The mutation of these task prompts is governed by mutation prompts that the LLM generates and improves throughout evolution in a self-referential way. Promptbreeder is also improving the mutation prompts that improve these task prompts. /12
RESEARCH
-
PromptBreeder: LLM-Driven Self-Improvement Mechanism for Task Prompts
By
–
Another very recent paper introduces PromptBreeder, a general-purpose self-referential self-improvement mechanism that evolves and adapts prompts for a given domain. Driven by an LLM, Promptbreeder mutates a population of task prompts and subsequently evaluates them /11
-
Automated LLM Improvement Through Quality Gap Maximization
By
–
Instead of merely maximizing the response quality for a given input, the focus is shifted to maximizing the quality gap of the response, thus promoting self-improvement. This method takes human preferences and creates an automated way to improve LLMs over time. /10
-
PIT Framework: Learning from Human Preferences and RLHF Reformulation
By
–
The PIT framework focuses on learning from human preference data, coupled with its unique reformulation of the RLHF objective. Humans indicate their preferences on LLM outputs and this data is used to train reward models. The RLHF objective is reformulated. /9
-
PIT Framework Enables LLMs Self-Improvement From Data
By
–
A more recent paper titled "Enable Language Models to Implicitly Learn Self-Improvement From Data" dives into an inventive framework known as ImPlicit Self-ImprovemenT (PIT), aimed at facilitating self-improvement in Large Language Models (LLMs) from data. /8
-
LLM Self-Improvement Through Fine-Tuning Reduces Supervision Needs
By
–
Having said that, the self-improvement of LLMs using fine-tuning can lead to better performance in various NLP tasks with less reliance on extensive supervision, which is a critical concern in the current state of LLM training. /7
-
LLM Learning Breakthrough: From Fine-tuning to On-the-Fly Adaptation
By
–
In some ways, the method is akin to teaching an LLM how to respond using CoT prompting and then feeding the answers as part of the fine-tuning process. The next big significant break through is when we invent a way for LLMs to learn on-the-fly. /6
-
LLM Fine-Tuning with Self-Generated Solutions Improves Reasoning
By
–
The LLM was then fine-tuned using these self-generated solutions as target outputs, which resulted in improved general reasoning ability of a 540B-parameter LLM across multiple benchmarks without requiring any ground truth labels /5
-
LLMs Self-Improve Using Unlabeled Data Without Fine-Tuning
By
–
Unlike humans, who are constantly learning and absorbing new things in real-time, LLMs need to be fine-tuned or prompted to learn something new. The paper "Large Language Models Can Self-Improve" demonstrates how LLM can self-improve with only unlabeled datasets. /3
-
Pre-trained LLM generates high-confidence rationale-augmented answers
By
–
They used a pre-trained LLM to generate "high-confidence" rationale-augmented answers for unlabeled questions using Chain-of-Thought prompting and self-consistency. /4