</think> Doesn't Stop Reasoning: Analysis of Spurious CoT Termination
cs.CL, cs.AI
Submitted: 2026-09-03
Updated: 2026-09-03
Comments: Accepted to EMNLP 2026 Main Conference
Code: https://github.com/Seunghee-Koh/Spurious-CoT-Termination
License: http://creativecommons.org/licenses/by/4.0/
The gist: Chain-of-thought (CoT) reasoning improves large reasoning models (LRMs) on complex tasks but often produces long, redundant traces.
Terminology
Abstract
Chain-of-thought (CoT) reasoning improves large reasoning models (LRMs) on complex tasks but often produces long, redundant traces. Recent training-free early-exit methods shorten these traces by choosing an intermediate point to stop reasoning. We study one such strategy that injects an end-of-think token (EoT, </think>) at this point to trigger the reasoning-to-answering transition, and find that the injected EoT does not always induce a clean answering phase. Answering-phase generation can continue before the model regenerates another EoT, with the span preceding this regenerated EoT scaling with the reasoning tokens saved by early exit and exhibiting continued reasoning behavior. We call this spurious CoT termination, where reasoning-like generation continues into the answering phase. We hypothesize that insufficient attention to the injected EoT contributes to spurious CoT termination and probe this hypothesis with Exit-token Attention Biasing (EAB). Across four LRMs, five benchmarks, and two early-exit methods, increasing attention to the injected EoT reduces spurious CoT termination and answering-phase length. These results reveal a limitation of controlling LRMs by externally matching their explicit think-block format. Inserting the EoT token conforms to this format but does not by itself guarantee the intended reasoning-to-answering transition. Our code is available at https://github.com/Seunghee-Koh/Spurious-CoT-Termination.
Sources
- Thought Anchors: Which LLM Reasoning Steps Matter?
- Training Verifiers to Solve Math Word Problems
- DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
- How Well do LLMs Compress Their Own Chain-of-Thought? A Token Complexity Approach
- Gradient-Stable Attention Heads Signal LLM Correctness
- Thinking Sparks!: Emergent Attention Heads in Reasoning Models During Post Training
- Stop Overthinking: A Survey on Efficient Reasoning for Large Language Models
- Kimi k1.5: Scaling Reinforcement Learning with LLMs
- Harnessing the Reasoning Economy: A Survey of Efficient Reasoning for Large Language Models
- Chain of Draft: Thinking Faster by Writing Less
- Qwen3 Technical Report
- Distilling System 2 into System 1
- When Can Large Reasoning Models Save Thinking? Mechanistic Analysis of Behavioral Divergence in Reasoning
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