Tips for the Few-Shot Setting:
Balance the examples across classes. For example, if you’re showing “good” and “bad” examples, include an equal number of good and bad examples.
Recency bias: shuffle the examples, so that e.g. all of the “bad” examples don’t appear at the end.
https://arxiv.org/abs/2102.09690
Balance the examples across classes. For example, if you’re showing “good” and “bad” examples, include an equal number of good and bad examples.
Recency bias: shuffle the examples, so that e.g. all of the “bad” examples don’t appear at the end.
https://arxiv.org/abs/2102.09690
Prompting tips:
(1) Use declarative and direct signifiers for tasks such as translate or rephrase this paragraph so that a 2nd grader can understand it.
(2) Use few-shot demonstrations when the task requires a bespoke format, recognizing that few-shot examples may be interpreted holistically by the model rather than as independent samples.
(3) Specify tasks using characters or characteristic situations as a proxy for an intention such as asking Gandhi or Nietzsche to solve a task. Here you are tapping into LLMs’ sophisticated understanding of analogies.
(4) Constrain the possible completion output using careful syntactic and lexical prompt formulations such as saying “Translate this French sentence to English” or by adding quotes around the French sentence.
(5) Encourage the model to break down problems into sub problems via step-by-step reasoning.
https://arxiv.org/abs/2102.07350
(1) Use declarative and direct signifiers for tasks such as translate or rephrase this paragraph so that a 2nd grader can understand it.
(2) Use few-shot demonstrations when the task requires a bespoke format, recognizing that few-shot examples may be interpreted holistically by the model rather than as independent samples.
(3) Specify tasks using characters or characteristic situations as a proxy for an intention such as asking Gandhi or Nietzsche to solve a task. Here you are tapping into LLMs’ sophisticated understanding of analogies.
(4) Constrain the possible completion output using careful syntactic and lexical prompt formulations such as saying “Translate this French sentence to English” or by adding quotes around the French sentence.
(5) Encourage the model to break down problems into sub problems via step-by-step reasoning.
https://arxiv.org/abs/2102.07350
👍1
Andrew Cantino’s prompt engineering tips:
(1) Make sure your inputs are grammatically correct and have good writing quality as LLMs tend to preserve stylistic consistency in their completions.
(2) Rather than generating a list of N items, generate a single item N times. This avoids the language model getting stuck in a repetitive loop.
(3) In order to improve output quality, generate many completions and then rank them heuristically.
https://blog.andrewcantino.com/blog/2021/04/21/prompt-engineering-tips-and-tricks/
(1) Make sure your inputs are grammatically correct and have good writing quality as LLMs tend to preserve stylistic consistency in their completions.
(2) Rather than generating a list of N items, generate a single item N times. This avoids the language model getting stuck in a repetitive loop.
(3) In order to improve output quality, generate many completions and then rank them heuristically.
https://blog.andrewcantino.com/blog/2021/04/21/prompt-engineering-tips-and-tricks/
❤1👍1
Mishra et al.’s prompt reframing techniques:
(1) Use low-level patterns from other examples to make a given prompt easier to understand for an LLM.
(2) Explictly itemize instructions into bulleted lists. Turn negative statements such as don’t create questions which are not to create questions which are.
(3) When possible, break down a top-level task into different sub-tasks that can be executed in parallel or sequentially.
(4) Avoid repeated and generic statements when trying to solve a very specific task. For example, instead of saying Answer the following question for a math problem, say Calculate answer to the following question. You need to either add or subtract numbers…
https://arxiv.org/abs/2109.07830
(1) Use low-level patterns from other examples to make a given prompt easier to understand for an LLM.
(2) Explictly itemize instructions into bulleted lists. Turn negative statements such as don’t create questions which are not to create questions which are.
(3) When possible, break down a top-level task into different sub-tasks that can be executed in parallel or sequentially.
(4) Avoid repeated and generic statements when trying to solve a very specific task. For example, instead of saying Answer the following question for a math problem, say Calculate answer to the following question. You need to either add or subtract numbers…
https://arxiv.org/abs/2109.07830
A scene from Curb Your Enthusiasm, in which Larry accidentally performs an LLM prompt injection attack, causing millions of unwanted packages to arrive at his LA home.
My company blocked chatgpt
Im a junior software engineer, in my team the seniors are allways occupied and they dont have time to explain so everytime im stuck chatgpt is my saviour. Today I arrived at my office and no one had access to chatgpt it was blocked by office wifi what should we do?
Im a junior software engineer, in my team the seniors are allways occupied and they dont have time to explain so everytime im stuck chatgpt is my saviour. Today I arrived at my office and no one had access to chatgpt it was blocked by office wifi what should we do?
👍2😁1