Revisiting Anthropomorphic Reflection Markers in Large Language Model Reasoning
This work clarifies the role of anthropomorphic markers in LLM reasoning for researchers studying model behavior and reasoning mechanisms, showing they are not reliable indicators of reflection.
LLMs often produce anthropomorphic reflection markers (e.g., 'wait', 'hmm') during reasoning, but their necessity is unclear. Suppressing these markers via prompt-level and token-level interventions preserves or improves performance on four benchmarks, especially with larger sampling budgets, indicating they are surface cues rather than essential for reflection.
Large Language Models (LLMs) often produce explicit reflective traces during complex reasoning, accompanied by anthropomorphic markers such as wait, hmm, and alternatively. Although these markers are commonly used as visible indicators of reflection, their mechanisms remain unclear, which leaves the risk of overthinking associated with redundant and repetitive reflection markers. In this work, we revisit anthropomorphic reflection markers, examining their necessity for reasoning and role in the reflection. We suppress these markers through prompt-level and token-level interventions, and analyze their effects on task performance across four benchmarks and two model scales. Our results show that anthropomorphic markers are not uniformly necessary for reasoning performance: suppressing them can preserve or improve performance in several settings, especially under larger sampling budgets. Meanwhile, marker suppression does not necessarily remove reflection behavior, as models can still perform marker-free verification. These suggest that anthropomorphic markers tend to be surface cues rather than reliable proxies for reflection itself, and motivate future research on reasoning mechanisms beyond explicit marker patterns.