SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2305.14493 · 2023

Do prompt positions really matter?

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 11 functions out of this paper's own repositories and ran 8 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
milliemaoo/prompt-position canonical 8 of 11
FunctionStatusWhere it lives
mean Ran milliemaoo/prompt-position/lm_eval/metrics.py
code served (permissive licence) · get_code("0928f497e20fb443")
escaped_split Ran milliemaoo/prompt-position/lm_eval/utils.py
code served (permissive licence) · get_code("57e732507bad40db")
get_result Ran milliemaoo/prompt-position/lm_eval/models/gpt3.py
code served (permissive licence) · get_code("d253589f2c78dad0")
group Ran milliemaoo/prompt-position/lm_eval/utils.py
code served (permissive licence) · get_code("359c35fdcd7a0ee5")
hash_args Ran milliemaoo/prompt-position/lm_eval/base.py
code served (permissive licence) · get_code("ea06eaae4fc1eaf0")
pop_stddev Ran milliemaoo/prompt-position/lm_eval/metrics.py
code served (permissive licence) · get_code("518b2201c66c2ad3")
sample_stddev Ran milliemaoo/prompt-position/lm_eval/metrics.py
code served (permissive licence) · get_code("d40f0123e8a8f5c0")
simple_parse_args_string Ran milliemaoo/prompt-position/lm_eval/utils.py
code served (permissive licence) · get_code("cd999ad59caf84a7")
anthropic_completion Not yet run milliemaoo/prompt-position/lm_eval/models/anthropic_llms.py
code served (permissive licence) · get_code("a5cf72836c37574e")
make_table Not yet run milliemaoo/prompt-position/lm_eval/evaluator.py
code served (permissive licence) · get_code("9d95c14130837ee4")
stop_sequences_criteria Not yet run milliemaoo/prompt-position/lm_eval/models/huggingface.py
code served (permissive licence) · get_code("069ffbe9aadaa2ae")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Prompt-based models have gathered a lot of attention from researchers due to their remarkable advancements in the fields of zero-shot and few-shot learning. Developing an effective prompt template plays a critical role. However, prior studies have mainly focused on prompt vocabulary searching or embedding initialization within a predefined template with the prompt position fixed. In this empirical study, we conduct the most comprehensive analysis to date of prompt position for diverse Natural Language Processing (NLP) tasks. Our findings quantify the substantial impact prompt position has on model performance. We observe that the prompt positions used in prior studies are often sub-optimal, and this observation is consistent even in widely used instruction-tuned models. These findings suggest prompt position optimisation as a valuable research direction to augment prompt engineering methodologies and prompt position-aware instruction tuning as a potential way to build more robust models in the future.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2305.14493")
get_code_for_paper("2305.14493")
have("2305.14493")

Connect an agent — have() is free.