-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathSELF-DISCOVER
23 lines (17 loc) · 2.01 KB
/
SELF-DISCOVER
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
From https://jrodthoughts.medium.com/meet-self-discover-google-deepminds-new-method-for-llm-reasoning-4f3fdc547926
The text provided is an article discussing a recent research paper by Google DeepMind on a novel reasoning technique called SELF-DISCOVER,
developed for large language models (LLMs). This method aims to enhance the problem-solving capabilities of LLMs by mimicking the human reasoning process.
DeepMind's approach to SELF-DISCOVER is inspired by human problem-solving strategies such as step-by-step problem-solving,
decomposition-based prompting, and step-back prompting.
It addresses the limitations of existing reasoning methods by tailoring the reasoning process to the specific demands of each task.
# SELF-DISCOVER operates in two fundamental stages:
Stage 1: In this stage, the LLM generates a task-specific reasoning structure, which is described using natural language terms
like "breakdown into subtasks" and "critical thinking." This process involves selecting, adapting, and implementing a coherent reasoning framework
using a structured data format (JSON) for interpretability and generation quality.
Stage 2: The LLM applies the reasoning structure generated in Stage 1 to solve individual task instances.
SELF-DISCOVER has shown promising results in enhancing the reasoning abilities of cutting-edge language models like PaLM 2-L and GPT-4 across various reasoning tasks,
including BBH, T4D, and MATH.
It significantly outperformed traditional methods such as chain-of-thought (CoT) and Plan-and-Solve (PS) in performance, demonstrating the effectiveness of its approach.
The article concludes that SELF-DISCOVER represents a significant step forward in the application of artificial intelligence to complex reasoning tasks.
By tailoring the reasoning process to the specific demands of each task and applying a structured reasoning process,
SELF-DISCOVER achieves higher accuracy and provides a more intuitive and logical path to solving problems, similar to the approach a human expert might take.