Work

Internal Chain-of-Thought: Empirical Evidence for Layer‐wise Subtask Scheduling in LLMs

Year: 2025

Type: conference-paper

Abstract: We show that large language models (LLMs) exhibit an internal chain-of-thought: they sequentially decompose and execute composite tasks layer-by-layer.Two claims ground our study: (i) distinct subtasks are learned at different network depths, and (ii) these subtasks are executed sequentially across layers.On a benchmark of 15 two-step composite tasks, we employ layer-from context-masking and propose a novel cross-task patching method, confirming (i).To examine claim (ii), we apply LogitLens to decode hidden states, revealing a consistent layerwise execution pattern.We further replicate our analysis on the real-world TRACE benchmark, observing the same stepwise dynamics.Together, our results enhance LLMs transparency by showing their capacity to internally plan and execute subtasks (or instructions), opening avenues for fine-grained, instruction-level activation steering.

Source: Conference on Empirical Methods in Natural Language Processing (EMNLP)

DOI: 10.18653/v1/2025.emnlp-main.1147

Authors: Zhipeng Yang, Junzhuo Li, Siyu Xia, Xuming Hu

Institutions: Hong Kong University of Science and Technology, The Hong Kong University of Science and Technology (Guangzhou), Southeast University

Language: English

Topic: Explainable Artificial Intelligence (XAI)

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: hybrid

Funders: Ant Group, National Natural Science Foundation of China, Hong Kong University of Science and Technology

OpenAlex ID: W4416035765