoning Model Thinks With Reinforcement Learning
- Llama-Nemotron: Efficient Reasoning Models
- Do NOT Think That Much for 2+3=? On the Overthinking of o1-Like LLMs
- The Entropy Mechanism of Reinforcement Learning for Reasoning Language Models
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- Thinkless: LLM Learns When to Think
- Concise Reasoning via Reinforcement Learning
- Finding Alignments Between Interpretable Causal Variables and Distributed Neural Representations
- Training Large Language Models to Reason in a Continuous Latent Space
- Think Only When You Need with Large Hybrid-Reasoning Models
- Tulu 3: Pushing Frontiers in Open Language Model Post-Training
- ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models
- Understanding R1-Zero-Like Training: A Critical Perspective
- Reasoning Models Can Be Effective Without Thinking
- s1: Simple test-time scaling
- Reinforcement Learning Finetunes Small Subnetworks in Large Language Models
- Proximal Policy Optimization Algorithms
- Spurious Rewards: Rethinking Training Signals in RLVR
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- RL's Razor: Why Online Reinforcement Learning Forgets Less
Related papers
- Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
- Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
- Subliminal Steering: Stronger Encoding of Hidden Signals
- MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
- The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
- Untangling the Mechanisms of Misleading Context in Medical Question Answering