Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Self-controller: Controlling LLMs with Multi-round Step-by-step Self-awareness

Xiao Peng, Xufan Geng

TL;DR

The Self-controller is proposed, a novel agentic framework bringing self-awareness into LLMs' reasoning logic, letting the LLM become self-aware of current status and think step by step in a multi-round chain-of-thought paradigm.

Abstract

The applications of large language models (LLMs) have been widely spread across all domains. However, the basic abilities such as the controllability of LLMs are still limited. To address this, we propose "Self-controller", a novel agentic framework bringing self-awareness into LLMs' reasoning logic. The core idea of this work is to maintain states based on the LLM's response, letting the LLM become self-aware of current status and think step by step in a multi-round chain-of-thought paradigm. Our experiment on the state of textual length has shown the controllability and effectiveness of the Self-controller. We further implement a binary search algorithm to accelerate the generation process based on the linearity and monotonicity of the textual length state. Another advantage of the Self-controller comes with DeepSeek's Context Caching technology, which significantly saves computational token consumption when a cluster of conversations shares the same prefix of context. Theoretically, we prove that in this scenario the extra time complexity is $O(c \log n)$. Results of the back-of-the-envelope estimation suggest that the token consumption of our method is no more than twice as much as that of the trivial single-round generation. Furthermore, our ablation study on word constraints demonstrates the Self-controller's consistent controllability across all foundation models.

Self-controller: Controlling LLMs with Multi-round Step-by-step Self-awareness

TL;DR

The Self-controller is proposed, a novel agentic framework bringing self-awareness into LLMs' reasoning logic, letting the LLM become self-aware of current status and think step by step in a multi-round chain-of-thought paradigm.

Abstract

The applications of large language models (LLMs) have been widely spread across all domains. However, the basic abilities such as the controllability of LLMs are still limited. To address this, we propose "Self-controller", a novel agentic framework bringing self-awareness into LLMs' reasoning logic. The core idea of this work is to maintain states based on the LLM's response, letting the LLM become self-aware of current status and think step by step in a multi-round chain-of-thought paradigm. Our experiment on the state of textual length has shown the controllability and effectiveness of the Self-controller. We further implement a binary search algorithm to accelerate the generation process based on the linearity and monotonicity of the textual length state. Another advantage of the Self-controller comes with DeepSeek's Context Caching technology, which significantly saves computational token consumption when a cluster of conversations shares the same prefix of context. Theoretically, we prove that in this scenario the extra time complexity is . Results of the back-of-the-envelope estimation suggest that the token consumption of our method is no more than twice as much as that of the trivial single-round generation. Furthermore, our ablation study on word constraints demonstrates the Self-controller's consistent controllability across all foundation models.
Paper Structure (20 sections, 5 equations, 6 figures, 2 tables, 1 algorithm)

This paper contains 20 sections, 5 equations, 6 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Works
  3. Large Language Models
  4. Self-awareness
  5. Controllability
  6. Methodology
  7. Overview of the Self-controller
  8. Binary Search on the Textual Length State
  9. Context Caching in Multi-round Sessions
  10. Experiments
  11. Length Control
  12. Binary Search on Length Control
  13. Ablation Study
  14. Efficiency
  15. Word Constraint
  16. ...and 5 more sections

Figures (6)

  • Figure 1: A simplified demonstration of the Self-controller summarizing for "Oedipus Rex"
  • Figure 2: Demonstration of the Self-controller workflow
  • Figure 3: The consumption of 3 different methods: single-round, multi-round, and binary search
  • Figure 4: Average BERTSCORE evaluation across different datasets and LLMs (word constraint = 250)
  • Figure 5: Efficiency study for the binary search multi-round generation method on the tldr dataset
  • ...and 1 more figures